Project_notebook.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "#Summary\n",
        "This notebook is a part of submission for CS598 DL4H group project.\n",
        "\n",
        "The objective of this reproducibility project is to replicate and verify the results presented in the paper \"*Benchmarking Deep Learning Architectures for predicting Readmission to the ICU and Describing patients-at-Risk*\" by Barbieri et al.\n",
        "\n",
        "Barbieri et al. aimed to solve the problem of predicting readmission to the Intensive Care Unit (ICU) and identifying patients-at-risk. Approximately 10 percent of patients discharged from an ICU in developed\n",
        "countries are readmitted during the same hospital stay (Rosenberg and Watts, 2000).\n",
        "\n",
        "The authors of the paper explored several deep learning architectures that included attention mechanisms, recurrent layers, neural ordinary differential equations (ODEs), and medical concept embeddings with time-aware attention. They proposed 14 models that were trained and evaluated on the MIMIC-III dataset.\n",
        "\n",
        "We were able to replicate the results of three out of 14 models at 80 epochs and found that the accuracy of the models was maintained while reducing the number of epochs from 80 to 10. The results also supported the original study's claim that models utilizing the recurrent component would have a higher accuracy than models utilizing only the attention layers. Additionally, our study confirmed the claim that the predictive accuracy of the deep learning models was considerably better than a baseline logistic regression model."
      ],
      "metadata": {
        "id": "ub_EMDPwBFJb"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Visualization of data\n",
        "\n",
        "The MIMIC-III v1.4 contains de-identified health-related data of patients admitted to the ICU at the Beth Israel Deaconess Medical Center in Boston, Massachusetts between 2001 and 2012. The dataset includes 45,298 ICU stays for 33,150 patients, labeled as either positive (N = 5,495) or negative (N = 39,803) depending on whether a patient did or did not experience readmission within 30 days from discharge.\n",
        "\n",
        " MIMIC-III is a relational database that contain detailed information about patients such as demographics, laboratory values, vital sign measurements, medication orders, procedures, clinical notes, and more.\n",
        "\n",
        "With a total of 26 interconnected tables, the database allows for complex data queries and analyses. For the purpose of illustration, we perform data analysis on two of these tables: *icu_stays* and *patients*, to gain insights into the characteristics of the patient population and their stays in the ICU.\n",
        "\n"
      ],
      "metadata": {
        "id": "Knwuw1n3BU6m"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "\n",
        "\n",
        "dir_path = \"/content/\"   # path to the downloaded MIMC-III files\n",
        "\n",
        "# Get ICU stays data\n",
        "icu_stays_df = pd.read_csv(dir_path + \"ICUSTAYS.csv\", parse_dates=['INTIME', 'OUTTIME'])\n",
        "icu_stays_df.head()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        },
        "id": "856PCBf4BoU_",
        "outputId": "7ef9a87b-c07f-4b33-cef6-7b382d2380cb"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "   ROW_ID  SUBJECT_ID  HADM_ID  ICUSTAY_ID DBSOURCE FIRST_CAREUNIT  \\\n",
              "0     365         268   110404      280836  carevue           MICU   \n",
              "1     366         269   106296      206613  carevue           MICU   \n",
              "2     367         270   188028      220345  carevue            CCU   \n",
              "3     368         271   173727      249196  carevue           MICU   \n",
              "4     369         272   164716      210407  carevue            CCU   \n",
              "\n",
              "  LAST_CAREUNIT  FIRST_WARDID  LAST_WARDID              INTIME  \\\n",
              "0          MICU            52           52 2198-02-14 23:27:38   \n",
              "1          MICU            52           52 2170-11-05 11:05:29   \n",
              "2           CCU            57           57 2128-06-24 15:05:20   \n",
              "3          SICU            52           23 2120-08-07 23:12:42   \n",
              "4           CCU            57           57 2186-12-25 21:08:04   \n",
              "\n",
              "              OUTTIME     LOS  \n",
              "0 2198-02-18 05:26:11  3.2490  \n",
              "1 2170-11-08 17:46:57  3.2788  \n",
              "2 2128-06-27 12:32:29  2.8939  \n",
              "3 2120-08-10 00:39:04  2.0600  \n",
              "4 2186-12-27 12:01:13  1.6202  "
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-06f99006-2a7e-4283-a3e8-ac05d0957775\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <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>ROW_ID</th>\n",
              "      <th>SUBJECT_ID</th>\n",
              "      <th>HADM_ID</th>\n",
              "      <th>ICUSTAY_ID</th>\n",
              "      <th>DBSOURCE</th>\n",
              "      <th>FIRST_CAREUNIT</th>\n",
              "      <th>LAST_CAREUNIT</th>\n",
              "      <th>FIRST_WARDID</th>\n",
              "      <th>LAST_WARDID</th>\n",
              "      <th>INTIME</th>\n",
              "      <th>OUTTIME</th>\n",
              "      <th>LOS</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>365</td>\n",
              "      <td>268</td>\n",
              "      <td>110404</td>\n",
              "      <td>280836</td>\n",
              "      <td>carevue</td>\n",
              "      <td>MICU</td>\n",
              "      <td>MICU</td>\n",
              "      <td>52</td>\n",
              "      <td>52</td>\n",
              "      <td>2198-02-14 23:27:38</td>\n",
              "      <td>2198-02-18 05:26:11</td>\n",
              "      <td>3.2490</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>366</td>\n",
              "      <td>269</td>\n",
              "      <td>106296</td>\n",
              "      <td>206613</td>\n",
              "      <td>carevue</td>\n",
              "      <td>MICU</td>\n",
              "      <td>MICU</td>\n",
              "      <td>52</td>\n",
              "      <td>52</td>\n",
              "      <td>2170-11-05 11:05:29</td>\n",
              "      <td>2170-11-08 17:46:57</td>\n",
              "      <td>3.2788</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>367</td>\n",
              "      <td>270</td>\n",
              "      <td>188028</td>\n",
              "      <td>220345</td>\n",
              "      <td>carevue</td>\n",
              "      <td>CCU</td>\n",
              "      <td>CCU</td>\n",
              "      <td>57</td>\n",
              "      <td>57</td>\n",
              "      <td>2128-06-24 15:05:20</td>\n",
              "      <td>2128-06-27 12:32:29</td>\n",
              "      <td>2.8939</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>368</td>\n",
              "      <td>271</td>\n",
              "      <td>173727</td>\n",
              "      <td>249196</td>\n",
              "      <td>carevue</td>\n",
              "      <td>MICU</td>\n",
              "      <td>SICU</td>\n",
              "      <td>52</td>\n",
              "      <td>23</td>\n",
              "      <td>2120-08-07 23:12:42</td>\n",
              "      <td>2120-08-10 00:39:04</td>\n",
              "      <td>2.0600</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>369</td>\n",
              "      <td>272</td>\n",
              "      <td>164716</td>\n",
              "      <td>210407</td>\n",
              "      <td>carevue</td>\n",
              "      <td>CCU</td>\n",
              "      <td>CCU</td>\n",
              "      <td>57</td>\n",
              "      <td>57</td>\n",
              "      <td>2186-12-25 21:08:04</td>\n",
              "      <td>2186-12-27 12:01:13</td>\n",
              "      <td>1.6202</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-06f99006-2a7e-4283-a3e8-ac05d0957775')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-06f99006-2a7e-4283-a3e8-ac05d0957775 button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-06f99006-2a7e-4283-a3e8-ac05d0957775');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 1
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Compute the mean and median LOS\n",
        "avg_los = icu_stays_df[\"LOS\"].mean()\n",
        "median_los = icu_stays_df[\"LOS\"].median()\n",
        "print(f'The average length of ICU stay in days: {avg_los:.2f}')\n",
        "print(f'The median length of ICU stay in days: {median_los:.2f}')\n",
        "\n",
        "# Plot the distribution\n",
        "plt.hist(icu_stays_df[\"LOS\"], bins=50)\n",
        "plt.xlabel('Length of ICU stay in days')\n",
        "plt.ylabel('Count')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 501
        },
        "id": "xPIFB4WTCFDf",
        "outputId": "49f365a8-e1aa-41a8-b94a-558c3cd6ce6b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "The average length of ICU stay in days: 4.92\n",
            "The median length of ICU stay in days: 2.09\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0, 0.5, 'Count')"
            ]
          },
          "metadata": {},
          "execution_count": 2
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Get patients data\n",
        "pat_df = pd.read_csv(dir_path + \"PATIENTS.csv\", parse_dates=['DOB', 'DOD'])\n",
        "pat_df.head()\n",
        "\n",
        "#  Please note that dates of birth for patients aged over 89 were shifted \n",
        "# to obscure their true age and comply with HIPAA regulations: \n",
        "# these patients appear in the database with ages of over 300 years."
