CNN_LSTM_stock_predict.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Copy of CNN_LSTM_stock_predict.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AXDHVzB0aQS2",
        "outputId": "d9a4d191-0f74-426e-e17d-7256072a27c6"
      },
      "source": [
        "from google.colab import drive\r\n",
        "drive.mount('/content/drive', force_remount=True)"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2GaUFq8HyVcv"
      },
      "source": [
        "# pip install shap"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lCHFRoO3aRy0"
      },
      "source": [
        "import os\r\n",
        "\r\n",
        "import torch\r\n",
        "import numpy as np\r\n",
        "import pandas as pd\r\n",
        "import seaborn as sns\r\n",
        "import matplotlib.pyplot as plt\r\n",
        "from torch import nn, optim\r\n",
        "from torch.utils.data import Dataset, DataLoader,TensorDataset\r\n",
        "import torch.nn.functional as F\r\n",
        "# from torch.autograd import Variable\r\n",
        "import torchvision.models as models  \r\n",
        "\r\n",
        "import warnings# 移除不必要的警告\r\n",
        "warnings.filterwarnings('ignore')\r\n",
        "from sklearn.preprocessing import MinMaxScaler\r\n",
        "from xgboost import XGBRegressor\r\n",
        "from sklearn.model_selection import RandomizedSearchCV\r\n",
        "from scipy.stats import randint as sp_randint\r\n",
        "from xgboost import plot_importance\r\n",
        "# import shap"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GOk45xnmq4p-"
      },
      "source": [
        "import matplotlib as mpl\r\n",
        "zhfont = mpl.font_manager.FontProperties(fname='/content/drive/MyDrive/chinese.simhei.ttf')#取得中文字型\r\n",
        "\r\n",
        "sns.set(font=zhfont.get_name())\r\n"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 505
        },
        "id": "C3p2LG0fackv",
        "outputId": "1db7e60d-af32-45ae-dc70-86796ad88e23"
      },
      "source": [
        "df = pd.read_csv('/content/drive/MyDrive/preprocess(1).csv')\r\n",
        "\r\n",
        "df.set_index(keys =df.columns[0],inplace=True)\r\n",
        "df['y_class']=(df.y)>1\r\n",
        "df.head()"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>('多方交易口數', '合計         ')</th>\n",
              "      <th>('多方交易口數', '外資         ')</th>\n",
              "      <th>('多方交易口數', '投信         ')</th>\n",
              "      <th>('多方交易口數', '自營商       ')</th>\n",
              "      <th>('多方交易契約金額', '合計         ')</th>\n",
              "      <th>('多方交易契約金額', '外資         ')</th>\n",
              "      <th>('多方交易契約金額', '投信         ')</th>\n",
              "      <th>('多方交易契約金額', '自營商       ')</th>\n",
              "      <th>('多方未平倉口數', '合計         ')</th>\n",
              "      <th>('多方未平倉口數', '外資         ')</th>\n",
              "      <th>('多方未平倉口數', '投信         ')</th>\n",
              "      <th>('多方未平倉口數', '自營商       ')</th>\n",
              "      <th>('多方未平倉契約金額', '合計         ')</th>\n",
              "      <th>('多方未平倉契約金額', '外資         ')</th>\n",
              "      <th>('多方未平倉契約金額', '投信         ')</th>\n",
              "      <th>('多方未平倉契約金額', '自營商       ')</th>\n",
              "      <th>('多空交易口數淨額', '合計         ')</th>\n",
              "      <th>('多空交易口數淨額', '外資         ')</th>\n",
              "      <th>('多空交易口數淨額', '投信         ')</th>\n",
              "      <th>('多空交易口數淨額', '自營商       ')</th>\n",
              "      <th>('多空交易契約金額淨額', '合計         ')</th>\n",
              "      <th>('多空交易契約金額淨額', '外資         ')</th>\n",
              "      <th>('多空交易契約金額淨額', '投信         ')</th>\n",
              "      <th>('多空交易契約金額淨額', '自營商       ')</th>\n",
              "      <th>('多空未平倉口數淨額', '合計         ')</th>\n",
              "      <th>('多空未平倉口數淨額', '外資         ')</th>\n",
              "      <th>('多空未平倉口數淨額', '投信         ')</th>\n",
              "      <th>('多空未平倉口數淨額', '自營商       ')</th>\n",
              "      <th>('多空未平倉契約淨額', '合計         ')</th>\n",
              "      <th>('多空未平倉契約淨額', '外資         ')</th>\n",
              "      <th>('多空未平倉契約淨額', '投信         ')</th>\n",
              "      <th>('多空未平倉契約淨額', '自營商       ')</th>\n",
              "      <th>('空方交易口數', '合計         ')</th>\n",
              "      <th>('空方交易口數', '外資         ')</th>\n",
              "      <th>('空方交易口數', '投信         ')</th>\n",
              "      <th>('空方交易口數', '自營商       ')</th>\n",
              "      <th>('空方交易契約金額', '合計         ')</th>\n",
              "      <th>('空方交易契約金額', '外資         ')</th>\n",
              "      <th>('空方交易契約金額', '投信         ')</th>\n",
              "      <th>('空方交易契約金額', '自營商       ')</th>\n",
              "      <th>...</th>\n",
              "      <th>('成交量(千股)', '紡纖         ')_div</th>\n",
              "      <th>('成交量(千股)', '紡纖         ')_div(%)</th>\n",
              "      <th>('成交量(千股)', '造紙類       ')_div</th>\n",
              "      <th>('成交量(千股)', '造紙類       ')_div(%)</th>\n",
              "      <th>('成交量(千股)', '金融類       ')_div</th>\n",
              "      <th>('成交量(千股)', '金融類       ')_div(%)</th>\n",
              "      <th>('成交量(千股)', '食品類       ')_div</th>\n",
              "      <th>('成交量(千股)', '食品類       ')_div(%)</th>\n",
              "      <th>('收盤價', '#DX')_ret</th>\n",
              "      <th>('收盤價', '#GSPC')_ret</th>\n",
              "      <th>('收盤價', '#IXIC')_ret</th>\n",
              "      <th>('收盤價', '#IXID')_ret</th>\n",
              "      <th>('收盤價', '#IXK')_ret</th>\n",
              "      <th>('收盤價', '#NDX')_ret</th>\n",
              "      <th>('收盤價', '#S5CD')_ret</th>\n",
              "      <th>('收盤價', '#S5CS')_ret</th>\n",
              "      <th>('收盤價', '#S5INDU')_ret</th>\n",
              "      <th>('收盤價', '#S5INFO')_ret</th>\n",
              "      <th>('收盤價', '#SOXX')_ret</th>\n",
              "      <th>('收盤價', '#VIX')_ret</th>\n",
              "      <th>('收盤價', '#VXN')_ret</th>\n",
              "      <th>收盤價(元)_x_ret</th>\n",
              "      <th>收盤價(元)_y_ret</th>\n",
              "      <th>('收盤價(元)', 'TR未金電     ')_ret</th>\n",
              "      <th>('收盤價(元)', '不含金融     ')_ret</th>\n",
              "      <th>('收盤價(元)', '不含電子     ')_ret</th>\n",
              "      <th>('收盤價(元)', '不含電子報酬 ')_ret</th>\n",
              "      <th>('收盤價(元)', '加權指數     ')_ret</th>\n",
              "      <th>('收盤價(元)', '報酬指數     ')_ret</th>\n",
              "      <th>('收盤價(元)', '塑膠化工     ')_ret</th>\n",
              "      <th>('收盤價(元)', '未含金電     ')_ret</th>\n",
              "      <th>('收盤價(元)', '機電類       ')_ret</th>\n",
              "      <th>('收盤價(元)', '水泥窯業     ')_ret</th>\n",
              "      <th>('收盤價(元)', '營建         ')_ret</th>\n",
              "      <th>('收盤價(元)', '紡纖         ')_ret</th>\n",
              "      <th>('收盤價(元)', '造紙類       ')_ret</th>\n",
              "      <th>('收盤價(元)', '金融類       ')_ret</th>\n",
              "      <th>('收盤價(元)', '食品類       ')_ret</th>\n",
              "      <th>y</th>\n",
              "      <th>y_class</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>年月日</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>2007-07-03</th>\n",
              "      <td>395169</td>\n",
              "      <td>22221</td>\n",
              "      <td>792</td>\n",
              "      <td>372156</td>\n",
              "      <td>31680000</td>\n",
              "      <td>7965000</td>\n",
              "      <td>1058000</td>\n",
              "      <td>22657000</td>\n",
              "      <td>592294</td>\n",
              "      <td>246553</td>\n",
              "      <td>1021</td>\n",
              "      <td>344720</td>\n",
              "      <td>94717000</td>\n",
              "      <td>40824000</td>\n",
              "      <td>1152000</td>\n",
              "      <td>52741000</td>\n",
              "      <td>4332</td>\n",
              "      <td>-2854</td>\n",
              "      <td>-483</td>\n",
              "      <td>7669</td>\n",
              "      <td>9437000</td>\n",
              "      <td>-848000</td>\n",
              "      <td>919000</td>\n",
              "      <td>9366000</td>\n",
              "      <td>-152605</td>\n",
              "      <td>-5762</td>\n",
              "      <td>-10747</td>\n",
              "      <td>-136096</td>\n",
              "      <td>9073000</td>\n",
              "      <td>-18678000</td>\n",
              "      <td>-10345000</td>\n",
              "      <td>38096000</td>\n",
              "      <td>390837</td>\n",
              "      <td>25075</td>\n",
              "      <td>1275</td>\n",
              "      <td>364487</td>\n",
              "      <td>22243000</td>\n",
              "      <td>8813000</td>\n",
              "      <td>139000</td>\n",
              "      <td>13291000</td>\n",
              "      <td>...</td>\n",
              "      <td>72685.0</td>\n",
              "      <td>0.438742</td>\n",
              "      <td>-17422.0</td>\n",
              "      <td>-0.291080</td>\n",
              "      <td>201916.0</td>\n",
              "      <td>0.569720</td>\n",
              "      <td>39272.0</td>\n",
              "      <td>0.379932</td>\n",
              "      <td>0.000614</td>\n",
              "      <td>0.003580</td>\n",
              "      <td>0.004806</td>\n",
              "      <td>0.001489</td>\n",
              "      <td>0.008824</td>\n",
              "      <td>0.005670</td>\n",
              "      <td>-0.000096</td>\n",
              "      <td>0.000071</td>\n",
              "      <td>0.003324</td>\n",
              "      <td>0.007358</td>\n",
              "      <td>0.000198</td>\n",
              "      <td>-0.007143</td>\n",
              "      <td>-0.017647</td>\n",
              "      <td>0.002815</td>\n",
              "      <td>0.009848</td>\n",
              "      <td>0.009797</td>\n",
              "      <td>0.005901</td>\n",
              "      <td>0.006356</td>\n",
              "      <td>0.009656</td>\n",
              "      <td>0.006344</td>\n",
              "      <td>0.007783</td>\n",
              "      <td>-0.002419</td>\n",
              "      <td>0.005062</td>\n",
              "      <td>0.006732</td>\n",
              "      <td>-0.001782</td>\n",
              "      <td>0.005712</td>\n",
              "      <td>0.014492</td>\n",
              "      <td>0.000207</td>\n",
              "      <td>0.009080</td>\n",
              "      <td>0.030225</td>\n",
              "      <td>0.008198</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2007-07-04</th>\n",
              "      <td>516232</td>\n",
              "      <td>47360</td>\n",
              "      <td>327</td>\n",
              "      <td>468545</td>\n",
              "      <td>44422000</td>\n",
              "      <td>14649000</td>\n",
              "      <td>375000</td>\n",
              "      <td>29398000</td>\n",
              "      <td>641980</td>\n",
              "      <td>256255</td>\n",
              "      <td>971</td>\n",
              "      <td>384754</td>\n",
              "      <td>105854000</td>\n",
              "      <td>44978000</td>\n",
              "      <td>1146000</td>\n",
              "      <td>59730000</td>\n",
              "      <td>-1182</td>\n",
              "      <td>5235</td>\n",
              "      <td>-12</td>\n",
              "      <td>-6405</td>\n",
              "      <td>2034000</td>\n",
              "      <td>-4388000</td>\n",
              "      <td>291000</td>\n",
              "      <td>6131000</td>\n",
              "      <td>-153798</td>\n",
              "      <td>-527</td>\n",
              "      <td>-10776</td>\n",
              "      <td>-142495</td>\n",
              "      <td>11205000</td>\n",
              "      <td>-22724000</td>\n",
              "      <td>-10180000</td>\n",
              "      <td>44109000</td>\n",
              "      <td>517414</td>\n",
              "      <td>42125</td>\n",
              "      <td>339</td>\n",
              "      <td>474950</td>\n",
              "      <td>42388000</td>\n",
              "      <td>19037000</td>\n",
              "      <td>84000</td>\n",
              "      <td>23267000</td>\n",
              "      <td>...</td>\n",
              "      <td>230061.0</td>\n",
              "      <td>0.965215</td>\n",
              "      <td>21335.0</td>\n",
              "      <td>0.502816</td>\n",
              "      <td>16821.0</td>\n",