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 206
        },
        "id": "6O_kGdBMCKZB",
        "outputId": "b68b9e1c-f04c-4f8f-adb0-0593c833d47c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "   ROW_ID  SUBJECT_ID GENDER        DOB        DOD             DOD_HOSP  \\\n",
              "0     234         249      F 2075-03-13        NaT                  NaN   \n",
              "1     235         250      F 2164-12-27 2188-11-22  2188-11-22 00:00:00   \n",
              "2     236         251      M 2090-03-15        NaT                  NaN   \n",
              "3     237         252      M 2078-03-06        NaT                  NaN   \n",
              "4     238         253      F 2089-11-26        NaT                  NaN   \n",
              "\n",
              "  DOD_SSN  EXPIRE_FLAG  \n",
              "0     NaN            0  \n",
              "1     NaN            1  \n",
              "2     NaN            0  \n",
              "3     NaN            0  \n",
              "4     NaN            0  "
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-1c6bb602-21f8-460c-9823-8cdc12dbd12f\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <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>ROW_ID</th>\n",
              "      <th>SUBJECT_ID</th>\n",
              "      <th>GENDER</th>\n",
              "      <th>DOB</th>\n",
              "      <th>DOD</th>\n",
              "      <th>DOD_HOSP</th>\n",
              "      <th>DOD_SSN</th>\n",
              "      <th>EXPIRE_FLAG</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>234</td>\n",
              "      <td>249</td>\n",
              "      <td>F</td>\n",
              "      <td>2075-03-13</td>\n",
              "      <td>NaT</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>235</td>\n",
              "      <td>250</td>\n",
              "      <td>F</td>\n",
              "      <td>2164-12-27</td>\n",
              "      <td>2188-11-22</td>\n",
              "      <td>2188-11-22 00:00:00</td>\n",
              "      <td>NaN</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>236</td>\n",
              "      <td>251</td>\n",
              "      <td>M</td>\n",
              "      <td>2090-03-15</td>\n",
              "      <td>NaT</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>237</td>\n",
              "      <td>252</td>\n",
              "      <td>M</td>\n",
              "      <td>2078-03-06</td>\n",
              "      <td>NaT</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>238</td>\n",
              "      <td>253</td>\n",
              "      <td>F</td>\n",
              "      <td>2089-11-26</td>\n",
              "      <td>NaT</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-1c6bb602-21f8-460c-9823-8cdc12dbd12f')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-1c6bb602-21f8-460c-9823-8cdc12dbd12f button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-1c6bb602-21f8-460c-9823-8cdc12dbd12f');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Link ICU stays and patients tables\n",
        "icu_pat = pd.merge(icu_stays_df, pat_df, how='inner', on='SUBJECT_ID')\n",
        "icu_pat.sort_values(by=['SUBJECT_ID', 'OUTTIME'], ascending=[True, False], inplace=True)\n",
        "\n",
        "# Time to next admission (discharge to admission)\n",
        "icu_pat['DAYS_TO_NEXT'] = (icu_pat.groupby(['SUBJECT_ID']).shift(1)['INTIME'] - icu_pat['OUTTIME']).dt.days\n",
        "# Plot the distribution\n",
        "plt.hist(icu_pat['DAYS_TO_NEXT'], bins=50)\n",
        "plt.xlabel('Time between discharge and next admission')\n",
        "plt.ylabel('Count')\n",
        "\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 466
        },
        "id": "ws0UqR8QCN-1",
        "outputId": "664d44b2-3f94-4209-c1b6-37df41816aae"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0, 0.5, 'Count')"
            ]
          },
          "metadata": {},
          "execution_count": 5
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# scatter plot\n",
        "icu_pat.plot.scatter(x='LOS', y='DAYS_TO_NEXT')\n",
        "plt.xlabel('Length of ICU stay in days')\n",
        "plt.ylabel('Time between discharge and next admission')\n",
        "# display the plot\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 449
        },
        "id": "TdTAvmYrD3vk",
        "outputId": "acf04884-c67f-45af-ecd0-59f1bfcf1ba3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qvb3GEUwOMsJ"
      },
      "source": [
        "During analysis, we utilized Google Colab Pro to improve the speed. We first confirm if we are using GPU and check the info of the GPU being utilized."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "n2IW3swLOKiY",
        "outputId": "acb5c0ed-b675-4042-f7f8-87988236c5c2"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "/device:GPU:0\n",
            "Sun May  7 18:02:07 2023       \n",
            "+-----------------------------------------------------------------------------+\n",
            "| NVIDIA-SMI 525.85.12    Driver Version: 525.85.12    CUDA Version: 12.0     |\n",
            "|-------------------------------+----------------------+----------------------+\n",
            "| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |\n",
            "| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |\n",
            "|                               |                      |               MIG M. |\n",
            "|===============================+======================+======================|\n",
            "|   0  Tesla T4            Off  | 00000000:00:04.0 Off |                    0 |\n",
            "| N/A   53C    P0    30W /  70W |    373MiB / 15360MiB |      0%      Default |\n",
            "|                               |                      |                  N/A |\n",
            "+-------------------------------+----------------------+----------------------+\n",
            "                                                                               \n",
            "+-----------------------------------------------------------------------------+\n",
            "| Processes:                                                                  |\n",
            "|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |\n",
            "|        ID   ID                                                   Usage      |\n",
            "|=============================================================================|\n",
            "+-----------------------------------------------------------------------------+\n"
          ]
        }
      ],
      "source": [
        "import tensorflow as tf\n",
        "print(tf.test.gpu_device_name())\n",
        "\n",
        "gpu_info = !nvidia-smi\n",
        "gpu_info = '\\n'.join(gpu_info)\n",
        "if gpu_info.find('failed') >= 0:\n",
        "  print('Not connected to a GPU')\n",
        "else:\n",
        "  print(gpu_info)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TpizCyUcYVv0"
      },
      "source": [
        "Determine the Memory being utilized"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-nIPjzUSXylH",
        "outputId": "ddd3b83f-d8af-4009-a8e0-41958baafdc2"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Your runtime has 27.3 gigabytes of available RAM\n",
            "\n",
            "You are using a high-RAM runtime!\n"
          ]
        }
      ],
      "source": [
        "from psutil import virtual_memory\n",
        "ram_gb = virtual_memory().total / 1e9\n",
        "print('Your runtime has {:.1f} gigabytes of available RAM\\n'.format(ram_gb))\n",
        "\n",
        "if ram_gb < 20:\n",
        "  print('Not using a high-RAM runtime')\n",
        "else:\n",
        "  print('You are using a high-RAM runtime!')"
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "wprYyXhioyFU"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jIy3hs2ISU4X"
      },
      "source": [
        "Now, we will get started with setting up the data. \n",
        "First, we requested access to PhysioNet, downloaded the MIMIC-III dataset, and then preprocessed the data on our local machines. \n",
        "Then we zipped the data files containing processed data and uploaded this zipped file to the root directory of our Google Drive. \n",
        "Now, we set up a connection to our data in Google Drive. "
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Rbba80udODJJ",
        "outputId": "96028926-1f6f-4d5b-bcb9-14f349b77b1f"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ]
        }
      ],
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zpgxk6c8OUYU"
      },
      "source": [
        "Unzip data.zip located in Google Drive root and move to Google Colab locally"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AGbEW_3nOT19",
        "outputId": "7a92a47c-f222-4d06-8992-8fb48b79f127"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Reading package lists... Done\n",
            "Building dependency tree       \n",
            "Reading state information... Done\n",
            "unzip is already the newest version (6.0-25ubuntu1.1).\n",
            "0 upgraded, 0 newly installed, 0 to remove and 24 not upgraded.\n",
            "Archive:  /content/drive/MyDrive/data.zip\n",
            "replace /content/data/chartevents_reduced.csv? [y]es, [n]o, [A]ll, [N]one, [r]ename: Archive:  /content/drive/MyDrive/logdir.zip\n",
            "replace /content/logdir/.DS_Store? [y]es, [n]o, [A]ll, [N]one, [r]ename: "
          ]
        }
      ],
      "source": [
        "!apt install unzip\n",
        "!unzip /content/drive/MyDrive/data.zip -d /content/\n",
        "!unzip /content/drive/MyDrive/logdir.zip -d /content/"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "C46qgGjrUjvw"
      },
      "source": [
        "First, we install and import all necessary dependencies. "
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LcKGVlNRZdjt",
        "outputId": "b1ce6c13-b148-462f-9116-8cc9f282efc4"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting torchdiffeq\n",
            "  Downloading torchdiffeq-0.2.3-py3-none-any.whl (31 kB)\n",
            "Requirement already satisfied: torch>=1.3.0 in /usr/local/lib/python3.10/dist-packages (from torchdiffeq) (2.0.0+cu118)\n",
            "Requirement already satisfied: scipy>=1.4.0 in /usr/local/lib/python3.10/dist-packages (from torchdiffeq) (1.10.1)\n",
            "Requirement already satisfied: numpy<1.27.0,>=1.19.5 in /usr/local/lib/python3.10/dist-packages (from scipy>=1.4.0->torchdiffeq) (1.22.4)\n",
            "Requirement already satisfied: sympy in /usr/local/lib/python3.10/dist-packages (from torch>=1.3.0->torchdiffeq) (1.11.1)\n",
            "Requirement already satisfied: jinja2 in /usr/local/lib/python3.10/dist-packages (from torch>=1.3.0->torchdiffeq) (3.1.2)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.10/dist-packages (from torch>=1.3.0->torchdiffeq) (4.5.0)\n",
            "Requirement already satisfied: networkx in /usr/local/lib/python3.10/dist-packages (from torch>=1.3.0->torchdiffeq) (3.1)\n",
            "Requirement already satisfied: triton==2.0.0 in /usr/local/lib/python3.10/dist-packages (from torch>=1.3.0->torchdiffeq) (2.0.0)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from torch>=1.3.0->torchdiffeq) (3.12.0)\n",
            "Requirement already satisfied: cmake in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch>=1.3.0->torchdiffeq) (3.25.2)\n",
            "Requirement already satisfied: lit in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch>=1.3.0->torchdiffeq) (16.0.2)\n",
            "Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.10/dist-packages (from jinja2->torch>=1.3.0->torchdiffeq) (2.1.2)\n",
            "Requirement already satisfied: mpmath>=0.19 in /usr/local/lib/python3.10/dist-packages (from sympy->torch>=1.3.0->torchdiffeq) (1.3.0)\n",
            "Installing collected packages: torchdiffeq\n",
            "Successfully installed torchdiffeq-0.2.3\n"
          ]
        }
      ],
      "source": [
        "pip install torchdiffeq"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "I5NAPhHsUtwg"
      },
      "outputs": [],
      "source": [
        "import torch\n",
        "import math\n",
        "import random \n",
        "import os\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "import torch.nn.functional as F\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import pickle\n",
        "import scipy.stats as st\n",
        "import torch.utils.data as utils\n",
        "from __future__ import print_function\n",
        "from pdb import set_trace as bp\n",
        "from math import pi\n",
        "from torchdiffeq import odeint, odeint_adjoint\n",
        "from pdb import set_trace as bp\n",
        "from torch.autograd import Variable\n",
        "from tqdm import tqdm\n",
        "from time import time\n",
        "from sklearn.metrics import accuracy_score, confusion_matrix, average_precision_score, roc_auc_score, f1_score\n",
        "from sklearn.metrics import *\n",
        "from sklearn.calibration import calibration_curve\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pWPcFKMGPOD7"
      },
      "source": [
        "Set Hyperparameters"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "_bs9QrDzOQp6"
      },
      "outputs": [],
      "source": [
        "'''Hyperparameters'''\n",
        "# data\n",
        "mimic_dir = '../../MIMIC-III Clinical Database/uncompressed/'\n",
        "#data_dir = '/content/drive/MyDrive/data/' # log directory  \n",
        "data_dir = 'data/' \n",
        "logdir = '/content/drive/MyDrive/logdir/' # log directory  \n",
        "#logdir = 'logdir/' \n",
        "\n",
        "# model\n",
        "min_count = 100 # words whose occurred less than min_cnt are encoded as OTHER\n",
        "  \n",
        "# training\n",
        "batch_size = 128\n",
        "num_epochs = 10\n",
        "#adjusted from 80 epochs\n",
        "dropout_rate = 0.5\n",
        "patience = 10 # early stopping\n",
        "\n",
        "# which data to load\n",
        "# on_the_cloud = False\n",
        "all_train = False\n",
        "# all_train = True\n",
        "\n",
        "# network variants\n",
        "# net_variant = 'birnn_concat_time_delta'\n",
        "# net_variant = 'birnn_concat_time_delta_attention'\n",
        "# net_variant = 'birnn_time_decay'#\n",
        "# net_variant = 'birnn_time_decay_attention'#\n",
        "# net_variant = 'ode_birnn'#\n",
        "# net_variant = 'ode_birnn_attention'#\n",
        "net_variant = 'ode_attention'#\n",
        "# net_variant = 'attention_concat_time'\n",
        "#net_variant = 'birnn_ode_decay'\n",
        "# net_variant = 'birnn_ode_decay_attention'\n",
        "# net_variant = 'mce_attention'#\n",
        "# net_variant = 'mce_birnn'#\n",
        "# net_variant = 'mce_birnn_attention'#\n",
        "\n",
        "# bootstrapping\n",
        "np_seed = 1234\n",
        "bootstrap_samples = 100\n",
        "\n",
        "# bayesian network\n",
        "pi = 0.5\n",
        "sigma1 = math.exp(-0)\n",
        "sigma2 = math.exp(-6)\n",
        "samples = 1\n",
        "test_samples = 10"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OJV0y6R6Pcgw"
      },
      "source": [
        "Data Load"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UuYxwONiPUcb",
        "outputId": "54c1d47c-fbf7-41e7-9214-4372ea1280bd"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Load data...\n"
          ]
        }
      ],
      "source": [
        "'''\n",
        "Aug 2018 by Sebastiano Barbieri\n",
        "s.barbieri@unsw.edu.au\n",
        "whitman5@illinois.edu adapted section below to test models with random data removed\n",
        "'''\n",
        "def get_data(data, type):\n",
        "  # Data\n",
        "  static       = data['static'].astype('float32')\n",
        "  label        = data['label'].astype('float32')\n",
        "  dp           = data['dp'].astype('int64') # diagnoses/procedures\n",
        "  cp           = data['cp'].astype('int64') # charts/prescriptions\n",
        "  dp_times     = data['dp_times'].astype('float32')\n",
        "  cp_times     = data['cp_times'].astype('float32')\n",
        "  train_ids    = data['train_ids']\n",
        "  validate_ids = data['validate_ids']\n",
        "  test_ids     = data['test_ids']  \n",
        "\n",
        "  # used for testing the models with 10 percent of data removed. 4500 is ~10 % of the data\n",
        "\n",
        "  #r = random.sample(range(0, len(train_ids)-1), 4500)\n",
        "  #o = np.full(len(train_ids), True)\n",
        "  #z = np.full(4500, False)\n",
        "  #np.put(o, r, z)\n",
        "  #train_ids = train_ids * o\n",
        "  #test_ids = test_ids * o\n",
        "\n",
        "  if (type == 'TRAIN'):\n",
        "    ids = train_ids\n",
        "  elif (type == 'VALIDATE'):\n",
        "    ids = validate_ids\n",
        "  elif (type == 'TEST'):\n",
        "    ids = test_ids\n",
        "  elif (type == 'ALL'):\n",
        "    ids = np.full_like(label, True, dtype=bool)\n",
        "\n",
        "  static   = static[ids, :]\n",
        "  label    = label[ids]\n",
        "  dp       = dp[ids, :]\n",
        "  cp       = cp[ids, :]\n",
        "  dp_times = dp_times[ids, :]\n",
        "  cp_times = cp_times[ids, :]\n",
        "  \n",
        "  return static, dp, cp, dp_times, cp_times, label\n",
        "\n",
        "\n",
        "def get_dictionaries(data):\n",
        "  return data['static_vars'], data['dict_dp'][()], data['dict_cp'][()]\n",
        "\n",
        "\n",
        "def num_static(data):\n",
        "  return data['static_vars'].shape[0]\n",
        "  \n",
        "\n",
        "def vocab_sizes(data):\n",
        "  return data['dp'].max()+1, data['cp'].max()+1\n",
        "\n",
        "  \n",
        "def get_trainloader(data, type, shuffle=True, idx=None):\n",
        "  # Data\n",
        "  static, dp, cp, dp_times, cp_times, label = get_data(data, type)\n",
        "\n",
        "  # Bootstrap\n",
        "  if idx is not None:\n",
        "    static, dp, cp, dp_times, cp_times, label = static[idx], dp[idx], cp[idx], dp_times[idx], cp_times[idx], label[idx]\n",
        "\n",
        "  # Compute total batch count\n",
        "  num_batches = len(label) // batch_size\n",
        "  \n",
        "  # Create dataset\n",
        "  dataset = utils.TensorDataset(torch.from_numpy(static), \n",
        "                                torch.from_numpy(dp),\n",
        "                                torch.from_numpy(cp),\n",
        "                                torch.from_numpy(dp_times),\n",
        "                                torch.from_numpy(cp_times),\n",
        "                                torch.from_numpy(label))\n",
        "\n",
        "  # Create batch queues\n",
        "  trainloader = utils.DataLoader(dataset,\n",
        "                                 batch_size = batch_size, \n",
        "                                 shuffle = shuffle,\n",
        "                                 sampler = None,\n",
        "                                 num_workers = 2,\n",
        "                                 drop_last = True)\n",
        "                                 \n",
        "  # Weight of positive samples for training\n",
        "  pos_weight = torch.tensor((len(label) - np.sum(label))/np.sum(label))\n",
        "  \n",
        "  return trainloader, num_batches, pos_weight\n",
        "  \n",
        "  \n",
        "if __name__ == '__main__':\n",
        "  print('Load data...')\n",
        "  data = np.load(data_dir + 'data_arrays.npz')\n",
        "  trainloader, num_batches, pos_weight = get_trainloader(data, 'TRAIN')\n",
        "  # vocab_diagnoses, vocab_procedures, vocab_prescriptions = vocab_sizes(data)\n",
        "  "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5Kk6hLg8PbRJ"
      },
      "source": [
        "ODE Modules"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "NGHgJotCPkQY"
      },
      "outputs": [],
      "source": [
        "'''\n",
        "Adapted from https://github.com/EmilienDupont/augmented-neural-odes/blob/master/anode/models.py\n",
        "'''\n",
        "\n",
        "MAX_NUM_STEPS = 1000  # Maximum number of steps for ODE solver\n",
        "\n",
        "class ODEFunc(nn.Module):\n",
        "    \"\"\"MLP modeling the derivative of ODE system.\n",
        "    Parameters\n",
        "    ----------\n",
        "    device : torch.device\n",
        "    data_dim : int\n",
        "        Dimension of data.\n",
        "    hidden_dim : int\n",
        "        Dimension of hidden layers.\n",
        "    augment_dim: int\n",
        "        Dimension of augmentation. If 0 does not augment ODE, otherwise augments\n",