              "      <td>0.030236</td>\n",
              "      <td>-1095.0</td>\n",
              "      <td>-0.007677</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.009832</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.011282</td>\n",
              "      <td>0.008481</td>\n",
              "      <td>0.009355</td>\n",
              "      <td>0.009593</td>\n",
              "      <td>0.008090</td>\n",
              "      <td>0.008198</td>\n",
              "      <td>0.008349</td>\n",
              "      <td>0.011106</td>\n",
              "      <td>0.007285</td>\n",
              "      <td>0.006287</td>\n",
              "      <td>0.036535</td>\n",
              "      <td>0.025500</td>\n",
              "      <td>0.011270</td>\n",
              "      <td>0.005686</td>\n",
              "      <td>0.017756</td>\n",
              "      <td>0.009611</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2007-07-05</th>\n",
              "      <td>340678</td>\n",
              "      <td>25268</td>\n",
              "      <td>709</td>\n",
              "      <td>314701</td>\n",
              "      <td>23970000</td>\n",
              "      <td>6345000</td>\n",
              "      <td>496000</td>\n",
              "      <td>17129000</td>\n",
              "      <td>664970</td>\n",
              "      <td>265112</td>\n",
              "      <td>893</td>\n",
              "      <td>398965</td>\n",
              "      <td>110191000</td>\n",
              "      <td>46534000</td>\n",
              "      <td>1103000</td>\n",
              "      <td>62554000</td>\n",
              "      <td>19741</td>\n",
              "      <td>1124</td>\n",
              "      <td>67</td>\n",
              "      <td>18550</td>\n",
              "      <td>-689000</td>\n",
              "      <td>-3278000</td>\n",
              "      <td>118000</td>\n",
              "      <td>2471000</td>\n",
              "      <td>-134201</td>\n",
              "      <td>597</td>\n",
              "      <td>-10709</td>\n",
              "      <td>-124089</td>\n",
              "      <td>10649000</td>\n",
              "      <td>-25718000</td>\n",
              "      <td>-10142000</td>\n",
              "      <td>46509000</td>\n",
              "      <td>320937</td>\n",
              "      <td>24144</td>\n",
              "      <td>642</td>\n",
              "      <td>296151</td>\n",
              "      <td>24659000</td>\n",
              "      <td>9623000</td>\n",
              "      <td>378000</td>\n",
              "      <td>14658000</td>\n",
              "      <td>...</td>\n",
              "      <td>-53324.0</td>\n",
              "      <td>-0.113840</td>\n",
              "      <td>-16402.0</td>\n",
              "      <td>-0.257222</td>\n",
              "      <td>-84249.0</td>\n",
              "      <td>-0.146993</td>\n",
              "      <td>-11431.0</td>\n",
              "      <td>-0.080760</td>\n",
              "      <td>0.001228</td>\n",
              "      <td>0.000348</td>\n",
              "      <td>0.004424</td>\n",
              "      <td>0.003026</td>\n",
              "      <td>0.008065</td>\n",
              "      <td>0.008070</td>\n",
              "      <td>0.006559</td>\n",
              "      <td>-0.000962</td>\n",
              "      <td>-0.000251</td>\n",
              "      <td>0.007025</td>\n",
              "      <td>0.007616</td>\n",
              "      <td>0.012426</td>\n",
              "      <td>0.023353</td>\n",
              "      <td>-0.006954</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.005634</td>\n",
              "      <td>0.009387</td>\n",
              "      <td>0.005130</td>\n",
              "      <td>0.006854</td>\n",
              "      <td>0.008799</td>\n",
              "      <td>0.009611</td>\n",
              "      <td>0.008598</td>\n",
              "      <td>0.005116</td>\n",
              "      <td>0.011422</td>\n",
              "      <td>-0.013147</td>\n",
              "      <td>0.003397</td>\n",
              "      <td>0.031191</td>\n",
              "      <td>-0.005317</td>\n",
              "      <td>0.005150</td>\n",
              "      <td>0.006243</td>\n",
              "      <td>0.004402</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2007-07-06</th>\n",
              "      <td>330124</td>\n",
              "      <td>16264</td>\n",
              "      <td>245</td>\n",
              "      <td>313615</td>\n",
              "      <td>30868000</td>\n",
              "      <td>8393000</td>\n",
              "      <td>269000</td>\n",
              "      <td>22206000</td>\n",
              "      <td>672632</td>\n",
              "      <td>267371</td>\n",
              "      <td>895</td>\n",
              "      <td>404366</td>\n",
              "      <td>110383000</td>\n",
              "      <td>48298000</td>\n",
              "      <td>1112000</td>\n",
              "      <td>60973000</td>\n",
              "      <td>1855</td>\n",
              "      <td>-7884</td>\n",
              "      <td>-59</td>\n",
              "      <td>9798</td>\n",
              "      <td>-1982000</td>\n",
              "      <td>-1934000</td>\n",
              "      <td>107000</td>\n",
              "      <td>-155000</td>\n",
              "      <td>-132344</td>\n",
              "      <td>-7279</td>\n",
              "      <td>-10768</td>\n",
              "      <td>-114297</td>\n",
              "      <td>8663000</td>\n",
              "      <td>-27618000</td>\n",
              "      <td>-10058000</td>\n",
              "      <td>46339000</td>\n",
              "      <td>328269</td>\n",
              "      <td>24148</td>\n",
              "      <td>304</td>\n",
              "      <td>303817</td>\n",
              "      <td>32851000</td>\n",
              "      <td>10328000</td>\n",
              "      <td>162000</td>\n",
              "      <td>22361000</td>\n",
              "      <td>...</td>\n",
              "      <td>-112612.0</td>\n",
              "      <td>-0.271296</td>\n",
              "      <td>-7125.0</td>\n",
              "      <td>-0.150431</td>\n",
              "      <td>226810.0</td>\n",
              "      <td>0.463918</td>\n",
              "      <td>-4053.0</td>\n",
              "      <td>-0.031150</td>\n",
              "      <td>-0.000858</td>\n",
              "      <td>0.000315</td>\n",
              "      <td>0.003711</td>\n",
              "      <td>0.008356</td>\n",
              "      <td>0.001488</td>\n",
              "      <td>0.003660</td>\n",
              "      <td>0.007915</td>\n",
              "      <td>0.002069</td>\n",
              "      <td>0.003592</td>\n",
              "      <td>0.003551</td>\n",
              "      <td>0.010386</td>\n",
              "      <td>-0.049096</td>\n",
              "      <td>-0.033353</td>\n",
              "      <td>-0.012606</td>\n",
              "      <td>0.007979</td>\n",
              "      <td>-0.001133</td>\n",
              "      <td>0.004374</td>\n",
              "      <td>0.000452</td>\n",
              "      <td>0.000516</td>\n",
              "      <td>0.004321</td>\n",
              "      <td>0.004402</td>\n",
              "      <td>-0.010621</td>\n",
              "      <td>-0.001228</td>\n",
              "      <td>0.006992</td>\n",
              "      <td>0.015539</td>\n",
              "      <td>0.003289</td>\n",
              "      <td>0.028047</td>\n",
              "      <td>0.001696</td>\n",
              "      <td>0.004000</td>\n",
              "      <td>-0.000532</td>\n",
              "      <td>0.019827</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2007-07-09</th>\n",
              "      <td>496122</td>\n",
              "      <td>31688</td>\n",
              "      <td>841</td>\n",
              "      <td>463593</td>\n",
              "      <td>40214000</td>\n",
              "      <td>9339000</td>\n",
              "      <td>716000</td>\n",
              "      <td>30159000</td>\n",
              "      <td>710388</td>\n",
              "      <td>269692</td>\n",
              "      <td>936</td>\n",
              "      <td>439760</td>\n",
              "      <td>122052000</td>\n",
              "      <td>53652000</td>\n",
              "      <td>1201000</td>\n",
              "      <td>67199000</td>\n",
              "      <td>45712</td>\n",
              "      <td>-17866</td>\n",
              "      <td>252</td>\n",
              "      <td>63326</td>\n",
              "      <td>5741000</td>\n",
              "      <td>1655000</td>\n",
              "      <td>107000</td>\n",
              "      <td>3979000</td>\n",
              "      <td>-86652</td>\n",
              "      <td>-25135</td>\n",
              "      <td>-10516</td>\n",
              "      <td>-51001</td>\n",
              "      <td>14592000</td>\n",
              "      <td>-25032000</td>\n",
              "      <td>-10227000</td>\n",
              "      <td>49851000</td>\n",
              "      <td>450410</td>\n",
              "      <td>49554</td>\n",
              "      <td>589</td>\n",
              "      <td>400267</td>\n",
              "      <td>34473000</td>\n",
              "      <td>7684000</td>\n",
              "      <td>609000</td>\n",
              "      <td>26180000</td>\n",
              "      <td>...</td>\n",
              "      <td>-79578.0</td>\n",
              "      <td>-0.263088</td>\n",
              "      <td>52903.0</td>\n",
              "      <td>1.314720</td>\n",
              "      <td>187845.0</td>\n",
              "      <td>0.262459</td>\n",
              "      <td>-39960.0</td>\n",
              "      <td>-0.316994</td>\n",
              "      <td>-0.000491</td>\n",
              "      <td>0.003912</td>\n",
              "      <td>0.001316</td>\n",
              "      <td>0.002214</td>\n",
              "      <td>0.001866</td>\n",
              "      <td>0.000448</td>\n",
              "      <td>-0.003658</td>\n",
              "      <td>0.001317</td>\n",
              "      <td>0.006243</td>\n",
              "      <td>-0.000276</td>\n",
              "      <td>0.011425</td>\n",
              "      <td>0.029891</td>\n",
              "      <td>0.033898</td>\n",
              "      <td>0.021278</td>\n",
              "      <td>0.021108</td>\n",
              "      <td>0.007703</td>\n",
              "      <td>0.017162</td>\n",
              "      <td>0.016760</td>\n",
              "      <td>0.016828</td>\n",
              "      <td>0.019757</td>\n",
              "      <td>0.019827</td>\n",
              "      <td>0.005804</td>\n",
              "      <td>0.007625</td>\n",
              "      <td>0.021666</td>\n",
              "      <td>0.002380</td>\n",
              "      <td>0.011682</td>\n",
              "      <td>-0.019448</td>\n",
              "      <td>0.022986</td>\n",
              "      <td>0.035889</td>\n",
              "      <td>-0.003512</td>\n",
              "      <td>0.002296</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 223 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "            ('多方交易口數', '合計         ')  ...  y_class\n",
              "年月日                                    ...         \n",
              "2007-07-03                     395169  ...    False\n",
              "2007-07-04                     516232  ...    False\n",
              "2007-07-05                     340678  ...    False\n",
              "2007-07-06                     330124  ...    False\n",
              "2007-07-09                     496122  ...    False\n",
              "\n",
              "[5 rows x 223 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XklbFNySaeGL",
        "outputId": "2986bb95-a204-41be-c532-edc2e9663b17"
      },
      "source": [
        "df.index=pd.to_datetime(df.index,format=\"%Y/%m/%d\")\r\n",
        "df.index.name='date'\r\n",
        "feature=df.columns[:-2]\r\n",
        "print(len(feature))\r\n",
        "\r\n",
        "feature_len=len(feature)"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "221\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mb0-aeT34FuL"
      },
      "source": [
        "## **`以過去20天預測第21天股價`**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zUrkRA--cq_p"
      },
      "source": [
        "DAYS_FOR_TRAIN=20\r\n",