        "        it with augment_dim dimensions.\n",
        "    time_dependent : bool\n",
        "        If True adds time as input, making ODE time dependent.\n",
        "    non_linearity : string\n",
        "        One of 'relu' and 'softplus'\n",
        "    \"\"\"\n",
        "    def __init__(self, device, data_dim, hidden_dim, augment_dim=0,\n",
        "                 time_dependent=False, non_linearity='relu'):\n",
        "        super(ODEFunc, self).__init__()\n",
        "        self.device = device\n",
        "        self.augment_dim = augment_dim\n",
        "        self.data_dim = data_dim\n",
        "        self.input_dim = data_dim + augment_dim\n",
        "        self.hidden_dim = hidden_dim\n",
        "        self.nfe = 0  # Number of function evaluations\n",
        "        self.time_dependent = time_dependent\n",
        "\n",
        "        if time_dependent:\n",
        "            self.fc1 = nn.Linear(self.input_dim + 1, hidden_dim)\n",
        "        else:\n",
        "            self.fc1 = nn.Linear(self.input_dim, hidden_dim)\n",
        "        self.fc2 = nn.Linear(hidden_dim, hidden_dim)\n",
        "        self.fc3 = nn.Linear(hidden_dim, self.input_dim)\n",
        "\n",
        "        if non_linearity == 'relu':\n",
        "            self.non_linearity = nn.ReLU(inplace=True)\n",
        "        elif non_linearity == 'softplus':\n",
        "            self.non_linearity = nn.Softplus()\n",
        "\n",
        "    def forward(self, t, x):\n",
        "        \"\"\"\n",
        "        Parameters\n",
        "        ----------\n",
        "        t : torch.Tensor\n",
        "            Current time. Shape (1,).\n",
        "        x : torch.Tensor\n",
        "            Shape (batch_size, input_dim)\n",
        "        \"\"\"\n",
        "        # Forward pass of model corresponds to one function evaluation, so\n",
        "        # increment counter\n",
        "        self.nfe += 1\n",
        "        if self.time_dependent:\n",
        "            # Shape (batch_size, 1)\n",
        "            t_vec = torch.ones(x.shape[0], 1).to(self.device) * t\n",
        "            # Shape (batch_size, data_dim + 1)\n",
        "            t_and_x = torch.cat([t_vec, x], 1)\n",
        "            # Shape (batch_size, hidden_dim)\n",
        "            out = self.fc1(t_and_x)\n",
        "        else:\n",
        "            out = self.fc1(x)\n",
        "        out = self.non_linearity(out)\n",
        "        out = self.fc2(out)\n",
        "        out = self.non_linearity(out)\n",
        "        out = self.fc3(out)\n",
        "        return out\n",
        "\n",
        "\n",
        "class ODEBlock(nn.Module):\n",
        "    \"\"\"Solves ODE defined by odefunc.\n",
        "    Parameters\n",
        "    ----------\n",
        "    device : torch.device\n",
        "    odefunc : ODEFunc instance or anode.conv_models.ConvODEFunc instance\n",
        "        Function defining dynamics of system.\n",
        "    is_conv : bool\n",
        "        If True, treats odefunc as a convolutional model.\n",
        "    tol : float\n",
        "        Error tolerance.\n",
        "    adjoint : bool\n",
        "        If True calculates gradient with adjoint method, otherwise\n",
        "        backpropagates directly through operations of ODE solver.\n",
        "    \"\"\"\n",
        "    def __init__(self, device, odefunc, is_conv=False, tol=1e-3, adjoint=False):\n",
        "        super(ODEBlock, self).__init__()\n",
        "        self.adjoint = adjoint\n",
        "        self.device = device\n",
        "        self.is_conv = is_conv\n",
        "        self.odefunc = odefunc\n",
        "        self.tol = tol\n",
        "\n",
        "    def forward(self, x, eval_times=None):\n",
        "        \"\"\"Solves ODE starting from x.\n",
        "        Parameters\n",
        "        ----------\n",
        "        x : torch.Tensor\n",
        "            Shape (batch_size, self.odefunc.data_dim)\n",
        "        eval_times : None or torch.Tensor\n",
        "            If None, returns solution of ODE at final time t=1. If torch.Tensor\n",
        "            then returns full ODE trajectory evaluated at points in eval_times.\n",
        "        \"\"\"\n",
        "        # Forward pass corresponds to solving ODE, so reset number of function\n",
        "        # evaluations counter\n",
        "        self.odefunc.nfe = 0\n",
        "        \n",
        "        if eval_times is None:\n",
        "            integration_time = torch.tensor([0, 1]).float().type_as(x)\n",
        "        else:\n",
        "            integration_time = eval_times.type_as(x)\n",
        "\n",
        "\n",
        "        if self.odefunc.augment_dim > 0:\n",
        "            if self.is_conv:\n",
        "                # Add augmentation\n",
        "                batch_size, channels, height, width = x.shape\n",
        "                aug = torch.zeros(batch_size, self.odefunc.augment_dim,\n",
        "                                  height, width).to(self.device)\n",
        "                # Shape (batch_size, channels + augment_dim, height, width)\n",
        "                x_aug = torch.cat([x, aug], 1)\n",
        "            else:\n",
        "                # Add augmentation\n",
        "                aug = torch.zeros(x.shape[0], self.odefunc.augment_dim).to(self.device)\n",
        "                # Shape (batch_size, data_dim + augment_dim)\n",
        "                x_aug = torch.cat([x, aug], 1)\n",
        "        else:\n",
        "            x_aug = x\n",
        "\n",
        "        if self.adjoint:\n",
        "            out = odeint_adjoint(self.odefunc, x_aug, integration_time,\n",
        "                                 rtol=self.tol, atol=self.tol, method='euler',\n",
        "                                 options={'max_num_steps': MAX_NUM_STEPS})\n",
        "        else:\n",
        "            out = odeint(self.odefunc, x_aug, integration_time,\n",
        "                         rtol=self.tol, atol=self.tol, method='euler',\n",
        "                         options={'max_num_steps': MAX_NUM_STEPS})\n",
        "\n",
        "        if eval_times is None:\n",
        "            return out[1]  # Return only final time\n",
        "        else:\n",
        "            return out\n",
        "\n",
        "\n",
        "class ODENet(nn.Module):\n",
        "    \"\"\"An ODEBlock followed by a Linear layer.\n",
        "    Parameters\n",
        "    ----------\n",
        "    device : torch.device\n",
        "    data_dim : int\n",
        "        Dimension of data.\n",
        "    hidden_dim : int\n",
        "        Dimension of hidden layers.\n",
        "    output_dim : int\n",
        "        Dimension of output after hidden layer. Should be 1 for regression or\n",
        "        num_classes for classification.\n",
        "    augment_dim: int\n",
        "        Dimension of augmentation. If 0 does not augment ODE, otherwise augments\n",
        "        it with augment_dim dimensions.\n",
        "    time_dependent : bool\n",
        "        If True adds time as input, making ODE time dependent.\n",
        "    non_linearity : string\n",
        "        One of 'relu' and 'softplus'\n",
        "    tol : float\n",
        "        Error tolerance.\n",
        "    adjoint : bool\n",
        "        If True calculates gradient with adjoint method, otherwise\n",
        "        backpropagates directly through operations of ODE solver.\n",
        "    \"\"\"\n",
        "    def __init__(self, device, data_dim, hidden_dim, output_dim=1,\n",
        "                 augment_dim=0, time_dependent=False, non_linearity='relu',\n",
        "                 tol=1e-3, adjoint=False):\n",
        "        super(ODENet, self).__init__()\n",
        "        self.device = device\n",
        "        self.data_dim = data_dim\n",
        "        self.hidden_dim = hidden_dim\n",
        "        self.augment_dim = augment_dim\n",
        "        self.output_dim = output_dim\n",
        "        self.time_dependent = time_dependent\n",
        "        self.tol = tol\n",
        "\n",
        "        odefunc = ODEFunc(device, data_dim, hidden_dim, augment_dim,\n",
        "                          time_dependent, non_linearity)\n",
        "\n",
        "        self.odeblock = ODEBlock(device, odefunc, tol=tol, adjoint=adjoint)\n",
        "\n",
        "    def forward(self, x, eval_times=None):\n",
        "        features = self.odeblock(x, eval_times)\n",
        "        return features\n",
        "        "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Yt3pf_trPvnp"
      },
      "source": [
        "Modules"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "NdW6Aj1APywK"
      },
      "outputs": [],
      "source": [
        "class Attention(torch.nn.Module):\n",
        "  \"\"\"\n",
        "  Dot-product attention module.\n",
        "  \n",
        "  Args:\n",
        "    inputs: A `Tensor` with embeddings in the last dimension.\n",
        "    mask: A `Tensor`. Dimensions are the same as inputs but without the embedding dimension.\n",
        "      Values are 0 for 0-padding in the input and 1 elsewhere.\n",
        "\n",
        "  Returns:\n",
        "    outputs: The input `Tensor` whose embeddings in the last dimension have undergone a weighted average.\n",
        "      The second-last dimension of the `Tensor` is removed.\n",
        "    attention_weights: weights given to each embedding.\n",
        "  \"\"\"\n",
        "  def __init__(self, embedding_dim):\n",
        "    super(Attention, self).__init__()\n",
        "    self.context = nn.Parameter(torch.Tensor(embedding_dim)) # context vector\n",
        "    self.linear_hidden = nn.Linear(embedding_dim, embedding_dim)\n",
        "    self.reset_parameters()\n",
        "    \n",
        "  def reset_parameters(self):\n",
        "    nn.init.normal_(self.context)\n",
        "\n",
        "  def forward(self, inputs, mask):\n",
        "    # Hidden representation of embeddings (no change in dimensions)\n",
        "    hidden = torch.tanh(self.linear_hidden(inputs))\n",
        "    # Compute weight of each embedding\n",
        "    importance = torch.sum(hidden * self.context, dim=-1)\n",