        "def deal_lstm(df,feature,n = DAYS_FOR_TRAIN):#以20天作為預測\r\n",
        "    \r\n",
        "    X = []\r\n",
        "    y = []\r\n",
        "    dataset_scaled_x = df[feature]\r\n",
        "    \r\n",
        "    for i in range(0, len(df)-n):\r\n",
        "        X.append(dataset_scaled_x[i:i+n])\r\n",
        "        y.append(df['y'].iloc[i+n-1])\r\n",
        "    \r\n",
        "    return (np.array(X),np.array(y))"
      ],
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VQWbOjpK4O7a"
      },
      "source": [
        "## 標準化(`MinMaxScaler`)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y3pOLSTUdkf0"
      },
      "source": [
        "def tol_normalize(ddf):\r\n",
        "\r\n",
        "  for col in feature:\r\n",
        "    ddf[col]=MinMaxScaler().fit_transform(ddf[col].values.reshape(-1, 1))\r\n",
        "\r\n",
        "  # ddf=MinMaxScaler().fit_transform(ddf)\r\n",
        "  return ddf"
      ],
      "execution_count": 30,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z9v5SJRDmI51"
      },
      "source": [
        "**Feature** **select(以xgboost 挑選最重要20特徵)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 306
        },
        "id": "Q-k8tbCrmR_t",
        "outputId": "3ab72902-1c87-4b44-95ba-99b17f5f51dd"
      },
      "source": [
        "xgb = XGBRegressor(n_estimators=100)\r\n",
        "xgb.fit(df[:'2013'][feature], df[:'2013']['y'])\r\n",
        "sorted_idx = xgb.feature_importances_.argsort()\r\n",
        "# plot_importance(xgb, max_num_features = 20)#xgb內建畫圖，但比較不漂亮\r\n",
        "sns.set(font=zhfont.get_name())\r\n",
        "plt.yticks(fontproperties=zhfont)\r\n",
        "\r\n",
        "sns.barplot(xgb.feature_importances_[sorted_idx][-20::],feature[sorted_idx][-20::] )\r\n",
        "\r\n",
        "feature_new=feature[sorted_idx][-20::]"
      ],
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[07:30:02] WARNING: /workspace/src/objective/regression_obj.cu:152: reg:linear is now deprecated in favor of reg:squarederror.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OaH485qDoMih"
      },
      "source": [
        "#shap解釋重要特徵(有亂碼)\r\n",
        "# explainer = shap.TreeExplainer(xgb)\r\n",
        "# shap_values = explainer.shap_values(df[:'2013'][feature])\r\n",
        "# shap.summary_plot(shap_values, df[:'2013'][feature], plot_type=\"bar\")"
      ],
      "execution_count": 32,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r37Fkde8dlDX"
      },
      "source": [
        "train_x,train_y=deal_lstm(tol_normalize(df[:'2013']),feature_new)\r\n",
        "valid_x,valid_y=deal_lstm((df['2013':'2017']),feature_new)\r\n",
        "test_x,test_y=deal_lstm((df['2017':]),feature_new)"
      ],
      "execution_count": 33,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YoVy6xCRffFf",
        "outputId": "9b51e9b5-6c7e-470c-ff0a-ddc64004031a"
      },
      "source": [
        "print(train_x.shape)\r\n",
        "print(train_y.shape)\r\n"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(1594, 20, 20)\n",
            "(1594,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "w6sZd5J1dAtr"
      },
      "source": [
        "class ElecDataset(Dataset):\r\n",
        "    def __init__(self,feature,target):\r\n",
        "        self.feature = feature\r\n",
        "        self.target = target\r\n",
        "    \r\n",
        "    def __len__(self):\r\n",
        "        return len(self.feature)\r\n",
        "    \r\n",
        "    def __getitem__(self,idx):\r\n",
        "        item = self.feature[idx]\r\n",
        "        label = self.target[idx]\r\n",
        "        \r\n",
        "        return item,label"
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7pBqt5SY779e"
      },
      "source": [
        "class LSTM(nn.Module):\r\n",
        "    def __init__(self, input_dim, hidden_dim=30, num_layers=1, output_dim=1):\r\n",
        "      super(LSTM, self).__init__()\r\n",
        "      # Hidden dimensions\r\n",
        "      self.hidden_dim = hidden_dim    \r\n",
        "\r\n",
        "      # Number of hidden layers\r\n",
        "      self.num_layers = num_layers\r\n",
        "\r\n",
        "      self.lstm = nn.LSTM(input_dim, hidden_dim, num_layers, batch_first=True)\r\n",
        "\r\n",
        "      self.fc = nn.Linear(hidden_dim, output_dim) \r\n",
        "\r\n",
        "    def forward(self, x):\r\n",
        "      # Initialize hidden state with zeros\r\n",
        "      h0 = torch.zeros(self.num_layers, x.size(0), self.hidden_dim).requires_grad_().to(device)\r\n",
        "\r\n",
        "      # Initialize cell state\r\n",
        "      c0 = torch.zeros(self.num_layers, x.size(0), self.hidden_dim).requires_grad_().to(device)\r\n",
        "\r\n",
        "      # One time step\r\n",
        "      # If we don't, we'll backprop all the way to the start even after going through another batch\r\n",
        "      out, (hn, cn) = self.lstm(x)\r\n",
        "\r\n",
        "      out = out.to(device)\r\n",
        "      #print('b',out.shape)\r\n",
        "      #print(type(out))\r\n",
        "\r\n",
        "      # Index hidden state of last time step\r\n",
        "      # out[:, -1, :] --> 100, 100 --> just want last time step hidden states! \r\n",
        "      out = self.fc(out[:, -1, :]) \r\n",
        "      # out.size() --> 100, 10\r\n",
        "      return out     "
      ],
      "execution_count": 36,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nPz5E-mYxy2p"
      },
      "source": [
        "class CNN(nn.Module):\r\n",
        "    def __init__(self):\r\n",
        "        super(CNN,self).__init__()\r\n",
        "\r\n",
        "        self.conv = nn.Sequential(\r\n",
        "            nn.Conv1d(20,20,kernel_size=1,stride=2),# in,out,k\r\n",
        "            nn.SELU(20),\r\n",
        "            # nn.Dropout(),\r\n",
        "            # nn.MaxPool2d(2, stride=2),\r\n",
        "            )\r\n",
        "        \r\n",
        "  \r\n",
        "        self.fc_model = nn.Sequential(\r\n",
        "            nn.Linear(20*10,50),\t\t\r\n",
        "            nn.BatchNorm1d(50,track_running_stats=False),\r\n",
        "            nn.SELU(50),\r\n",
        "            nn.Dropout(),\r\n",
        "            nn.Linear(50,50),\t\t\r\n",
        "            nn.BatchNorm1d(50),\r\n",
        "            nn.SELU(50),\r\n",
        "            # nn.Dropout(),\r\n",
        "            # nn.Linear(50,30),\t\t\r\n",
        "            # nn.BatchNorm1d(30),\r\n",
        "            # nn.ReLU(),\r\n",
        "            nn.Dropout(),\r\n",
        "            nn.Linear(50,50))#5為輸出幾類\r\n",
        "\r\n",
        "        self.lstm = LSTM(50) \r\n",
        "        '''\r\n",
        "        #\r\n",
        "        Input: W Filter: K Padding: P Stride: S Output: (W –K + 2P) / S + 1\r\n",
        "        '''\r\n",
        "        \r\n",
        "    def forward(self,x):\r\n",
        "\r\n",
        "        x = self.conv(x)\r\n",
        "        x = x.view(x.size(0), -1)\r\n",
        "        out = self.fc_model(x)\r\n",
        "        out = self.lstm(out.unsqueeze(0)) \r\n",
        "\r\n",
        "        return out"
      ],
      "execution_count": 49,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iWRk-elLdQ48",
        "outputId": "bedfdcaa-836b-4fa5-a337-6ac317a859b2"
      },
      "source": [
        "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\r\n",
        "model = CNN().to(device)\r\n",
        "model = model.double()\r\n",
        "\r\n",
        "\r\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=1e-2)\r\n",
        "# scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=10, gamma=0.1, last_epoch=-1) #gamma=0.1)\r\n",
        "# scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer,gamma=0.9) #gamma=0.1)\r\n",
        "\r\n",
        "criterion = nn.MSELoss()\r\n",
        "model.to(device)"
      ],
      "execution_count": 50,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "CNN(\n",
              "  (conv): Sequential(\n",
              "    (0): Conv1d(20, 20, kernel_size=(1,), stride=(2,))\n",
              "    (1): SELU(inplace=True)\n",
              "  )\n",
              "  (fc_model): Sequential(\n",
              "    (0): Linear(in_features=200, out_features=50, bias=True)\n",
              "    (1): BatchNorm1d(50, eps=1e-05, momentum=0.1, affine=True, track_running_stats=False)\n",
              "    (2): SELU(inplace=True)\n",
              "    (3): Dropout(p=0.5, inplace=False)\n",
              "    (4): Linear(in_features=50, out_features=50, bias=True)\n",
              "    (5): BatchNorm1d(50, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
              "    (6): SELU(inplace=True)\n",
              "    (7): Dropout(p=0.5, inplace=False)\n",
              "    (8): Linear(in_features=50, out_features=50, bias=True)\n",
              "  )\n",
              "  (lstm): LSTM(\n",
              "    (lstm): LSTM(50, 30, batch_first=True)\n",
              "    (fc): Linear(in_features=30, out_features=1, bias=True)\n",
              "  )\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 50
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nR3Jy_kckAD7",
        "outputId": "526d46a6-aab6-4720-c1e2-6fa331b4c18a"
      },
      "source": [
        "print((train_x.reshape(train_x.shape[0],-1,train_x.shape[1])).shape)"
      ],
      "execution_count": 51,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(1594, 20, 20)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "O54b1jlodSR3"
      },
      "source": [
        "train = ElecDataset(train_x.reshape(train_x.shape[0],-1,train_x.shape[1]),train_y)\r\n",
        "valid = ElecDataset(valid_x.reshape(valid_x.shape[0],-1,valid_x.shape[1]),valid_y)\r\n",
        "test = ElecDataset(test_x.reshape(test_x.shape[0],-1,test_x.shape[1]),test_y)\r\n",
        "\r\n",
        "train_loader = torch.utils.data.DataLoader(train,batch_size=128,shuffle=False)\r\n",
        "valid_loader = torch.utils.data.DataLoader(valid,batch_size=128,shuffle=False)\r\n",
        "test_loader = torch.utils.data.DataLoader(test,batch_size=128,shuffle=False)"
      ],
      "execution_count": 52,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DKlbn4YTdcZT"
      },
      "source": [
        "train_losses = []\r\n",
        "valid_losses = []\r\n",
        "def Train():\r\n",
        "    \r\n",
        "    running_loss = .0\r\n",
        "    \r\n",
        "    model.train()\r\n",
        "    \r\n",
        "    for idx, (inputs,labels) in enumerate(train_loader):\r\n",
        "        inputs = inputs.cuda()\r\n",
        "        labels = labels.cuda()\r\n",
        "        optimizer.zero_grad()\r\n",
        "        preds = model(inputs)\r\n",
        "        loss = criterion(preds,labels)\r\n",
        "        torch.nn.utils.clip_grad_norm_(model.parameters(),1)\r\n",
        "\r\n",
        "        loss.backward()\r\n",
        "        optimizer.step()\r\n",
        "        # scheduler.step()\r\n",
        "        running_loss += loss\r\n",
        "        \r\n",
        "    train_loss = running_loss/len(train_loader)\r\n",
        "    train_losses.append(train_loss.detach().cpu().numpy())\r\n",
        "    \r\n",
        "    print(f'train_loss {train_loss}')"
      ],
      "execution_count": 53,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nojlwI3QgED4"
      },
      "source": [
        "def Valid():\r\n",
        "    running_loss = .0\r\n",
        "    \r\n",
        "    model.eval()\r\n",
        "    \r\n",
        "    with torch.no_grad():\r\n",
        "        for idx, (inputs, labels) in enumerate(valid_loader):\r\n",
        "            inputs = inputs.to(device)\r\n",
        "            labels = labels.to(device)\r\n",
        "            optimizer.zero_grad()\r\n",
        "            preds = model(inputs)\r\n",
        "            loss = criterion(preds,labels)\r\n",
        "            running_loss += loss\r\n",
        "            \r\n",
        "        valid_loss = running_loss/len(valid_loader)\r\n",
        "        valid_losses.append(valid_loss.detach().cpu().numpy())\r\n",
        "        print(f'valid_loss {valid_loss}')"
      ],
      "execution_count": 54,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eFlCPaCJgGX1",
        "outputId": "ec3e323d-6d2a-416e-e922-96de780771a9"
      },
      "source": [
        "epochs =450\r\n",
        "for epoch in range(epochs):\r\n",
        "    print('epochs {}/{}'.format(epoch+1,epochs))\r\n",
        "    Train()\r\n",
        "    Valid()\r\n",
        "    # gc.collect()"
      ],
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "epochs 1/450\n",
            "train_loss 0.055411337985568056\n",
            "valid_loss 0.015681707952822855\n",
            "epochs 2/450\n",
            "train_loss 0.03283838309952828\n",
            "valid_loss 0.013666641325329252\n",
            "epochs 3/450\n",
            "train_loss 0.013021692116378561\n",
            "valid_loss 0.0014963542477338181\n",
            "epochs 4/450\n",
            "train_loss 0.014603695036249373\n",
            "valid_loss 0.0027142913464800506\n",
            "epochs 5/450\n",
            "train_loss 0.007627706525172708\n",
            "valid_loss 0.003306778231652389\n",
            "epochs 6/450\n",
            "train_loss 0.005437492492731121\n",
            "valid_loss 0.004221517157634049\n",
            "epochs 7/450\n",
            "train_loss 0.0030249627222981228\n",
            "valid_loss 0.0008643400942070062\n",
            "epochs 8/450\n",
            "train_loss 0.001482415958325794\n",
            "valid_loss 0.0007340961548749232\n",
            "epochs 9/450\n",
            "train_loss 0.0007125251356204462\n",
            "valid_loss 0.0009644937734481172\n",
            "epochs 10/450\n",
            "train_loss 0.0005160581117610213\n",
            "valid_loss 0.0005212375125875226\n",
            "epochs 11/450\n",
            "train_loss 0.0005197623149254804\n",
            "valid_loss 0.0007734028542983541\n",
            "epochs 12/450\n",
            "train_loss 0.0002759860868272096\n",
            "valid_loss 0.0007651700149118561\n",
            "epochs 13/450\n",
            "train_loss 0.00042658093023455516\n",
            "valid_loss 0.0004137394865819679\n",
            "epochs 14/450\n",
            "train_loss 0.0006373381999326605\n",
            "valid_loss 0.0005198646549868013\n",
            "epochs 15/450\n",
            "train_loss 0.000350067805723722\n",
            "valid_loss 0.00037723303358118345\n",
            "epochs 16/450\n",
            "train_loss 0.00035390286788335786\n",
            "valid_loss 0.0002723903919054352\n",
            "epochs 17/450\n",
            "train_loss 0.00029872889410577824\n",
            "valid_loss 0.0002427081039049998\n",
            "epochs 18/450\n",
            "train_loss 0.0002750274431084472\n",
            "valid_loss 0.00026388604923354315\n",
            "epochs 19/450\n",
            "train_loss 0.00030110048402830854\n",
            "valid_loss 0.0001874902552278205\n",
            "epochs 20/450\n",
            "train_loss 0.0002448080513598331\n",
            "valid_loss 0.0001876286512315291\n",
            "epochs 21/450\n",
            "train_loss 0.00022824116214693036\n",
            "valid_loss 0.00018446472409092062\n",
            "epochs 22/450\n",
            "train_loss 0.0002145976662776281\n",
            "valid_loss 0.00014183503275424464\n",
            "epochs 23/450\n",
            "train_loss 0.00022818381186022593\n",
            "valid_loss 0.00012488069402383814\n",
            "epochs 24/450\n",
            "train_loss 0.0002482108972136082\n",
            "valid_loss 0.0001167864956156642\n",
            "epochs 25/450\n",
            "train_loss 0.00023146908721991398\n",
            "valid_loss 0.00018163748290680537\n",
            "epochs 26/450\n",
            "train_loss 0.00023230423542233166\n",
            "valid_loss 0.00011798526909875321\n",
            "epochs 27/450\n",
            "train_loss 0.0002186597558994463\n",
            "valid_loss 0.0001151331897743196\n",
            "epochs 28/450\n",
            "train_loss 0.0002052616911499475\n",
            "valid_loss 0.00011299122107004774\n",
            "epochs 29/450\n",
            "train_loss 0.00024082996845010756\n",
            "valid_loss 0.00023814007647392175\n",
            "epochs 30/450\n",
            "train_loss 0.00028934727739522774\n",
            "valid_loss 9.201431615756344e-05\n",
            "epochs 31/450\n",
            "train_loss 0.00021038514927981792\n",
            "valid_loss 9.03835745452338e-05\n",
            "epochs 32/450\n",
            "train_loss 0.000205782030025038\n",
            "valid_loss 9.849096370233076e-05\n",
            "epochs 33/450\n",
            "train_loss 0.00021059732069670588\n",
            "valid_loss 8.402713505530863e-05\n",
            "epochs 34/450\n",
            "train_loss 0.0002135110341535944\n",
            "valid_loss 0.0001402485367701493\n",
            "epochs 35/450\n",
            "train_loss 0.00020795439762018842\n",
            "valid_loss 8.584277684562305e-05\n",
            "epochs 36/450\n",
            "train_loss 0.00020852570022818255\n",
            "valid_loss 9.113120732707298e-05\n",
            "epochs 37/450\n",
            "train_loss 0.0001982161127855182\n",
            "valid_loss 8.668551053709403e-05\n",
            "epochs 38/450\n",
            "train_loss 0.00020636535777345192\n",
            "valid_loss 7.787134662179501e-05\n",
            "epochs 39/450\n",
            "train_loss 0.00019154869430612265\n",
            "valid_loss 7.806290087031964e-05\n",
            "epochs 40/450\n",
            "train_loss 0.00019865474658729696\n",
            "valid_loss 7.971267768888568e-05\n",
            "epochs 41/450\n",
            "train_loss 0.00021295365276985788\n",
            "valid_loss 9.39752477661012e-05\n",
            "epochs 42/450\n",
            "train_loss 0.00019894677260227513\n",
            "valid_loss 7.381824678867381e-05\n",
            "epochs 43/450\n",
            "train_loss 0.00020087313815747928\n",
            "valid_loss 7.420479072857833e-05\n",
            "epochs 44/450\n",
            "train_loss 0.0001970351823654974\n",
            "valid_loss 6.881781078425275e-05\n",
            "epochs 45/450\n",
            "train_loss 0.00019756656938987283\n",
            "valid_loss 8.693032952658925e-05\n",
            "epochs 46/450\n",
            "train_loss 0.00020477966741589494\n",
            "valid_loss 0.00010371178213021415\n",
            "epochs 47/450\n",
            "train_loss 0.00021358026533871917\n",
            "valid_loss 8.729027953193391e-05\n",
            "epochs 48/450\n",
            "train_loss 0.00022352864812517272\n",
            "valid_loss 8.835426154930984e-05\n",
            "epochs 49/450\n",
            "train_loss 0.00023385240610319954\n",
            "valid_loss 0.00012166532918567427\n",
            "epochs 50/450\n",
            "train_loss 0.00024105122342737517\n",
            "valid_loss 0.00016040134133464196\n",
            "epochs 51/450\n",
            "train_loss 0.0002631521356108833\n",
            "valid_loss 7.352702787933202e-05\n",
            "epochs 52/450\n",
            "train_loss 0.00022472679412542673\n",
            "valid_loss 9.063135027174738e-05\n",
            "epochs 53/450\n",
            "train_loss 0.00020801158421720784\n",
            "valid_loss 7.652593584553436e-05\n",
            "epochs 54/450\n",
            "train_loss 0.00020263420770025436\n",
            "valid_loss 7.961627142171921e-05\n",
            "epochs 55/450\n",
            "train_loss 0.00020991374988719232\n",
            "valid_loss 8.694940469370622e-05\n",
            "epochs 56/450\n",
            "train_loss 0.00019404763112684418\n",
            "valid_loss 6.992714515980115e-05\n",
            "epochs 57/450\n",
            "train_loss 0.00019322874485135596\n",
            "valid_loss 6.629752148567005e-05\n",
            "epochs 58/450\n",
            "train_loss 0.00020083763596075914\n",
            "valid_loss 6.693517151309208e-05\n",
            "epochs 59/450\n",
            "train_loss 0.00019453449703608664\n",
            "valid_loss 6.39910260437185e-05\n",
            "epochs 60/450\n",
            "train_loss 0.00019587033156071845\n",
            "valid_loss 6.24938668980203e-05\n",
            "epochs 61/450\n",
            "train_loss 0.00019420279446296092\n",
            "valid_loss 6.248942917138309e-05\n",
            "epochs 62/450\n",
            "train_loss 0.00019545560760838028\n",
            "valid_loss 5.947920332457207e-05\n",
            "epochs 63/450\n",
            "train_loss 0.00019189321688458803\n",
            "valid_loss 6.0466956088854205e-05\n",
            "epochs 64/450\n",
            "train_loss 0.0001927429972463032\n",
            "valid_loss 5.968176528290037e-05\n",
            "epochs 65/450\n",
            "train_loss 0.0001943382317502347\n",
            "valid_loss 6.21536692095671e-05\n",
            "epochs 66/450\n",
            "train_loss 0.0001941219033758517\n",
            "valid_loss 6.188810964729527e-05\n",
            "epochs 67/450\n",
            "train_loss 0.00019317233574690066\n",
            "valid_loss 6.290789623805358e-05\n",
            "epochs 68/450\n",
            "train_loss 0.00019402098881834444\n",
            "valid_loss 5.958522135106145e-05\n",
            "epochs 69/450\n",
            "train_loss 0.0001918060404244662\n",
            "valid_loss 6.120603798962101e-05\n",
            "epochs 70/450\n",
            "train_loss 0.00019397705244213821\n",
            "valid_loss 5.902693178588996e-05\n",
            "epochs 71/450\n",
            "train_loss 0.00019395662209405036\n",
            "valid_loss 6.165249195105076e-05\n",
            "epochs 72/450\n",
            "train_loss 0.00019221947134725776\n",
            "valid_loss 6.142594002999802e-05\n",
            "epochs 73/450\n",
            "train_loss 0.0001936117982284364\n",
            "valid_loss 6.549179375230003e-05\n",
            "epochs 74/450\n",
            "train_loss 0.00019684538985039303\n",
            "valid_loss 7.007062330849836e-05\n",
            "epochs 75/450\n",
            "train_loss 0.0002060706287418322\n",
            "valid_loss 6.610752900633583e-05\n",
            "epochs 76/450\n",
            "train_loss 0.00019576472724657935\n",
            "valid_loss 6.341111314992749e-05\n",
            "epochs 77/450\n",
            "train_loss 0.00020144546818193205\n",
            "valid_loss 6.348834399981489e-05\n",
            "epochs 78/450\n",
            "train_loss 0.00019446135433758336\n",
            "valid_loss 6.789129112148914e-05\n",
            "epochs 79/450\n",
            "train_loss 0.00019244311390312352\n",
            "valid_loss 6.0474780185365454e-05\n",
            "epochs 80/450\n",
            "train_loss 0.00019876232516691258\n",
            "valid_loss 7.917107600524068e-05\n",
            "epochs 81/450\n",
            "train_loss 0.00019441950495663966\n",
            "valid_loss 6.674999560221786e-05\n",
            "epochs 82/450\n",
            "train_loss 0.0001959238946936592\n",
            "valid_loss 6.183006236095414e-05\n",
            "epochs 83/450\n",
            "train_loss 0.00019713371106557323\n",
            "valid_loss 6.882881409223762e-05\n",
            "epochs 84/450\n",
            "train_loss 0.00019289718526271606\n",
            "valid_loss 6.074125456923892e-05\n",
            "epochs 85/450\n",
            "train_loss 0.00019663031160666836\n",
            "valid_loss 7.180954340544404e-05\n",
            "epochs 86/450\n",
            "train_loss 0.0001924770133753396\n",
            "valid_loss 6.093410549328027e-05\n",
            "epochs 87/450\n",
            "train_loss 0.00019415046946192499\n",
            "valid_loss 6.189617555088026e-05\n",
            "epochs 88/450\n",
            "train_loss 0.00019182067755400858\n",
            "valid_loss 5.9665221885654265e-05\n",
            "epochs 89/450\n",
            "train_loss 0.00020298989030581656\n",
            "valid_loss 6.902217798065696e-05\n",
            "epochs 90/450\n",
            "train_loss 0.00020180200950339327\n",
            "valid_loss 6.162984034545874e-05\n",
            "epochs 91/450\n",
            "train_loss 0.00020701352011110068\n",
            "valid_loss 6.178493011258045e-05\n",
            "epochs 92/450\n",
            "train_loss 0.00021236195256838978\n",
            "valid_loss 7.033732880104534e-05\n",
            "epochs 93/450\n",
            "train_loss 0.00021018999513641376\n",
            "valid_loss 6.227005220063554e-05\n",
            "epochs 94/450\n",
            "train_loss 0.00019816925459121155\n",
            "valid_loss 6.654993637423542e-05\n",