        "    importance = importance.masked_fill(mask == 0, -1e9)\n",
        "    # Softmax so that weights sum up to one\n",
        "    attention_weights = F.softmax(importance, dim=-1)\n",
        "    # Weighted sum of embeddings\n",
        "    weighted_projection = inputs * torch.unsqueeze(attention_weights, dim=-1)\n",
        "    # Output\n",
        "    outputs = torch.sum(weighted_projection, dim=-2)\n",
        "    return outputs, attention_weights\n",
        "\n",
        "\n",
        "class GRUExponentialDecay(nn.Module):\n",
        "  \"\"\"\n",
        "  GRU RNN module where the hidden state decays exponentially\n",
        "  (see e.g. Che et al. 2018, Recurrent Neural Networks for Multivariate Time Series\n",
        "  with Missing Values).\n",
        "  \n",
        "  Args:\n",
        "    inputs: A `Tensor` with embeddings in the last dimension.\n",
        "    times: A `Tensor` with the same shape as inputs containing the recorded times (but no embedding dimension).\n",
        "\n",
        "  Returns:\n",
        "    outs: Hidden states of the RNN.\n",
        "  \"\"\"\n",
        "  def __init__(self, input_size, hidden_size, bias=True):\n",
        "    super(GRUExponentialDecay, self).__init__()\n",
        "    self.input_size = input_size\n",
        "    self.hidden_size = hidden_size\n",
        "    self.gru_cell = nn.GRUCell(input_size, hidden_size)\n",
        "    self.decays = nn.Parameter(torch.Tensor(hidden_size)) # exponential decays vector\n",
        "  \n",
        "  def forward(self, inputs, times):\n",
        "    # initializing and then calling cuda() later isn't working for some reason\n",
        "    if torch.cuda.is_available():\n",
        "      hn = torch.zeros(inputs.size(0), self.hidden_size).cuda() # batch_size x hidden_size\n",
        "      outs = torch.zeros(inputs.size(0), inputs.size(1), self.hidden_size).cuda() # batch_size x seq_len x hidden_size\n",
        "    else:\n",
        "      hn = torch.zeros(inputs.size(0), self.hidden_size) # batch_size x hidden_size\n",
        "      outs = torch.zeros(inputs.size(0), inputs.size(1), self.hidden_size) # batch_size x seq_len x hidden_size\n",
        "    \n",
        "    # this is slow\n",
        "    for seq in range(inputs.size(1)):\n",
        "      hn = self.gru_cell(inputs[:,seq,:], hn)\n",
        "      outs[:,seq,:] = hn\n",
        "      hn = hn*torch.exp(-torch.clamp(torch.unsqueeze(times[:,seq], dim=-1)*self.decays, min=0))\n",
        "    return outs\n",
        "\n",
        "\n",
        "class GRUOdeDecay(nn.Module):\n",
        "  \"\"\"\n",
        "  GRU RNN module where the hidden state decays according to an ODE.\n",
        "  (see Rubanova et al. 2019, Latent ODEs for Irregularly-Sampled Time Series)\n",
        "  \n",
        "  Args:\n",
        "    inputs: A `Tensor` with embeddings in the last dimension.\n",
        "    times: A `Tensor` with the same shape as inputs containing the recorded times (but no embedding dimension).\n",
        "\n",
        "  Returns:\n",
        "    outs: Hidden states of the RNN.\n",
        "  \"\"\"\n",
        "  def __init__(self, input_size, hidden_size, bias=True):\n",
        "    super(GRUOdeDecay, self).__init__()\n",
        "    self.input_size = input_size\n",
        "    self.hidden_size = hidden_size\n",
        "    self.gru_cell = nn.GRUCell(input_size, hidden_size)\n",
        "    self.decays = nn.Parameter(torch.Tensor(hidden_size)) # exponential decays vector\n",
        "    \n",
        "    # ODE\n",
        "    self.device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')\n",
        "    self.ode_net = ODENet(self.device, self.input_size, self.input_size, output_dim=self.input_size, augment_dim=0, time_dependent=False, non_linearity='softplus', tol=1e-3, adjoint=True)\n",
        "  \n",
        "  def forward(self, inputs, times):\n",
        "    # initializing and then calling cuda() later isn't working for some reason\n",
        "    if torch.cuda.is_available():\n",
        "      hn = torch.zeros(inputs.size(0), self.hidden_size).cuda() # batch_size x hidden_size\n",
        "      outs = torch.zeros(inputs.size(0), inputs.size(1), self.hidden_size).cuda() # batch_size x seq_len x hidden_size\n",
        "    else:\n",
        "      hn = torch.zeros(inputs.size(0), self.hidden_size) # batch_size x hidden_size\n",
        "      outs = torch.zeros(inputs.size(0), inputs.size(1), self.hidden_size) # batch_size x seq_len x hidden_size\n",
        "\n",
        "    # this is slow\n",
        "    for seq in range(inputs.size(1)):\n",
        "      hn = self.gru_cell(inputs[:,seq,:], hn)\n",
        "      outs[:,seq,:] = hn\n",
        "      \n",
        "      times_unique, inverse_indices = torch.unique(times[:,seq], sorted=True, return_inverse=True)\n",
        "      if times_unique.size(0) > 1:\n",
        "        hn = self.ode_net(hn, times_unique)\n",
        "        hn = hn[inverse_indices, torch.arange(0, inverse_indices.size(0)), :]\n",
        "    return outs\n",
        "\n",
        "\n",
        "def abs_time_to_delta(times):\n",
        "  delta = torch.cat((torch.unsqueeze(times[:, 0], dim=-1), times[:, 1:] - times[:, :-1]), dim=1)\n",
        "  delta = torch.clamp(delta, min=0)\n",
        "  return delta\n",
        "\n",
        "\n",
        "if net_variant == 'birnn_concat_time_delta':\n",
        "  # GRU\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(np.ceil(num_dp_codes**0.25))\n",
        "      self.embed_cp_dim = int(np.ceil(num_cp_codes**0.25))\n",
        "\n",
        "      # Embedding layers\n",
        "      self.embed_dp = nn.Embedding(num_embeddings=num_dp_codes, embedding_dim=self.embed_dp_dim, padding_idx=0)\n",
        "      self.embed_cp = nn.Embedding(num_embeddings=num_cp_codes, embedding_dim=self.embed_cp_dim, padding_idx=0)\n",
        "\n",
        "      # GRU layers\n",
        "      self.gru_dp_fw = nn.GRU(input_size=self.embed_dp_dim+1, hidden_size=self.embed_dp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_fw = nn.GRU(input_size=self.embed_cp_dim+1, hidden_size=self.embed_cp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_dp_bw = nn.GRU(input_size=self.embed_dp_dim+1, hidden_size=self.embed_dp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_bw = nn.GRU(input_size=self.embed_cp_dim+1, hidden_size=self.embed_cp_dim+1, num_layers=1, batch_first=True)\n",
        "      \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(2*(self.embed_dp_dim+1), 1)\n",
        "      self.fc_cp  = nn.Linear(2*(self.embed_cp_dim+1), 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Compute time delta\n",
        "      ## output dim: batch_size x seq_len\n",
        "      dp_t_delta_fw = abs_time_to_delta(dp_t)\n",
        "      cp_t_delta_fw = abs_time_to_delta(cp_t)\n",
        "      dp_t_delta_bw = abs_time_to_delta(torch.flip(dp_t, [1]))\n",
        "      cp_t_delta_bw = abs_time_to_delta(torch.flip(cp_t, [1]))    \n",
        "    \n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp_fw = self.embed_dp(dp)\n",
        "      embedded_cp_fw = self.embed_cp(cp)\n",
        "      embedded_dp_bw = torch.flip(embedded_dp_fw, [1])\n",
        "      embedded_cp_bw = torch.flip(embedded_cp_fw, [1])\n",
        "      \n",
        "      # Concatate with time\n",
        "      ## output dim: batch_size x seq_len x (embedding_dim+1)\n",
        "      concat_dp_fw = torch.cat((embedded_dp_fw, torch.unsqueeze(dp_t_delta_fw, dim=-1)), dim=-1)\n",
        "      concat_cp_fw = torch.cat((embedded_cp_fw, torch.unsqueeze(cp_t_delta_fw, dim=-1)), dim=-1)\n",
        "      concat_dp_bw = torch.cat((embedded_dp_bw, torch.unsqueeze(dp_t_delta_bw, dim=-1)), dim=-1)\n",
        "      concat_cp_bw = torch.cat((embedded_cp_bw, torch.unsqueeze(cp_t_delta_bw, dim=-1)), dim=-1)\n",
        "      ## Dropout\n",
        "      concat_dp_fw = self.dropout(concat_dp_fw)\n",
        "      concat_cp_fw = self.dropout(concat_cp_fw)\n",
        "      concat_dp_bw = self.dropout(concat_dp_bw)\n",
        "      concat_cp_bw = self.dropout(concat_cp_bw)\n",
        "      \n",
        "      # GRU\n",
        "      ## output dim rnn:        batch_size x seq_len x (embedding_dim+1)\n",
        "      ## output dim rnn_hidden: batch_size x 1 x (embedding_dim+1)\n",
        "      rnn_dp_fw, rnn_hidden_dp_fw = self.gru_dp_fw(concat_dp_fw)\n",
        "      rnn_cp_fw, rnn_hidden_cp_fw = self.gru_cp_fw(concat_cp_fw)\n",
        "      rnn_dp_bw, rnn_hidden_dp_bw = self.gru_dp_bw(concat_dp_bw)\n",
        "      rnn_cp_bw, rnn_hidden_cp_bw = self.gru_cp_bw(concat_cp_bw)      \n",
        "      ## output dim rnn_hidden: batch_size x (embedding_dim+1)\n",
        "      rnn_hidden_dp_fw = rnn_hidden_dp_fw.view(-1, self.embed_dp_dim+1)\n",
        "      rnn_hidden_cp_fw = rnn_hidden_cp_fw.view(-1, self.embed_cp_dim+1)\n",
        "      rnn_hidden_dp_bw = rnn_hidden_dp_bw.view(-1, self.embed_dp_dim+1)\n",
        "      rnn_hidden_cp_bw = rnn_hidden_cp_bw.view(-1, self.embed_cp_dim+1)\n",
        "      ## concatenate forward and backward: batch_size x 2*(embedding_dim+1)\n",
        "      rnn_hidden_dp = torch.cat((rnn_hidden_dp_fw, rnn_hidden_dp_bw), dim=-1)\n",
        "      rnn_hidden_cp = torch.cat((rnn_hidden_cp_fw, rnn_hidden_cp_bw), dim=-1)\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(rnn_hidden_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(rnn_hidden_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []\n",
        "\n",
        "\n",
        "elif net_variant == 'birnn_concat_time_delta_attention':\n",
        "  # GRU\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(np.ceil(num_dp_codes**0.25))\n",
        "      self.embed_cp_dim = int(np.ceil(num_cp_codes**0.25))\n",
        "\n",
        "      # Embedding layers\n",
        "      self.embed_dp = nn.Embedding(num_embeddings=num_dp_codes, embedding_dim=self.embed_dp_dim, padding_idx=0)\n",
        "      self.embed_cp = nn.Embedding(num_embeddings=num_cp_codes, embedding_dim=self.embed_cp_dim, padding_idx=0)\n",
        "\n",
        "      # GRU layers\n",