            "epochs 95/450\n",
            "train_loss 0.00019625935302153006\n",
            "valid_loss 6.265903539106345e-05\n",
            "epochs 96/450\n",
            "train_loss 0.0001981803753462078\n",
            "valid_loss 6.0128558966445325e-05\n",
            "epochs 97/450\n",
            "train_loss 0.00019717498901844344\n",
            "valid_loss 8.375840087235767e-05\n",
            "epochs 98/450\n",
            "train_loss 0.0002180450029745581\n",
            "valid_loss 0.00014427934208359865\n",
            "epochs 99/450\n",
            "train_loss 0.0002300898299732419\n",
            "valid_loss 0.0001855764113544354\n",
            "epochs 100/450\n",
            "train_loss 0.0002939853962828254\n",
            "valid_loss 0.000561262445297097\n",
            "epochs 101/450\n",
            "train_loss 0.00045271673147012013\n",
            "valid_loss 0.0005907266925988409\n",
            "epochs 102/450\n",
            "train_loss 0.0005988937622367339\n",
            "valid_loss 0.00011658578873758272\n",
            "epochs 103/450\n",
            "train_loss 0.0003797233812753891\n",
            "valid_loss 0.0001561181039832342\n",
            "epochs 104/450\n",
            "train_loss 0.0003458998708751329\n",
            "valid_loss 0.00020467057943119953\n",
            "epochs 105/450\n",
            "train_loss 0.00032032873130033326\n",
            "valid_loss 0.00029061606545185075\n",
            "epochs 106/450\n",
            "train_loss 0.00038897139133217684\n",
            "valid_loss 0.0002432065246359173\n",
            "epochs 107/450\n",
            "train_loss 0.0004967565372790956\n",
            "valid_loss 0.0006491871597661833\n",
            "epochs 108/450\n",
            "train_loss 0.00043543193625919806\n",
            "valid_loss 0.00012478902317038032\n",
            "epochs 109/450\n",
            "train_loss 0.00031311892592561074\n",
            "valid_loss 0.00011033537945359336\n",
            "epochs 110/450\n",
            "train_loss 0.0003282076952048587\n",
            "valid_loss 8.769249552669432e-05\n",
            "epochs 111/450\n",
            "train_loss 0.00024722857460402506\n",
            "valid_loss 8.745249046082158e-05\n",
            "epochs 112/450\n",
            "train_loss 0.0002449456235906886\n",
            "valid_loss 8.279074143109607e-05\n",
            "epochs 113/450\n",
            "train_loss 0.00020839676350976923\n",
            "valid_loss 7.680617668330022e-05\n",
            "epochs 114/450\n",
            "train_loss 0.00020329772184158324\n",
            "valid_loss 0.00014330402233656955\n",
            "epochs 115/450\n",
            "train_loss 0.00021289010169150396\n",
            "valid_loss 6.844025029610226e-05\n",
            "epochs 116/450\n",
            "train_loss 0.00020726765844344703\n",
            "valid_loss 6.654856026832785e-05\n",
            "epochs 117/450\n",
            "train_loss 0.00020576272207615986\n",
            "valid_loss 9.723748749244315e-05\n",
            "epochs 118/450\n",
            "train_loss 0.00021440299405685944\n",
            "valid_loss 7.320765567735245e-05\n",
            "epochs 119/450\n",
            "train_loss 0.00022915952867090216\n",
            "valid_loss 7.373655420323749e-05\n",
            "epochs 120/450\n",
            "train_loss 0.0002853888913458494\n",
            "valid_loss 8.803450022289404e-05\n",
            "epochs 121/450\n",
            "train_loss 0.0002491607189107923\n",
            "valid_loss 8.012253658103475e-05\n",
            "epochs 122/450\n",
            "train_loss 0.00022284884045389687\n",
            "valid_loss 0.00010014752332099572\n",
            "epochs 123/450\n",
            "train_loss 0.0002371807827988631\n",
            "valid_loss 8.818501412072569e-05\n",
            "epochs 124/450\n",
            "train_loss 0.00023216909391633928\n",
            "valid_loss 6.751339985838302e-05\n",
            "epochs 125/450\n",
            "train_loss 0.00020094422259376763\n",
            "valid_loss 6.533558698620268e-05\n",
            "epochs 126/450\n",
            "train_loss 0.0001998357787100262\n",
            "valid_loss 7.055504502272949e-05\n",
            "epochs 127/450\n",
            "train_loss 0.0002071201619641782\n",
            "valid_loss 9.896186964612118e-05\n",
            "epochs 128/450\n",
            "train_loss 0.00020914837865404801\n",
            "valid_loss 6.957394155390145e-05\n",
            "epochs 129/450\n",
            "train_loss 0.00020061666881293757\n",
            "valid_loss 9.299714012904126e-05\n",
            "epochs 130/450\n",
            "train_loss 0.00021077532924640514\n",
            "valid_loss 6.994730158151468e-05\n",
            "epochs 131/450\n",
            "train_loss 0.0002067399240003313\n",
            "valid_loss 7.885781087719609e-05\n",
            "epochs 132/450\n",
            "train_loss 0.00019569586161327459\n",
            "valid_loss 7.638291392264942e-05\n",
            "epochs 133/450\n",
            "train_loss 0.00019503059375840004\n",
            "valid_loss 7.223797982978409e-05\n",
            "epochs 134/450\n",
            "train_loss 0.00020072608719643665\n",
            "valid_loss 6.608257784408787e-05\n",
            "epochs 135/450\n",
            "train_loss 0.00020367788541915548\n",
            "valid_loss 6.68743064987967e-05\n",
            "epochs 136/450\n",
            "train_loss 0.00021572063304827027\n",
            "valid_loss 9.26273883809019e-05\n",
            "epochs 137/450\n",
            "train_loss 0.00020290510813244483\n",
            "valid_loss 7.077068439426972e-05\n",
            "epochs 138/450\n",
            "train_loss 0.00019915184095539354\n",
            "valid_loss 6.441310701388779e-05\n",
            "epochs 139/450\n",
            "train_loss 0.000199245126401165\n",
            "valid_loss 6.04648022030399e-05\n",
            "epochs 140/450\n",
            "train_loss 0.00019799823937170647\n",
            "valid_loss 6.0396593035850105e-05\n",
            "epochs 141/450\n",
            "train_loss 0.0002175741069194373\n",
            "valid_loss 0.00011528097137099882\n",
            "epochs 142/450\n",
            "train_loss 0.00020582026807579694\n",
            "valid_loss 0.00013490132243534772\n",
            "epochs 143/450\n",
            "train_loss 0.0002408916901243113\n",
            "valid_loss 6.743570008186312e-05\n",
            "epochs 144/450\n",
            "train_loss 0.0002226491207142999\n",
            "valid_loss 8.769925741252484e-05\n",
            "epochs 145/450\n",
            "train_loss 0.00021134438996211046\n",
            "valid_loss 7.683206919675412e-05\n",
            "epochs 146/450\n",
            "train_loss 0.00019777893784011144\n",
            "valid_loss 7.674373860333837e-05\n",
            "epochs 147/450\n",
            "train_loss 0.00019825938237381745\n",
            "valid_loss 6.771744468772496e-05\n",
            "epochs 148/450\n",
            "train_loss 0.00020063173553329657\n",
            "valid_loss 6.710852852327528e-05\n",
            "epochs 149/450\n",
            "train_loss 0.00020212003221189921\n",
            "valid_loss 8.027288717001968e-05\n",
            "epochs 150/450\n",
            "train_loss 0.00020179892957901258\n",
            "valid_loss 6.315676357444406e-05\n",
            "epochs 151/450\n",
            "train_loss 0.00019603872884578857\n",
            "valid_loss 6.12194699490801e-05\n",
            "epochs 152/450\n",
            "train_loss 0.00019102813716540855\n",
            "valid_loss 5.953990042425378e-05\n",
            "epochs 153/450\n",
            "train_loss 0.00019077038399294344\n",
            "valid_loss 6.0062882698608185e-05\n",
            "epochs 154/450\n",
            "train_loss 0.00019483926387881733\n",
            "valid_loss 6.411686078625326e-05\n",
            "epochs 155/450\n",
            "train_loss 0.00019528517249379105\n",
            "valid_loss 6.50476350082983e-05\n",
            "epochs 156/450\n",
            "train_loss 0.00019599740618050975\n",
            "valid_loss 5.8471002323313996e-05\n",
            "epochs 157/450\n",
            "train_loss 0.00019251494597753962\n",
            "valid_loss 6.239055180781182e-05\n",
            "epochs 158/450\n",
            "train_loss 0.00019214012216677893\n",
            "valid_loss 6.757467024617616e-05\n",
            "epochs 159/450\n",
            "train_loss 0.00019574417564931058\n",
            "valid_loss 5.949545047215326e-05\n",
            "epochs 160/450\n",
            "train_loss 0.00020228257584200144\n",
            "valid_loss 7.065046631528372e-05\n",
            "epochs 161/450\n",
            "train_loss 0.00020941784356457092\n",
            "valid_loss 6.529178224735764e-05\n",
            "epochs 162/450\n",
            "train_loss 0.00019371700268312604\n",
            "valid_loss 6.786661434165503e-05\n",
            "epochs 163/450\n",
            "train_loss 0.00019481003192780143\n",
            "valid_loss 6.95329632951672e-05\n",
            "epochs 164/450\n",
            "train_loss 0.00019381159444789827\n",
            "valid_loss 6.953760106944519e-05\n",
            "epochs 165/450\n",
            "train_loss 0.00020054945620138242\n",
            "valid_loss 6.216260231779843e-05\n",
            "epochs 166/450\n",
            "train_loss 0.00019515699965068383\n",
            "valid_loss 5.926673019552827e-05\n",
            "epochs 167/450\n",
            "train_loss 0.0001912496858737473\n",
            "valid_loss 5.956160919579794e-05\n",
            "epochs 168/450\n",
            "train_loss 0.0001964735767130512\n",
            "valid_loss 5.8605341279985716e-05\n",
            "epochs 169/450\n",
            "train_loss 0.00019579550794850562\n",
            "valid_loss 5.919706289590853e-05\n",
            "epochs 170/450\n",
            "train_loss 0.0001964566838838599\n",
            "valid_loss 6.049833725408708e-05\n",
            "epochs 171/450\n",
            "train_loss 0.00019411858272282893\n",
            "valid_loss 6.005557472381489e-05\n",
            "epochs 172/450\n",
            "train_loss 0.00019195671691483505\n",
            "valid_loss 6.087437509245177e-05\n",
            "epochs 173/450\n",
            "train_loss 0.00019837188998982452\n",
            "valid_loss 6.959302258962228e-05\n",
            "epochs 174/450\n",
            "train_loss 0.00020203798110991007\n",
            "valid_loss 6.403019964915589e-05\n",
            "epochs 175/450\n",
            "train_loss 0.000201574733781952\n",
            "valid_loss 6.618764024551366e-05\n",
            "epochs 176/450\n",
            "train_loss 0.00019380069936729392\n",
            "valid_loss 6.219020988686364e-05\n",
            "epochs 177/450\n",
            "train_loss 0.00019915164068858685\n",
            "valid_loss 5.9027496769541155e-05\n",
            "epochs 178/450\n",
            "train_loss 0.0001963787438713942\n",
            "valid_loss 6.374418728856354e-05\n",
            "epochs 179/450\n",
            "train_loss 0.0001955507156655533\n",
            "valid_loss 6.124240239540205e-05\n",
            "epochs 180/450\n",
            "train_loss 0.00019347082274266127\n",
            "valid_loss 6.257031285240119e-05\n",
            "epochs 181/450\n",
            "train_loss 0.00019117482918906388\n",
            "valid_loss 6.247132623577718e-05\n",
            "epochs 182/450\n",
            "train_loss 0.00019560727873840515\n",
            "valid_loss 6.383827649407782e-05\n",
            "epochs 183/450\n",
            "train_loss 0.00019225569179338292\n",
            "valid_loss 5.9976012423737735e-05\n",
            "epochs 184/450\n",
            "train_loss 0.00018973552330561316\n",
            "valid_loss 5.97660026134237e-05\n",
            "epochs 185/450\n",
            "train_loss 0.00019273295705062913\n",
            "valid_loss 6.118357581995139e-05\n",
            "epochs 186/450\n",
            "train_loss 0.00019039773910859334\n",
            "valid_loss 6.364921435026078e-05\n",
            "epochs 187/450\n",
            "train_loss 0.0001948314587992037\n",
            "valid_loss 6.068261143958845e-05\n",
            "epochs 188/450\n",
            "train_loss 0.000197054482277174\n",
            "valid_loss 7.621030141381885e-05\n",
            "epochs 189/450\n",