        "      self.gru_dp_fw = nn.GRU(input_size=self.embed_dp_dim+1, hidden_size=self.embed_dp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_fw = nn.GRU(input_size=self.embed_cp_dim+1, hidden_size=self.embed_cp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_dp_bw = nn.GRU(input_size=self.embed_dp_dim+1, hidden_size=self.embed_dp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_bw = nn.GRU(input_size=self.embed_cp_dim+1, hidden_size=self.embed_cp_dim+1, num_layers=1, batch_first=True)\n",
        "\n",
        "      # Attention layers\n",
        "      self.attention_dp = Attention(embedding_dim=2*(self.embed_dp_dim+1)) #+1 for the concatenated time\n",
        "      self.attention_cp = Attention(embedding_dim=2*(self.embed_cp_dim+1))\n",
        "            \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(2*(self.embed_dp_dim+1), 1)\n",
        "      self.fc_cp  = nn.Linear(2*(self.embed_cp_dim+1), 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Compute time delta\n",
        "      ## output dim: batch_size x seq_len\n",
        "      dp_t_delta_fw = abs_time_to_delta(dp_t)\n",
        "      cp_t_delta_fw = abs_time_to_delta(cp_t)\n",
        "      dp_t_delta_bw = abs_time_to_delta(torch.flip(dp_t, [1]))\n",
        "      cp_t_delta_bw = abs_time_to_delta(torch.flip(cp_t, [1]))    \n",
        "    \n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp_fw = self.embed_dp(dp)\n",
        "      embedded_cp_fw = self.embed_cp(cp)\n",
        "      embedded_dp_bw = torch.flip(embedded_dp_fw, [1])\n",
        "      embedded_cp_bw = torch.flip(embedded_cp_fw, [1])\n",
        "      \n",
        "      # Concatate with time\n",
        "      ## output dim: batch_size x seq_len x (embedding_dim+1)\n",
        "      concat_dp_fw = torch.cat((embedded_dp_fw, torch.unsqueeze(dp_t_delta_fw, dim=-1)), dim=-1)\n",
        "      concat_cp_fw = torch.cat((embedded_cp_fw, torch.unsqueeze(cp_t_delta_fw, dim=-1)), dim=-1)\n",
        "      concat_dp_bw = torch.cat((embedded_dp_bw, torch.unsqueeze(dp_t_delta_bw, dim=-1)), dim=-1)\n",
        "      concat_cp_bw = torch.cat((embedded_cp_bw, torch.unsqueeze(cp_t_delta_bw, dim=-1)), dim=-1)\n",
        "      ## Dropout\n",
        "      concat_dp_fw = self.dropout(concat_dp_fw)\n",
        "      concat_cp_fw = self.dropout(concat_cp_fw)\n",
        "      concat_dp_bw = self.dropout(concat_dp_bw)\n",
        "      concat_cp_bw = self.dropout(concat_cp_bw)\n",
        "      \n",
        "      # GRU\n",
        "      ## output dim rnn:        batch_size x seq_len x (embedding_dim+1)\n",
        "      ## output dim rnn_hidden: batch_size x 1 x (embedding_dim+1)\n",
        "      rnn_dp_fw, rnn_hidden_dp_fw = self.gru_dp_fw(concat_dp_fw)\n",
        "      rnn_cp_fw, rnn_hidden_cp_fw = self.gru_cp_fw(concat_cp_fw)\n",
        "      rnn_dp_bw, rnn_hidden_dp_bw = self.gru_dp_bw(concat_dp_bw)\n",
        "      rnn_cp_bw, rnn_hidden_cp_bw = self.gru_cp_bw(concat_cp_bw)      \n",
        "      # concatenate forward and backward\n",
        "      ## output dim: batch_size x seq_len x 2*(embedding_dim+1)\n",
        "      rnn_dp = torch.cat((rnn_dp_fw, torch.flip(rnn_dp_bw, [1])), dim=-1)\n",
        "      rnn_cp = torch.cat((rnn_cp_fw, torch.flip(rnn_cp_bw, [1])), dim=-1)\n",
        "\n",
        "      # Attention\n",
        "      ## output dim: batch_size x 2*(embedding_dim+1)\n",
        "      attended_dp, weights_dp = self.attention_dp(rnn_dp, (dp > 0).float())\n",
        "      attended_cp, weights_cp = self.attention_cp(rnn_cp, (cp > 0).float())\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(attended_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(attended_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []\n",
        "\n",
        "\n",
        "elif net_variant == 'birnn_time_decay':\n",
        "  # GRU\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(np.ceil(num_dp_codes**0.25))+1\n",
        "      self.embed_cp_dim = int(np.ceil(num_cp_codes**0.25))+1\n",
        "\n",
        "      # Embedding layers\n",
        "      self.embed_dp = nn.Embedding(num_embeddings=num_dp_codes, embedding_dim=self.embed_dp_dim, padding_idx=0)\n",
        "      self.embed_cp = nn.Embedding(num_embeddings=num_cp_codes, embedding_dim=self.embed_cp_dim, padding_idx=0)\n",
        "\n",
        "      # GRU layers\n",
        "      self.gru_dp_fw = GRUExponentialDecay(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim)\n",
        "      self.gru_cp_fw = GRUExponentialDecay(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim)\n",
        "      self.gru_dp_bw = GRUExponentialDecay(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim)\n",
        "      self.gru_cp_bw = GRUExponentialDecay(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim)\n",
        "      \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(2*self.embed_dp_dim, 1)\n",
        "      self.fc_cp  = nn.Linear(2*self.embed_cp_dim, 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Compute time delta\n",
        "      ## output dim: batch_size x seq_len\n",
        "      dp_t_delta_fw = abs_time_to_delta(dp_t)\n",
        "      cp_t_delta_fw = abs_time_to_delta(cp_t)\n",
        "      dp_t_delta_bw = abs_time_to_delta(torch.flip(dp_t, [1]))\n",
        "      cp_t_delta_bw = abs_time_to_delta(torch.flip(cp_t, [1]))    \n",
        "    \n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp_fw = self.embed_dp(dp)\n",
        "      embedded_cp_fw = self.embed_cp(cp)\n",
        "      embedded_dp_bw = torch.flip(embedded_dp_fw, [1])\n",
        "      embedded_cp_bw = torch.flip(embedded_cp_fw, [1])\n",
        "      ## Dropout\n",
        "      embedded_dp_fw = self.dropout(embedded_dp_fw)\n",
        "      embedded_cp_fw = self.dropout(embedded_cp_fw)\n",
        "      embedded_dp_bw = self.dropout(embedded_dp_bw)\n",
        "      embedded_cp_bw = self.dropout(embedded_cp_bw)\n",
        "      \n",
        "      # GRU\n",
        "      ## output dim rnn:        batch_size x seq_len x embedding_dim\n",
        "      rnn_dp_fw = self.gru_dp_fw(embedded_dp_fw, dp_t_delta_fw)\n",
        "      rnn_cp_fw = self.gru_cp_fw(embedded_cp_fw, cp_t_delta_fw)\n",
        "      rnn_dp_bw = self.gru_dp_bw(embedded_dp_bw, dp_t_delta_bw)\n",
        "      rnn_cp_bw = self.gru_cp_bw(embedded_cp_bw, cp_t_delta_bw)      \n",
        "      ## output dim rnn_hidden: batch_size x embedding_dim\n",
        "      rnn_dp_fw = rnn_dp_fw[:,-1,:]\n",
        "      rnn_cp_fw = rnn_cp_fw[:,-1,:]\n",
        "      rnn_dp_bw = rnn_dp_bw[:,-1,:]\n",
        "      rnn_cp_bw = rnn_cp_bw[:,-1,:]\n",
        "      ## concatenate forward and backward: batch_size x 2*embedding_dim\n",
        "      rnn_dp = torch.cat((rnn_dp_fw, rnn_dp_bw), dim=-1)\n",
        "      rnn_cp = torch.cat((rnn_cp_fw, rnn_cp_bw), dim=-1)\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(rnn_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(rnn_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []\n",
        "\n",
        "\n",
        "elif net_variant == 'birnn_time_decay_attention':\n",
        "  # GRU\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(np.ceil(num_dp_codes**0.25))+1\n",
        "      self.embed_cp_dim = int(np.ceil(num_cp_codes**0.25))+1\n",
        "\n",
        "      # Embedding layers\n",
        "      self.embed_dp = nn.Embedding(num_embeddings=num_dp_codes, embedding_dim=self.embed_dp_dim, padding_idx=0)\n",
        "      self.embed_cp = nn.Embedding(num_embeddings=num_cp_codes, embedding_dim=self.embed_cp_dim, padding_idx=0)\n",
        "\n",
        "      # GRU layers\n",
        "      self.gru_dp_fw = GRUExponentialDecay(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim)\n",
        "      self.gru_cp_fw = GRUExponentialDecay(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim)\n",
        "      self.gru_dp_bw = GRUExponentialDecay(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim)\n",
        "      self.gru_cp_bw = GRUExponentialDecay(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim)\n",
        "\n",
        "      # Attention layers\n",
        "      self.attention_dp = Attention(embedding_dim=2*self.embed_dp_dim)\n",
        "      self.attention_cp = Attention(embedding_dim=2*self.embed_cp_dim)\n",
        "      \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(2*self.embed_dp_dim, 1)\n",
        "      self.fc_cp  = nn.Linear(2*self.embed_cp_dim, 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Compute time delta\n",
        "      ## output dim: batch_size x seq_len\n",
        "      dp_t_delta_fw = abs_time_to_delta(dp_t)\n",
        "      cp_t_delta_fw = abs_time_to_delta(cp_t)\n",
        "      dp_t_delta_bw = abs_time_to_delta(torch.flip(dp_t, [1]))\n",
        "      cp_t_delta_bw = abs_time_to_delta(torch.flip(cp_t, [1]))    \n",
        "    \n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp_fw = self.embed_dp(dp)\n",
        "      embedded_cp_fw = self.embed_cp(cp)\n",
        "      embedded_dp_bw = torch.flip(embedded_dp_fw, [1])\n",
        "      embedded_cp_bw = torch.flip(embedded_cp_fw, [1])\n",
        "      ## Dropout\n",
        "      embedded_dp_fw = self.dropout(embedded_dp_fw)\n",
        "      embedded_cp_fw = self.dropout(embedded_cp_fw)\n",
        "      embedded_dp_bw = self.dropout(embedded_dp_bw)\n",
        "      embedded_cp_bw = self.dropout(embedded_cp_bw)\n",
        "      \n",
        "      # GRU\n",
        "      ## output dim rnn:        batch_size x seq_len x embedding_dim\n",
        "      rnn_dp_fw = self.gru_dp_fw(embedded_dp_fw, dp_t_delta_fw)\n",
        "      rnn_cp_fw = self.gru_cp_fw(embedded_cp_fw, cp_t_delta_fw)\n",
        "      rnn_dp_bw = self.gru_dp_bw(embedded_dp_bw, dp_t_delta_bw)\n",
        "      rnn_cp_bw = self.gru_cp_bw(embedded_cp_bw, cp_t_delta_bw)\n",
        "      # concatenate forward and backward\n",
        "      ## output dim: batch_size x seq_len x 2*embedding_dim\n",
        "      rnn_dp = torch.cat((rnn_dp_fw, torch.flip(rnn_dp_bw, [1])), dim=-1)\n",
        "      rnn_cp = torch.cat((rnn_cp_fw, torch.flip(rnn_cp_bw, [1])), dim=-1)\n",
        "\n",
        "      # Attention\n",
        "      ## output dim: batch_size x 2*embedding_dim\n",
        "      attended_dp, weights_dp = self.attention_dp(rnn_dp, (dp > 0).float())\n",
        "      attended_cp, weights_cp = self.attention_cp(rnn_cp, (cp > 0).float())\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(attended_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(attended_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []\n",