            "train_loss 0.0001923220711861155\n",
            "valid_loss 6.146100616410499e-05\n",
            "epochs 190/450\n",
            "train_loss 0.0001945264754286465\n",
            "valid_loss 6.412363151469169e-05\n",
            "epochs 191/450\n",
            "train_loss 0.0001955535981922608\n",
            "valid_loss 6.026275555904521e-05\n",
            "epochs 192/450\n",
            "train_loss 0.00019085256814341733\n",
            "valid_loss 6.443412552658847e-05\n",
            "epochs 193/450\n",
            "train_loss 0.00019541479905450015\n",
            "valid_loss 7.048199032313194e-05\n",
            "epochs 194/450\n",
            "train_loss 0.00020343390851844113\n",
            "valid_loss 7.067322432196335e-05\n",
            "epochs 195/450\n",
            "train_loss 0.0001946518986879362\n",
            "valid_loss 6.40244836906233e-05\n",
            "epochs 196/450\n",
            "train_loss 0.00020094568688990035\n",
            "valid_loss 6.004196089965338e-05\n",
            "epochs 197/450\n",
            "train_loss 0.00019216293816247672\n",
            "valid_loss 6.727105145738676e-05\n",
            "epochs 198/450\n",
            "train_loss 0.0001982800898191849\n",
            "valid_loss 5.938721909934098e-05\n",
            "epochs 199/450\n",
            "train_loss 0.00020025955257247402\n",
            "valid_loss 6.206757193821332e-05\n",
            "epochs 200/450\n",
            "train_loss 0.0001933084041405904\n",
            "valid_loss 6.632254114641404e-05\n",
            "epochs 201/450\n",
            "train_loss 0.0001922032974392016\n",
            "valid_loss 7.24562744303606e-05\n",
            "epochs 202/450\n",
            "train_loss 0.00019888611884393914\n",
            "valid_loss 6.338504078177866e-05\n",
            "epochs 203/450\n",
            "train_loss 0.0001969363805201133\n",
            "valid_loss 6.801422097719428e-05\n",
            "epochs 204/450\n",
            "train_loss 0.00020309563391711024\n",
            "valid_loss 6.13327200663175e-05\n",
            "epochs 205/450\n",
            "train_loss 0.00019753591622318005\n",
            "valid_loss 6.207691793511951e-05\n",
            "epochs 206/450\n",
            "train_loss 0.00019209379589930602\n",
            "valid_loss 8.006194853981647e-05\n",
            "epochs 207/450\n",
            "train_loss 0.00019304758712256986\n",
            "valid_loss 7.406261040487791e-05\n",
            "epochs 208/450\n",
            "train_loss 0.00019527768234252197\n",
            "valid_loss 6.707540692750564e-05\n",
            "epochs 209/450\n",
            "train_loss 0.00019476623869232762\n",
            "valid_loss 6.979445448308656e-05\n",
            "epochs 210/450\n",
            "train_loss 0.000198272857266947\n",
            "valid_loss 6.150269978809408e-05\n",
            "epochs 211/450\n",
            "train_loss 0.00019467133653894178\n",
            "valid_loss 6.222980704729454e-05\n",
            "epochs 212/450\n",
            "train_loss 0.0001984103777488801\n",
            "valid_loss 7.253407560511685e-05\n",
            "epochs 213/450\n",
            "train_loss 0.00019439776279823055\n",
            "valid_loss 6.195091713643304e-05\n",
            "epochs 214/450\n",
            "train_loss 0.00019448429557451357\n",
            "valid_loss 5.990683617284088e-05\n",
            "epochs 215/450\n",
            "train_loss 0.0001971620626011203\n",
            "valid_loss 6.416229311298052e-05\n",
            "epochs 216/450\n",
            "train_loss 0.00019601905985947216\n",
            "valid_loss 6.285502604828098e-05\n",
            "epochs 217/450\n",
            "train_loss 0.00020005084968285504\n",
            "valid_loss 6.419964633110937e-05\n",
            "epochs 218/450\n",
            "train_loss 0.00019377230876405774\n",
            "valid_loss 6.15774960600447e-05\n",
            "epochs 219/450\n",
            "train_loss 0.00019722887953018748\n",
            "valid_loss 6.47285695453907e-05\n",
            "epochs 220/450\n",
            "train_loss 0.00020470784835281868\n",
            "valid_loss 6.961760191991175e-05\n",
            "epochs 221/450\n",
            "train_loss 0.00019691160398512385\n",
            "valid_loss 6.722699382895407e-05\n",
            "epochs 222/450\n",
            "train_loss 0.00019261657427448696\n",
            "valid_loss 7.435644342103003e-05\n",
            "epochs 223/450\n",
            "train_loss 0.00020349892978484477\n",
            "valid_loss 7.062072310692469e-05\n",
            "epochs 224/450\n",
            "train_loss 0.00020593823726952417\n",
            "valid_loss 7.163396927778797e-05\n",
            "epochs 225/450\n",
            "train_loss 0.000193980660940026\n",
            "valid_loss 8.339171260072847e-05\n",
            "epochs 226/450\n",
            "train_loss 0.00019801472357443238\n",
            "valid_loss 6.76455110255687e-05\n",
            "epochs 227/450\n",
            "train_loss 0.000197708785591711\n",
            "valid_loss 6.726736265989749e-05\n",
            "epochs 228/450\n",
            "train_loss 0.00020722092946524664\n",
            "valid_loss 6.480898112623652e-05\n",
            "epochs 229/450\n",
            "train_loss 0.00019665366813487458\n",
            "valid_loss 6.650152442430593e-05\n",
            "epochs 230/450\n",
            "train_loss 0.00019650110983353682\n",
            "valid_loss 7.087998450265847e-05\n",
            "epochs 231/450\n",
            "train_loss 0.00021031522112422292\n",
            "valid_loss 6.542125962397404e-05\n",
            "epochs 232/450\n",
            "train_loss 0.00019760707759272392\n",
            "valid_loss 6.38918501992589e-05\n",
            "epochs 233/450\n",
            "train_loss 0.00019167819331977897\n",
            "valid_loss 6.284377680840686e-05\n",
            "epochs 234/450\n",
            "train_loss 0.00019464782934661669\n",
            "valid_loss 6.434288778906939e-05\n",
            "epochs 235/450\n",
            "train_loss 0.00019353803084056794\n",
            "valid_loss 8.263666620365359e-05\n",
            "epochs 236/450\n",
            "train_loss 0.00019512310830179042\n",
            "valid_loss 6.030175392821523e-05\n",
            "epochs 237/450\n",
            "train_loss 0.00019548484771084672\n",
            "valid_loss 6.51844988209346e-05\n",
            "epochs 238/450\n",
            "train_loss 0.0001926674043518323\n",
            "valid_loss 6.230766195973853e-05\n",
            "epochs 239/450\n",
            "train_loss 0.0001993802098390659\n",
            "valid_loss 6.267583761448367e-05\n",
            "epochs 240/450\n",
            "train_loss 0.00021799905695070579\n",
            "valid_loss 9.696202017516153e-05\n",
            "epochs 241/450\n",
            "train_loss 0.00020655811527494513\n",
            "valid_loss 7.688117115714705e-05\n",
            "epochs 242/450\n",
            "train_loss 0.00019724210765498004\n",
            "valid_loss 6.147160060562931e-05\n",
            "epochs 243/450\n",
            "train_loss 0.00019487606267783083\n",
            "valid_loss 7.32447382669278e-05\n",
            "epochs 244/450\n",
            "train_loss 0.0002038514429873902\n",
            "valid_loss 6.53391719489906e-05\n",
            "epochs 245/450\n",
            "train_loss 0.00020049669470919673\n",
            "valid_loss 7.491332025005401e-05\n",
            "epochs 246/450\n",
            "train_loss 0.0001979185035511831\n",
            "valid_loss 8.076672879810936e-05\n",
            "epochs 247/450\n",
            "train_loss 0.00022134572456752278\n",
            "valid_loss 0.00018221369014493916\n",
            "epochs 248/450\n",
            "train_loss 0.00024006589428124628\n",
            "valid_loss 0.00010558926717305918\n",
            "epochs 249/450\n",
            "train_loss 0.00024151858690650582\n",
            "valid_loss 0.00012633161011255703\n",
            "epochs 250/450\n",
            "train_loss 0.00023189999765984456\n",
            "valid_loss 0.0001292884438873426\n",
            "epochs 251/450\n",
            "train_loss 0.00023081104552219987\n",
            "valid_loss 7.588329085735284e-05\n",
            "epochs 252/450\n",
            "train_loss 0.00021953594464223477\n",
            "valid_loss 7.690371818204295e-05\n",
            "epochs 253/450\n",
            "train_loss 0.00021327483440426427\n",
            "valid_loss 7.77637702562848e-05\n",
            "epochs 254/450\n",
            "train_loss 0.0003155857409244552\n",
            "valid_loss 0.0002515006467767563\n",
            "epochs 255/450\n",
            "train_loss 0.000372785649631245\n",
            "valid_loss 0.00012412759801024126\n",
            "epochs 256/450\n",
            "train_loss 0.00020980958000174768\n",
            "valid_loss 9.41450331738829e-05\n",
            "epochs 257/450\n",
            "train_loss 0.00030620458430934296\n",
            "valid_loss 0.0002801333253731985\n",
            "epochs 258/450\n",
            "train_loss 0.000291098864310656\n",
            "valid_loss 0.00016668365532340726\n",
            "epochs 259/450\n",
            "train_loss 0.0002528150220073598\n",
            "valid_loss 0.00013559563565191695\n",
            "epochs 260/450\n",
            "train_loss 0.0002532831635497753\n",
            "valid_loss 8.42752948327871e-05\n",
            "epochs 261/450\n",
            "train_loss 0.00020374268334526522\n",
            "valid_loss 7.926960563719674e-05\n",
            "epochs 262/450\n",
            "train_loss 0.00021647798351779446\n",
            "valid_loss 0.00020293488688743512\n",
            "epochs 263/450\n",
            "train_loss 0.00025946079300639074\n",
            "valid_loss 0.00015466593660535099\n",
            "epochs 264/450\n",
            "train_loss 0.00029114049854980816\n",
            "valid_loss 0.00014357646871888098\n",
            "epochs 265/450\n",
            "train_loss 0.00021550198152134995\n",
            "valid_loss 0.000303042803749636\n",
            "epochs 266/450\n",
            "train_loss 0.0002672256388288962\n",
            "valid_loss 8.654667518087065e-05\n",
            "epochs 267/450\n",
            "train_loss 0.0002219942074721275\n",
            "valid_loss 0.00015041924284887713\n",
            "epochs 268/450\n",
            "train_loss 0.0003219406241992129\n",
            "valid_loss 0.0001681904456580118\n",
            "epochs 269/450\n",
            "train_loss 0.00023300057009211362\n",
            "valid_loss 0.0001106037291114165\n",
            "epochs 270/450\n",
            "train_loss 0.0002828717826058493\n",
            "valid_loss 0.00010803421265708803\n",
            "epochs 271/450\n",
            "train_loss 0.00021579195252924388\n",
            "valid_loss 7.306284847373905e-05\n",
            "epochs 272/450\n",
            "train_loss 0.00019832690589707618\n",
            "valid_loss 7.715555740642523e-05\n",
            "epochs 273/450\n",
            "train_loss 0.0001973461019502006\n",
            "valid_loss 8.287501478373584e-05\n",
            "epochs 274/450\n",
            "train_loss 0.00022638931259803555\n",
            "valid_loss 9.719462967435074e-05\n",
            "epochs 275/450\n",
            "train_loss 0.0001989124423915726\n",
            "valid_loss 9.596972884337388e-05\n",
            "epochs 276/450\n",
            "train_loss 0.0002065059353811874\n",
            "valid_loss 8.100263531733324e-05\n",
            "epochs 277/450\n",
            "train_loss 0.00023007885177860773\n",
            "valid_loss 6.52143510880202e-05\n",
            "epochs 278/450\n",
            "train_loss 0.00020895679264562445\n",
            "valid_loss 0.00011987241475897514\n",
            "epochs 279/450\n",
            "train_loss 0.0002339099274409264\n",
            "valid_loss 7.315358968974207e-05\n",
            "epochs 280/450\n",
            "train_loss 0.00020503528794499075\n",
            "valid_loss 8.74869597283133e-05\n",
            "epochs 281/450\n",
            "train_loss 0.00019185410076148359\n",
            "valid_loss 8.784311322235161e-05\n",
            "epochs 282/450\n",
            "train_loss 0.00019092828228545826\n",
            "valid_loss 9.191907372285101e-05\n",
            "epochs 283/450\n",
            "train_loss 0.000192125261608947\n",
            "valid_loss 9.760870863957769e-05\n",