        "\n",
        "\n",
        "elif net_variant == 'attention_concat_time':\n",
        "  # Attention Only\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(2*np.ceil(num_dp_codes**0.25))\n",
        "      self.embed_cp_dim = int(2*np.ceil(num_cp_codes**0.25))\n",
        "\n",
        "      # Embedding layers\n",
        "      self.embed_dp = nn.Embedding(num_embeddings=num_dp_codes, embedding_dim=self.embed_dp_dim, padding_idx=0)\n",
        "      self.embed_cp = nn.Embedding(num_embeddings=num_cp_codes, embedding_dim=self.embed_cp_dim, padding_idx=0)\n",
        "      \n",
        "      # Attention layers\n",
        "      self.attention_dp = Attention(embedding_dim=self.embed_dp_dim+1) #+1 for the concatenated time\n",
        "      self.attention_cp = Attention(embedding_dim=self.embed_cp_dim+1)\n",
        "      \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(self.embed_dp_dim+1, 1)\n",
        "      self.fc_cp  = nn.Linear(self.embed_cp_dim+1, 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp = self.embed_dp(dp)\n",
        "      embedded_cp = self.embed_cp(cp)\n",
        "      \n",
        "      # Concatate with time\n",
        "      ## output dim: batch_size x seq_len x (embedding_dim+1)\n",
        "      concat_dp = torch.cat((embedded_dp, torch.unsqueeze(dp_t, dim=-1)), dim=-1)\n",
        "      concat_cp = torch.cat((embedded_cp, torch.unsqueeze(cp_t, dim=-1)), dim=-1)\n",
        "      ## Dropout\n",
        "      concat_dp = self.dropout(concat_dp)\n",
        "      concat_cp = self.dropout(concat_cp)\n",
        "      \n",
        "      # Attention\n",
        "      ## output dim: batch_size x (embedding_dim+1)\n",
        "      attended_dp, weights_dp = self.attention_dp(concat_dp, (dp > 0).float())\n",
        "      attended_cp, weights_cp = self.attention_cp(concat_cp, (cp > 0).float())\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(attended_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(attended_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []\n",
        "\n",
        "\n",
        "elif net_variant == 'ode_birnn':\n",
        "  # Attention Only\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(np.ceil(num_dp_codes**0.25))+1\n",
        "      self.embed_cp_dim = int(np.ceil(num_cp_codes**0.25))+1\n",
        "\n",
        "      # Embedding layers\n",
        "      self.embed_dp = nn.Embedding(num_embeddings=num_dp_codes, embedding_dim=self.embed_dp_dim, padding_idx=0)\n",
        "      self.embed_cp = nn.Embedding(num_embeddings=num_cp_codes, embedding_dim=self.embed_cp_dim, padding_idx=0)\n",
        "      \n",
        "      # ODE layers\n",
        "      self.device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')\n",
        "      self.ode_dp = ODENet(self.device, self.embed_dp_dim, self.embed_dp_dim, output_dim=self.embed_dp_dim, augment_dim=0, time_dependent=False, non_linearity='softplus', tol=1e-3, adjoint=True)\n",
        "      self.ode_cp = ODENet(self.device, self.embed_cp_dim, self.embed_cp_dim, output_dim=self.embed_cp_dim, augment_dim=0, time_dependent=False, non_linearity='softplus', tol=1e-3, adjoint=True)\n",
        "\n",
        "      # GRU layers\n",
        "      self.gru_dp_fw = nn.GRU(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_fw = nn.GRU(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim, num_layers=1, batch_first=True)\n",
        "      self.gru_dp_bw = nn.GRU(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_bw = nn.GRU(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim, num_layers=1, batch_first=True)\n",
        "      \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(2*self.embed_dp_dim, 1)\n",
        "      self.fc_cp  = nn.Linear(2*self.embed_cp_dim, 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp = self.embed_dp(dp)\n",
        "      embedded_cp = self.embed_cp(cp)\n",
        "      \n",
        "      # ODE\n",
        "      ## Round times\n",
        "      dp_t = torch.round(100*dp_t)/100\n",
        "      cp_t = torch.round(100*cp_t)/100\n",
        "      \n",
        "      embedded_dp_long = embedded_dp.view(-1, self.embed_dp_dim)\n",
        "      dp_t_long = dp_t.view(-1)\n",
        "      dp_t_long_unique, inverse_indices = torch.unique(dp_t_long, sorted=True, return_inverse=True)\n",
        "      ode_dp_long = self.ode_dp(embedded_dp_long, dp_t_long_unique)\n",
        "      ode_dp_long = ode_dp_long[inverse_indices, torch.arange(0, inverse_indices.size(0)), :]\n",
        "      ode_dp = ode_dp_long.view(dp.size(0), dp.size(1), self.embed_dp_dim)\n",
        "\n",
        "      embedded_cp_long = embedded_cp.view(-1, self.embed_cp_dim)\n",
        "      cp_t_long = cp_t.view(-1)\n",
        "      cp_t_long_unique, inverse_indices = torch.unique(cp_t_long, sorted=True, return_inverse=True)\n",
        "      ode_cp_long = self.ode_cp(embedded_cp_long, cp_t_long_unique)\n",
        "      ode_cp_long = ode_cp_long[inverse_indices, torch.arange(0, inverse_indices.size(0)), :]\n",
        "      ode_cp = ode_cp_long.view(cp.size(0), cp.size(1), self.embed_cp_dim)\n",
        "      \n",
        "      ## Dropout\n",
        "      ode_dp = self.dropout(ode_dp)\n",
        "      ode_cp = self.dropout(ode_cp)\n",
        "\n",
        "      # Forward and backward sequences\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      ode_dp_fw = ode_dp\n",
        "      ode_cp_fw = ode_cp\n",
        "      ode_dp_bw = torch.flip(ode_dp_fw, [1])\n",
        "      ode_cp_bw = torch.flip(ode_cp_fw, [1])\n",
        "      \n",
        "      # GRU\n",
        "      ## output dim rnn:        batch_size x seq_len x embedding_dim\n",
        "      ## output dim rnn_hidden: batch_size x 1 x embedding_dim\n",
        "      rnn_dp_fw, rnn_hidden_dp_fw = self.gru_dp_fw(ode_dp_fw)\n",
        "      rnn_cp_fw, rnn_hidden_cp_fw = self.gru_cp_fw(ode_cp_fw)\n",
        "      rnn_dp_bw, rnn_hidden_dp_bw = self.gru_dp_bw(ode_dp_bw)\n",
        "      rnn_cp_bw, rnn_hidden_cp_bw = self.gru_cp_bw(ode_cp_bw)      \n",
        "      ## output dim rnn_hidden: batch_size x embedding_dim\n",
        "      rnn_hidden_dp_fw = rnn_hidden_dp_fw.view(-1, self.embed_dp_dim)\n",
        "      rnn_hidden_cp_fw = rnn_hidden_cp_fw.view(-1, self.embed_cp_dim)\n",
        "      rnn_hidden_dp_bw = rnn_hidden_dp_bw.view(-1, self.embed_dp_dim)\n",
        "      rnn_hidden_cp_bw = rnn_hidden_cp_bw.view(-1, self.embed_cp_dim)\n",
        "      ## concatenate forward and backward: batch_size x 2*embedding_dim\n",
        "      rnn_hidden_dp = torch.cat((rnn_hidden_dp_fw, rnn_hidden_dp_bw), dim=-1)\n",
        "      rnn_hidden_cp = torch.cat((rnn_hidden_cp_fw, rnn_hidden_cp_bw), dim=-1)\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(rnn_hidden_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(rnn_hidden_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []\n",
        "\n",
        "\n",
        "elif net_variant == 'ode_birnn_attention':\n",
        "  # Attention Only\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(np.ceil(num_dp_codes**0.25))+1\n",
        "      self.embed_cp_dim = int(np.ceil(num_cp_codes**0.25))+1\n",
        "\n",
        "      # Embedding layers\n",
        "      self.embed_dp = nn.Embedding(num_embeddings=num_dp_codes, embedding_dim=self.embed_dp_dim, padding_idx=0)\n",
        "      self.embed_cp = nn.Embedding(num_embeddings=num_cp_codes, embedding_dim=self.embed_cp_dim, padding_idx=0)\n",
        "      \n",
        "      # ODE layers\n",
        "      self.device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')\n",
        "      self.ode_dp = ODENet(self.device, self.embed_dp_dim, self.embed_dp_dim, output_dim=self.embed_dp_dim, augment_dim=0, time_dependent=False, non_linearity='softplus', tol=1e-3, adjoint=True)\n",
        "      self.ode_cp = ODENet(self.device, self.embed_cp_dim, self.embed_cp_dim, output_dim=self.embed_cp_dim, augment_dim=0, time_dependent=False, non_linearity='softplus', tol=1e-3, adjoint=True)\n",
        "\n",
        "      # GRU layers\n",
        "      self.gru_dp_fw = nn.GRU(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_fw = nn.GRU(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim, num_layers=1, batch_first=True)\n",
        "      self.gru_dp_bw = nn.GRU(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_bw = nn.GRU(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim, num_layers=1, batch_first=True)\n",
        "\n",
        "      # Attention layers\n",
        "      self.attention_dp = Attention(embedding_dim=2*self.embed_dp_dim)\n",
        "      self.attention_cp = Attention(embedding_dim=2*self.embed_cp_dim)\n",
        "      \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(2*self.embed_dp_dim, 1)\n",
        "      self.fc_cp  = nn.Linear(2*self.embed_cp_dim, 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp = self.embed_dp(dp)\n",
        "      embedded_cp = self.embed_cp(cp)\n",
        "      \n",
        "      # ODE\n",
        "      ## Round times\n",
        "      dp_t = torch.round(100*dp_t)/100\n",
        "      cp_t = torch.round(100*cp_t)/100\n",
        "      \n",
        "      embedded_dp_long = embedded_dp.view(-1, self.embed_dp_dim)\n",
        "      dp_t_long = dp_t.view(-1)\n",
        "      dp_t_long_unique, inverse_indices = torch.unique(dp_t_long, sorted=True, return_inverse=True)\n",
        "      ode_dp_long = self.ode_dp(embedded_dp_long, dp_t_long_unique)\n",
        "      ode_dp_long = ode_dp_long[inverse_indices, torch.arange(0, inverse_indices.size(0)), :]\n",
        "      ode_dp = ode_dp_long.view(dp.size(0), dp.size(1), self.embed_dp_dim)\n",
        "\n",
        "      embedded_cp_long = embedded_cp.view(-1, self.embed_cp_dim)\n",
        "      cp_t_long = cp_t.view(-1)\n",
        "      cp_t_long_unique, inverse_indices = torch.unique(cp_t_long, sorted=True, return_inverse=True)\n",
        "      ode_cp_long = self.ode_cp(embedded_cp_long, cp_t_long_unique)\n",
        "      ode_cp_long = ode_cp_long[inverse_indices, torch.arange(0, inverse_indices.size(0)), :]\n",
        "      ode_cp = ode_cp_long.view(cp.size(0), cp.size(1), self.embed_cp_dim)\n",
        "      \n",
        "      ## Dropout\n",
        "      ode_dp = self.dropout(ode_dp)\n",
        "      ode_cp = self.dropout(ode_cp)\n",