            "epochs 284/450\n",
            "train_loss 0.00019101683898454773\n",
            "valid_loss 8.832559518913437e-05\n",
            "epochs 285/450\n",
            "train_loss 0.00019036207865710307\n",
            "valid_loss 9.101923930676295e-05\n",
            "epochs 286/450\n",
            "train_loss 0.00019183241454419353\n",
            "valid_loss 8.912813533533371e-05\n",
            "epochs 287/450\n",
            "train_loss 0.00018916251164788225\n",
            "valid_loss 9.427087525524165e-05\n",
            "epochs 288/450\n",
            "train_loss 0.0001906512892368212\n",
            "valid_loss 8.719944263043392e-05\n",
            "epochs 289/450\n",
            "train_loss 0.00019076877271753299\n",
            "valid_loss 8.352924071263587e-05\n",
            "epochs 290/450\n",
            "train_loss 0.00018962660577593463\n",
            "valid_loss 8.522846830066884e-05\n",
            "epochs 291/450\n",
            "train_loss 0.00018992187964264539\n",
            "valid_loss 8.583941770718635e-05\n",
            "epochs 292/450\n",
            "train_loss 0.00018977164613115117\n",
            "valid_loss 8.410484003491007e-05\n",
            "epochs 293/450\n",
            "train_loss 0.00019026340313138147\n",
            "valid_loss 8.744543248942099e-05\n",
            "epochs 294/450\n",
            "train_loss 0.00019031202101163237\n",
            "valid_loss 8.865412795996488e-05\n",
            "epochs 295/450\n",
            "train_loss 0.00019192140623604354\n",
            "valid_loss 7.785434692325347e-05\n",
            "epochs 296/450\n",
            "train_loss 0.00018953587372129373\n",
            "valid_loss 8.381468162151994e-05\n",
            "epochs 297/450\n",
            "train_loss 0.00019021989536516001\n",
            "valid_loss 8.033960502189407e-05\n",
            "epochs 298/450\n",
            "train_loss 0.00019135180119333792\n",
            "valid_loss 8.164034593115836e-05\n",
            "epochs 299/450\n",
            "train_loss 0.00019181332056811452\n",
            "valid_loss 7.355918970200689e-05\n",
            "epochs 300/450\n",
            "train_loss 0.00019050733453560698\n",
            "valid_loss 7.499088059387021e-05\n",
            "epochs 301/450\n",
            "train_loss 0.00018934641539553574\n",
            "valid_loss 7.811680764599161e-05\n",
            "epochs 302/450\n",
            "train_loss 0.00019005976734053607\n",
            "valid_loss 7.878696235445483e-05\n",
            "epochs 303/450\n",
            "train_loss 0.0001904699453750587\n",
            "valid_loss 7.529997234272131e-05\n",
            "epochs 304/450\n",
            "train_loss 0.00019095762249597372\n",
            "valid_loss 7.211225899192615e-05\n",
            "epochs 305/450\n",
            "train_loss 0.00019102961376273226\n",
            "valid_loss 6.959250204164928e-05\n",
            "epochs 306/450\n",
            "train_loss 0.0001914730110556913\n",
            "valid_loss 6.707255642976759e-05\n",
            "epochs 307/450\n",
            "train_loss 0.00018954256114081893\n",
            "valid_loss 7.595202210984161e-05\n",
            "epochs 308/450\n",
            "train_loss 0.00019059041187004274\n",
            "valid_loss 7.971096783430408e-05\n",
            "epochs 309/450\n",
            "train_loss 0.00019069454401660557\n",
            "valid_loss 8.915035862532354e-05\n",
            "epochs 310/450\n",
            "train_loss 0.00019331935953382354\n",
            "valid_loss 7.425882942937604e-05\n",
            "epochs 311/450\n",
            "train_loss 0.0001906756670734409\n",
            "valid_loss 7.165800595109264e-05\n",
            "epochs 312/450\n",
            "train_loss 0.00018982301667391736\n",
            "valid_loss 7.107678730315463e-05\n",
            "epochs 313/450\n",
            "train_loss 0.0001916519932647307\n",
            "valid_loss 6.432960783745198e-05\n",
            "epochs 314/450\n",
            "train_loss 0.00019136285580542302\n",
            "valid_loss 6.366583894471292e-05\n",
            "epochs 315/450\n",
            "train_loss 0.00019128621371102494\n",
            "valid_loss 6.114444428241165e-05\n",
            "epochs 316/450\n",
            "train_loss 0.0001925127021688168\n",
            "valid_loss 6.0858648589136214e-05\n",
            "epochs 317/450\n",
            "train_loss 0.00018917687034240263\n",
            "valid_loss 6.191459679111053e-05\n",
            "epochs 318/450\n",
            "train_loss 0.00018992950492244888\n",
            "valid_loss 6.555592043992742e-05\n",
            "epochs 319/450\n",
            "train_loss 0.00019170160573563053\n",
            "valid_loss 6.332249151543802e-05\n",
            "epochs 320/450\n",
            "train_loss 0.00019332920110771035\n",
            "valid_loss 6.118644787112284e-05\n",
            "epochs 321/450\n",
            "train_loss 0.000192492519256685\n",
            "valid_loss 6.087257862039325e-05\n",
            "epochs 322/450\n",
            "train_loss 0.00019062785578511472\n",
            "valid_loss 6.07946847060103e-05\n",
            "epochs 323/450\n",
            "train_loss 0.00018972497516417806\n",
            "valid_loss 6.341039169850431e-05\n",
            "epochs 324/450\n",
            "train_loss 0.0001905836686785906\n",
            "valid_loss 6.616789663376549e-05\n",
            "epochs 325/450\n",
            "train_loss 0.00019192321425922105\n",
            "valid_loss 6.235818792555937e-05\n",
            "epochs 326/450\n",
            "train_loss 0.00019963947211104957\n",
            "valid_loss 7.110461383712375e-05\n",
            "epochs 327/450\n",
            "train_loss 0.00019279574074761285\n",
            "valid_loss 7.31810872049601e-05\n",
            "epochs 328/450\n",
            "train_loss 0.00019465223680984115\n",
            "valid_loss 7.348803202888635e-05\n",
            "epochs 329/450\n",
            "train_loss 0.0001938535880068832\n",
            "valid_loss 6.812186598114903e-05\n",
            "epochs 330/450\n",
            "train_loss 0.00019334467312466514\n",
            "valid_loss 6.379016914264345e-05\n",
            "epochs 331/450\n",
            "train_loss 0.00019462864218380778\n",
            "valid_loss 6.415194588265005e-05\n",
            "epochs 332/450\n",
            "train_loss 0.00019467432551048083\n",
            "valid_loss 6.279957476404767e-05\n",
            "epochs 333/450\n",
            "train_loss 0.00019378701570490773\n",
            "valid_loss 6.332418988011159e-05\n",
            "epochs 334/450\n",
            "train_loss 0.00018958477345117544\n",
            "valid_loss 6.684004203468884e-05\n",
            "epochs 335/450\n",
            "train_loss 0.00019434457524847315\n",
            "valid_loss 6.251577825017941e-05\n",
            "epochs 336/450\n",
            "train_loss 0.00019943973345375697\n",
            "valid_loss 7.623280311304999e-05\n",
            "epochs 337/450\n",
            "train_loss 0.00019917554934735133\n",
            "valid_loss 6.16024491319392e-05\n",
            "epochs 338/450\n",
            "train_loss 0.00019093401949879011\n",
            "valid_loss 6.172840541947013e-05\n",
            "epochs 339/450\n",
            "train_loss 0.00018965567130246365\n",
            "valid_loss 6.6809029486105e-05\n",
            "epochs 340/450\n",
            "train_loss 0.00019033212391640912\n",
            "valid_loss 6.537051863502687e-05\n",
            "epochs 341/450\n",
            "train_loss 0.00018970608471348912\n",
            "valid_loss 7.087085722197406e-05\n",
            "epochs 342/450\n",
            "train_loss 0.0001914712692996556\n",
            "valid_loss 7.16835860631393e-05\n",
            "epochs 343/450\n",
            "train_loss 0.00019377580815136745\n",
            "valid_loss 7.263291722851213e-05\n",
            "epochs 344/450\n",
            "train_loss 0.00019396344988457812\n",
            "valid_loss 6.597733547286582e-05\n",
            "epochs 345/450\n",
            "train_loss 0.0001918989584435339\n",
            "valid_loss 6.423629808454292e-05\n",
            "epochs 346/450\n",
            "train_loss 0.0001913122592796506\n",
            "valid_loss 6.266762174016086e-05\n",
            "epochs 347/450\n",
            "train_loss 0.00019176274895620903\n",
            "valid_loss 6.22023825230709e-05\n",
            "epochs 348/450\n",
            "train_loss 0.00019031894960379648\n",
            "valid_loss 6.701398235888276e-05\n",
            "epochs 349/450\n",
            "train_loss 0.00019014890054245027\n",
            "valid_loss 7.203269941614055e-05\n",
            "epochs 350/450\n",
            "train_loss 0.00019386016450265782\n",
            "valid_loss 7.95645289975878e-05\n",
            "epochs 351/450\n",
            "train_loss 0.00019482821972385396\n",
            "valid_loss 7.296143369596484e-05\n",
            "epochs 352/450\n",
            "train_loss 0.00019257855937184656\n",
            "valid_loss 7.708530722106884e-05\n",
            "epochs 353/450\n",
            "train_loss 0.0001909895006245557\n",
            "valid_loss 7.443414189395695e-05\n",
            "epochs 354/450\n",
            "train_loss 0.00019100563889609882\n",
            "valid_loss 7.750350737267485e-05\n",
            "epochs 355/450\n",
            "train_loss 0.00019440306207670977\n",
            "valid_loss 7.099616153072672e-05\n",
            "epochs 356/450\n",
            "train_loss 0.00019255125818462812\n",
            "valid_loss 6.75230884106632e-05\n",
            "epochs 357/450\n",
            "train_loss 0.00018977530239314738\n",
            "valid_loss 7.061449480387876e-05\n",
            "epochs 358/450\n",
            "train_loss 0.00019067358197982526\n",
            "valid_loss 7.220582678658613e-05\n",
            "epochs 359/450\n",
            "train_loss 0.00019153074940430663\n",
            "valid_loss 7.49629352345021e-05\n",
            "epochs 360/450\n",
            "train_loss 0.00019181054066242905\n",
            "valid_loss 7.402781467636557e-05\n",
            "epochs 361/450\n",
            "train_loss 0.0001929841039959341\n",
            "valid_loss 6.725294236442046e-05\n",
            "epochs 362/450\n",
            "train_loss 0.00019016327085714745\n",
            "valid_loss 6.976640864308943e-05\n",
            "epochs 363/450\n",
            "train_loss 0.00019232846135060317\n",
            "valid_loss 7.480094011568769e-05\n",
            "epochs 364/450\n",
            "train_loss 0.00019290947900468274\n",
            "valid_loss 7.296600327392852e-05\n",
            "epochs 365/450\n",
            "train_loss 0.00019159437540275417\n",
            "valid_loss 7.404135965558714e-05\n",
            "epochs 366/450\n",
            "train_loss 0.00019500568553001374\n",
            "valid_loss 6.730937022561032e-05\n",
            "epochs 367/450\n",
            "train_loss 0.00019293003114713515\n",
            "valid_loss 7.10805426648301e-05\n",
            "epochs 368/450\n",
            "train_loss 0.00019062774044146746\n",
            "valid_loss 7.329307787464753e-05\n",
            "epochs 369/450\n",
            "train_loss 0.00019102139314487553\n",
            "valid_loss 7.756013338131773e-05\n",
            "epochs 370/450\n",
            "train_loss 0.00019198668981812265\n",
            "valid_loss 7.262502598503089e-05\n",
            "epochs 371/450\n",
            "train_loss 0.000195359604575771\n",
            "valid_loss 8.73734695447238e-05\n",
            "epochs 372/450\n",
            "train_loss 0.00019608060881077326\n",
            "valid_loss 9.413930909461674e-05\n",
            "epochs 373/450\n",
            "train_loss 0.0001973569797188131\n",
            "valid_loss 7.394312525042228e-05\n",
            "epochs 374/450\n",
            "train_loss 0.00019621295713955145\n",
            "valid_loss 7.005377167268124e-05\n",
            "epochs 375/450\n",
            "train_loss 0.00019588887742324576\n",
            "valid_loss 7.271210986894382e-05\n",
            "epochs 376/450\n",
            "train_loss 0.0001945516157796064\n",
            "valid_loss 7.256097922878078e-05\n",
            "epochs 377/450\n",
            "train_loss 0.00019548334159126017\n",
            "valid_loss 7.301304917948976e-05\n",
            "epochs 378/450\n",
            "train_loss 0.00019438832521220686\n",
            "valid_loss 6.962841700826562e-05\n",
            "epochs 379/450\n",
            "train_loss 0.00019847026177112186\n",
            "valid_loss 7.703809870194544e-05\n",
            "epochs 380/450\n",
            "train_loss 0.00019680644780151755\n",
            "valid_loss 8.588931237759859e-05\n",
            "epochs 381/450\n",
            "train_loss 0.00019284020569995078\n",
            "valid_loss 8.314166294754775e-05\n",
            "epochs 382/450\n",
            "train_loss 0.00019243630756498868\n",
            "valid_loss 7.67271864126239e-05\n",
            "epochs 383/450\n",
            "train_loss 0.00019163725004545378\n",
            "valid_loss 7.310403447194573e-05\n",
            "epochs 384/450\n",
            "train_loss 0.00019080490743618724\n",
            "valid_loss 7.82849000486977e-05\n",
            "epochs 385/450\n",
            "train_loss 0.00019021947119953354\n",
            "valid_loss 8.759663319079563e-05\n",
            "epochs 386/450\n",
            "train_loss 0.0001913368211053739\n",
            "valid_loss 9.338868418794406e-05\n",
            "epochs 387/450\n",
            "train_loss 0.00019204883699085073\n",
            "valid_loss 8.963330400592326e-05\n",
            "epochs 388/450\n",
            "train_loss 0.00019246477222275332\n",
            "valid_loss 7.801216197410489e-05\n",
            "epochs 389/450\n",
            "train_loss 0.0001906949413518997\n",
            "valid_loss 7.400985233746604e-05\n",
            "epochs 390/450\n",
            "train_loss 0.00019052262688891292\n",
            "valid_loss 8.022636971119514e-05\n",
            "epochs 391/450\n",
            "train_loss 0.0001929792414869079\n",
            "valid_loss 8.272924470270894e-05\n",
            "epochs 392/450\n",
            "train_loss 0.00019387541043802068\n",
            "valid_loss 7.832495664941214e-05\n",
            "epochs 393/450\n",
            "train_loss 0.00019362298535547463\n",
            "valid_loss 8.229996210014215e-05\n",
            "epochs 394/450\n",
            "train_loss 0.0001941177788049031\n",
            "valid_loss 7.488520475407716e-05\n",
            "epochs 395/450\n",
            "train_loss 0.0001924428137377492\n",
            "valid_loss 6.878844408289239e-05\n",
            "epochs 396/450\n",
            "train_loss 0.0001921747997574292\n",
            "valid_loss 6.849080321867991e-05\n",
            "epochs 397/450\n",
            "train_loss 0.0001912137692274819\n",
            "valid_loss 7.317543594309333e-05\n",
            "epochs 398/450\n",
            "train_loss 0.00019179905578160978\n",
            "valid_loss 7.426648895500543e-05\n",
            "epochs 399/450\n",
            "train_loss 0.0001906886423026642\n",
            "valid_loss 7.797032233200935e-05\n",
            "epochs 400/450\n",
            "train_loss 0.00019303148077709265\n",
            "valid_loss 7.7890643629983e-05\n",
            "epochs 401/450\n",
            "train_loss 0.00019153004094003983\n",
            "valid_loss 7.678286248527387e-05\n",
            "epochs 402/450\n",
            "train_loss 0.00019137338797595358\n",
            "valid_loss 8.815203714075417e-05\n",
            "epochs 403/450\n",
            "train_loss 0.00019318414014253077\n",
            "valid_loss 8.108596024215384e-05\n",
            "epochs 404/450\n",
            "train_loss 0.00019483568649079536\n",
            "valid_loss 8.607595801845324e-05\n",
            "epochs 405/450\n",
            "train_loss 0.0001937596092372307\n",
            "valid_loss 7.652696835149027e-05\n",
            "epochs 406/450\n",
            "train_loss 0.00019178345326469965\n",
            "valid_loss 7.930791827039628e-05\n",
            "epochs 407/450\n",
            "train_loss 0.0001947189880042456\n",
            "valid_loss 8.002732930496891e-05\n",
            "epochs 408/450\n",
            "train_loss 0.00019435876321941968\n",
            "valid_loss 7.496620104772144e-05\n",
            "epochs 409/450\n",
            "train_loss 0.00019380842137804022\n",
            "valid_loss 7.829895397148685e-05\n",
            "epochs 410/450\n",
            "train_loss 0.00019359935703248607\n",
            "valid_loss 7.329091437485331e-05\n",
            "epochs 411/450\n",
            "train_loss 0.0001919778838665679\n",
            "valid_loss 7.575660616587097e-05\n",
            "epochs 412/450\n",
            "train_loss 0.00019819264032927919\n",
            "valid_loss 7.061655529792261e-05\n",
            "epochs 413/450\n",
            "train_loss 0.0001938474036573028\n",
            "valid_loss 6.490626548002786e-05\n",
            "epochs 414/450\n",
            "train_loss 0.0001915479519356775\n",
            "valid_loss 7.390173760426327e-05\n",
            "epochs 415/450\n",
            "train_loss 0.0001924381429952528\n",
            "valid_loss 8.973422232656195e-05\n",
            "epochs 416/450\n",
            "train_loss 0.00019950369515206996\n",
            "valid_loss 8.276840679974952e-05\n",
            "epochs 417/450\n",
            "train_loss 0.00020021644459404042\n",
            "valid_loss 7.755685814868175e-05\n",
            "epochs 418/450\n",
            "train_loss 0.00020176925847044099\n",
            "valid_loss 8.241806335521365e-05\n",
            "epochs 419/450\n",
            "train_loss 0.0002014245614856105\n",
            "valid_loss 7.730437349776222e-05\n",
            "epochs 420/450\n",
            "train_loss 0.00019990336705050772\n",
            "valid_loss 7.516296042353883e-05\n",
            "epochs 421/450\n",
            "train_loss 0.00020565962872616452\n",
            "valid_loss 7.570158200670037e-05\n",
            "epochs 422/450\n",
            "train_loss 0.0001947564297825573\n",
            "valid_loss 7.719773880480943e-05\n",
            "epochs 423/450\n",
            "train_loss 0.00019849518405218647\n",
            "valid_loss 7.776843862370759e-05\n",
            "epochs 424/450\n",
            "train_loss 0.00019477416491745886\n",
            "valid_loss 7.386198666317059e-05\n",
            "epochs 425/450\n",
            "train_loss 0.00019222717112011867\n",
            "valid_loss 7.08201306766788e-05\n",
            "epochs 426/450\n",
            "train_loss 0.0001944802403252919\n",
            "valid_loss 7.31152659846287e-05\n",
            "epochs 427/450\n",
            "train_loss 0.0001975091665787551\n",
            "valid_loss 7.577384777831953e-05\n",
            "epochs 428/450\n",
            "train_loss 0.00019709383184948753\n",
            "valid_loss 8.937213793985492e-05\n",
            "epochs 429/450\n",
            "train_loss 0.00019403789536093184\n",
            "valid_loss 9.029516909501303e-05\n",
            "epochs 430/450\n",
            "train_loss 0.00019401396047005742\n",
            "valid_loss 9.870928650173894e-05\n",
            "epochs 431/450\n",
            "train_loss 0.00019120857717400212\n",
            "valid_loss 9.483741783770428e-05\n",
            "epochs 432/450\n",
            "train_loss 0.00019194104999636466\n",
            "valid_loss 0.00010238788390158874\n",
            "epochs 433/450\n",
            "train_loss 0.0002034431685376915\n",
            "valid_loss 9.931087744430474e-05\n",
            "epochs 434/450\n",
            "train_loss 0.00019133512223617772\n",
            "valid_loss 8.804723728425106e-05\n",
            "epochs 435/450\n",
            "train_loss 0.0001935956970977715\n",
            "valid_loss 9.254063791646812e-05\n",
            "epochs 436/450\n",
            "train_loss 0.00019332325434672968\n",
            "valid_loss 8.868789410487019e-05\n",
            "epochs 437/450\n",
            "train_loss 0.00019198849550300446\n",
            "valid_loss 8.807060032167261e-05\n",
            "epochs 438/450\n",
            "train_loss 0.00019451717983100649\n",
            "valid_loss 9.296627539781461e-05\n",
            "epochs 439/450\n",
            "train_loss 0.0001932088273512531\n",
            "valid_loss 9.015430203256568e-05\n",
            "epochs 440/450\n",
            "train_loss 0.00019192058744761645\n",
            "valid_loss 9.639825073628808e-05\n",
            "epochs 441/450\n",
            "train_loss 0.00019256835199935385\n",
            "valid_loss 9.273243890590619e-05\n",
            "epochs 442/450\n",
            "train_loss 0.00019243076686921662\n",
            "valid_loss 8.22678983373254e-05\n",
            "epochs 443/450\n",
            "train_loss 0.00019508354094496393\n",
            "valid_loss 8.061703452826244e-05\n",
            "epochs 444/450\n",
            "train_loss 0.0001961763496706107\n",
            "valid_loss 8.754119043435944e-05\n",
            "epochs 445/450\n",
            "train_loss 0.00021152597009985657\n",
            "valid_loss 7.791785407286027e-05\n",
            "epochs 446/450\n",
            "train_loss 0.00020881555685340106\n",
            "valid_loss 8.692834992194169e-05\n",
            "epochs 447/450\n",
            "train_loss 0.0002157941354397519\n",
            "valid_loss 6.337386996006722e-05\n",
            "epochs 448/450\n",
            "train_loss 0.00022208076157874507\n",
            "valid_loss 6.322584630854454e-05\n",
            "epochs 449/450\n",
            "train_loss 0.00020093190823176923\n",
            "valid_loss 6.608636502412389e-05\n",
            "epochs 450/450\n",
            "train_loss 0.00019802319250914355\n",
            "valid_loss 6.326446841499018e-05\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZNRidIh3gKwl",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 318
        },
        "outputId": "d2ff5294-c02b-4824-e91f-a6591d87bc88"
      },
      "source": [
        "# plt.figure(figsize=(25,10))\r\n",
        "\r\n",
        "plt.plot(train_losses, label='train loss')\r\n",
        "plt.plot(valid_losses, label='validation loss')\r\n",
        "\r\n",
        "plt.title('Training history')\r\n",
        "plt.ylabel('Loss')\r\n",
        "plt.xlabel('Epoch')\r\n",
        "plt.legend()\r\n",
        "# plt.ylim([0, 0.001]);"
      ],
      "execution_count": 56,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f19de15add8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 56
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VCDwdr2E0ldA"
      },
      "source": [
        "test_losses=[]\r\n",
        "real_test =[]\r\n",
        "predict_test = []\r\n",
        "def test():\r\n",
        "    running_loss = .0\r\n",
        "    \r\n",
        "    model.eval()\r\n",
        "    \r\n",
        "    with torch.no_grad():\r\n",
        "        for idx, (inputs, labels) in enumerate(test_loader):\r\n",
        "            inputs = inputs.to(device)\r\n",
        "            labels = labels.to(device)\r\n",
        "            optimizer.zero_grad()\r\n",
        "            preds = model(inputs)\r\n",
        "            loss = criterion(preds,labels)\r\n",
        "            running_loss += loss\r\n",
        "            real_test.extend(labels.to('cpu').numpy().tolist())\r\n",
        "            predict_test.extend(preds.to('cpu').numpy().tolist())\r\n",
        "        test_loss = running_loss/len(test_loader)\r\n",
        "        test_losses.append(test_loss.detach().cpu().numpy())\r\n",
        "        \r\n",
        "        return print(f'test_loss {test_loss}')"
      ],
      "execution_count": 57,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jNjqN90xo1P6",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d1e2bdaa-2286-4d24-ad9b-06fa79d3df5d"
      },
      "source": [
        "test()"
      ],
      "execution_count": 58,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "test_loss 0.0001051586722075397\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OX2q9Xgpo3bT",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 318
        },
        "outputId": "3301abef-b1c9-43e7-dbac-ee8c681e549a"
      },
      "source": [
        "plt.title('Predict return')\r\n",
        "\r\n",
        "plt.plot(real_test, label='real')\r\n",
        "plt.plot(predict_test, label='predict')\r\n",
        "plt.ylabel('return')\r\n",
        "plt.xlabel('time')\r\n",
        "plt.legend()"
      ],
      "execution_count": 59,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f19de07edd8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 59
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    }
  ]
}