        "\n",
        "      # Forward and backward sequences\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      ode_dp_fw = ode_dp\n",
        "      ode_cp_fw = ode_cp\n",
        "      ode_dp_bw = torch.flip(ode_dp_fw, [1])\n",
        "      ode_cp_bw = torch.flip(ode_cp_fw, [1])\n",
        "      \n",
        "      # GRU\n",
        "      ## output dim rnn:        batch_size x seq_len x embedding_dim\n",
        "      ## output dim rnn_hidden: batch_size x 1 x embedding_dim\n",
        "      rnn_dp_fw, rnn_hidden_dp_fw = self.gru_dp_fw(ode_dp_fw)\n",
        "      rnn_cp_fw, rnn_hidden_cp_fw = self.gru_cp_fw(ode_cp_fw)\n",
        "      rnn_dp_bw, rnn_hidden_dp_bw = self.gru_dp_bw(ode_dp_bw)\n",
        "      rnn_cp_bw, rnn_hidden_cp_bw = self.gru_cp_bw(ode_cp_bw)      \n",
        "      # concatenate forward and backward\n",
        "      ## output dim: batch_size x seq_len x 2*embedding_dim\n",
        "      rnn_dp = torch.cat((rnn_dp_fw, torch.flip(rnn_dp_bw, [1])), dim=-1)\n",
        "      rnn_cp = torch.cat((rnn_cp_fw, torch.flip(rnn_cp_bw, [1])), dim=-1)\n",
        "\n",
        "      # Attention\n",
        "      ## output dim: batch_size x 2*embedding_dim\n",
        "      attended_dp, weights_dp = self.attention_dp(rnn_dp, (dp > 0).float())\n",
        "      attended_cp, weights_cp = self.attention_cp(rnn_cp, (cp > 0).float())\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(attended_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(attended_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []\n",
        "\n",
        "\n",
        "elif net_variant == 'ode_attention':\n",
        "  # Attention Only\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(2*np.ceil(num_dp_codes**0.25))+1\n",
        "      self.embed_cp_dim = int(2*np.ceil(num_cp_codes**0.25))+1\n",
        "\n",
        "      # Embedding layers\n",
        "      self.embed_dp = nn.Embedding(num_embeddings=num_dp_codes, embedding_dim=self.embed_dp_dim, padding_idx=0)\n",
        "      self.embed_cp = nn.Embedding(num_embeddings=num_cp_codes, embedding_dim=self.embed_cp_dim, padding_idx=0)\n",
        "      \n",
        "      # ODE layers\n",
        "      self.device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')\n",
        "      self.ode_dp = ODENet(self.device, self.embed_dp_dim, self.embed_dp_dim, output_dim=self.embed_dp_dim, augment_dim=0, time_dependent=False, non_linearity='softplus', tol=1e-3, adjoint=True)\n",
        "      self.ode_cp = ODENet(self.device, self.embed_cp_dim, self.embed_cp_dim, output_dim=self.embed_cp_dim, augment_dim=0, time_dependent=False, non_linearity='softplus', tol=1e-3, adjoint=True)\n",
        "      \n",
        "      # Attention layers\n",
        "      self.attention_dp = Attention(embedding_dim=self.embed_dp_dim)\n",
        "      self.attention_cp = Attention(embedding_dim=self.embed_cp_dim)\n",
        "      \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(self.embed_dp_dim, 1)\n",
        "      self.fc_cp  = nn.Linear(self.embed_cp_dim, 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp = self.embed_dp(dp)\n",
        "      embedded_cp = self.embed_cp(cp)\n",
        "      \n",
        "      # ODE\n",
        "      ## Round times\n",
        "      dp_t = torch.round(100*dp_t)/100\n",
        "      cp_t = torch.round(100*cp_t)/100\n",
        "      \n",
        "      embedded_dp_long = embedded_dp.view(-1, self.embed_dp_dim)\n",
        "      dp_t_long = dp_t.view(-1)\n",
        "      dp_t_long_unique, inverse_indices = torch.unique(dp_t_long, sorted=True, return_inverse=True)\n",
        "      ode_dp_long = self.ode_dp(embedded_dp_long, dp_t_long_unique)\n",
        "      ode_dp_long = ode_dp_long[inverse_indices, torch.arange(0, inverse_indices.size(0)), :]\n",
        "      ode_dp = ode_dp_long.view(dp.size(0), dp.size(1), self.embed_dp_dim)\n",
        "\n",
        "      embedded_cp_long = embedded_cp.view(-1, self.embed_cp_dim)\n",
        "      cp_t_long = cp_t.view(-1)\n",
        "      cp_t_long_unique, inverse_indices = torch.unique(cp_t_long, sorted=True, return_inverse=True)\n",
        "      ode_cp_long = self.ode_cp(embedded_cp_long, cp_t_long_unique)\n",
        "      ode_cp_long = ode_cp_long[inverse_indices, torch.arange(0, inverse_indices.size(0)), :]\n",
        "      ode_cp = ode_cp_long.view(cp.size(0), cp.size(1), self.embed_cp_dim)\n",
        "\n",
        "      ## Dropout\n",
        "      ode_dp = self.dropout(ode_dp)\n",
        "      ode_cp = self.dropout(ode_cp)\n",
        "      \n",
        "      # Attention\n",
        "      ## output dim: batch_size x (embedding_dim+1)\n",
        "      attended_dp, weights_dp = self.attention_dp(ode_dp, (dp > 0).float())\n",
        "      attended_cp, weights_cp = self.attention_cp(ode_cp, (cp > 0).float())\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(attended_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(attended_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []\n",
        "\n",
        "\n",
        "elif net_variant == 'birnn_ode_decay':\n",
        "  # GRU\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(np.ceil(num_dp_codes**0.25))+1\n",
        "      self.embed_cp_dim = int(np.ceil(num_cp_codes**0.25))+1\n",
        "\n",
        "      # Embedding layers\n",
        "      self.embed_dp = nn.Embedding(num_embeddings=num_dp_codes, embedding_dim=self.embed_dp_dim, padding_idx=0)\n",
        "      self.embed_cp = nn.Embedding(num_embeddings=num_cp_codes, embedding_dim=self.embed_cp_dim, padding_idx=0)\n",
        "\n",
        "      # GRU layers\n",
        "      self.gru_dp_fw = GRUOdeDecay(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim)\n",
        "      self.gru_cp_fw = GRUOdeDecay(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim)\n",
        "      self.gru_dp_bw = GRUOdeDecay(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim)\n",
        "      self.gru_cp_bw = GRUOdeDecay(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim)\n",
        "      \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(2*self.embed_dp_dim, 1)\n",
        "      self.fc_cp  = nn.Linear(2*self.embed_cp_dim, 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Compute time delta\n",
        "      ## output dim: batch_size x seq_len\n",
        "      dp_t_delta_fw = abs_time_to_delta(dp_t)\n",
        "      cp_t_delta_fw = abs_time_to_delta(cp_t)\n",
        "      ## Round\n",
        "      dp_t_delta_fw = torch.round(100*dp_t_delta_fw)/100\n",
        "      cp_t_delta_fw = torch.round(100*cp_t_delta_fw)/100            \n",
        "      dp_t_delta_bw = abs_time_to_delta(torch.flip(dp_t, [1]))\n",
        "      cp_t_delta_bw = abs_time_to_delta(torch.flip(cp_t, [1]))    \n",
        "    \n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp_fw = self.embed_dp(dp)\n",
        "      embedded_cp_fw = self.embed_cp(cp)\n",
        "      embedded_dp_bw = torch.flip(embedded_dp_fw, [1])\n",
        "      embedded_cp_bw = torch.flip(embedded_cp_fw, [1])\n",
        "      ## Dropout\n",
        "      embedded_dp_fw = self.dropout(embedded_dp_fw)\n",
        "      embedded_cp_fw = self.dropout(embedded_cp_fw)\n",
        "      embedded_dp_bw = self.dropout(embedded_dp_bw)\n",
        "      embedded_cp_bw = self.dropout(embedded_cp_bw)\n",
        "      \n",
        "      # GRU\n",
        "      ## output dim rnn:        batch_size x seq_len x embedding_dim\n",
        "      rnn_dp_fw = self.gru_dp_fw(embedded_dp_fw, dp_t_delta_fw)\n",
        "      rnn_cp_fw = self.gru_cp_fw(embedded_cp_fw, cp_t_delta_fw)\n",
        "      rnn_dp_bw = self.gru_dp_bw(embedded_dp_bw, dp_t_delta_bw)\n",
        "      rnn_cp_bw = self.gru_cp_bw(embedded_cp_bw, cp_t_delta_bw)      \n",
        "      ## output dim rnn_hidden: batch_size x embedding_dim\n",
        "      rnn_dp_fw = rnn_dp_fw[:,-1,:]\n",
        "      rnn_cp_fw = rnn_cp_fw[:,-1,:]\n",
        "      rnn_dp_bw = rnn_dp_bw[:,-1,:]\n",
        "      rnn_cp_bw = rnn_cp_bw[:,-1,:]\n",
        "      ## concatenate forward and backward: batch_size x 2*embedding_dim\n",
        "      rnn_dp = torch.cat((rnn_dp_fw, rnn_dp_bw), dim=-1)\n",
        "      rnn_cp = torch.cat((rnn_cp_fw, rnn_cp_bw), dim=-1)\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(rnn_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(rnn_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []      \n",
        "      \n",
        "\n",
        "elif net_variant == 'birnn_ode_decay_attention':\n",
        "  # GRU\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(np.ceil(num_dp_codes**0.25))+1\n",
        "      self.embed_cp_dim = int(np.ceil(num_cp_codes**0.25))+1\n",
        "\n",
        "      # Embedding layers\n",
        "      self.embed_dp = nn.Embedding(num_embeddings=num_dp_codes, embedding_dim=self.embed_dp_dim, padding_idx=0)\n",
        "      self.embed_cp = nn.Embedding(num_embeddings=num_cp_codes, embedding_dim=self.embed_cp_dim, padding_idx=0)\n",
        "\n",
        "      # GRU layers\n",
        "      self.gru_dp_fw = GRUOdeDecay(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim)\n",
        "      self.gru_cp_fw = GRUOdeDecay(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim)\n",
        "      self.gru_dp_bw = GRUOdeDecay(input_size=self.embed_dp_dim, hidden_size=self.embed_dp_dim)\n",
        "      self.gru_cp_bw = GRUOdeDecay(input_size=self.embed_cp_dim, hidden_size=self.embed_cp_dim)\n",
        "\n",
        "      # Attention layers\n",
        "      self.attention_dp = Attention(embedding_dim=2*self.embed_dp_dim)\n",
        "      self.attention_cp = Attention(embedding_dim=2*self.embed_cp_dim)\n",
        "      \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(2*self.embed_dp_dim, 1)\n",
        "      self.fc_cp  = nn.Linear(2*self.embed_cp_dim, 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Compute time delta\n",
        "      ## output dim: batch_size x seq_len\n",
        "      dp_t_delta_fw = abs_time_to_delta(dp_t)\n",
        "      cp_t_delta_fw = abs_time_to_delta(cp_t)\n",
        "      ## Round\n",
        "      dp_t_delta_fw = torch.round(100*dp_t_delta_fw)/100\n",
        "      cp_t_delta_fw = torch.round(100*cp_t_delta_fw)/100      \n",
        "      dp_t_delta_bw = abs_time_to_delta(torch.flip(dp_t, [1]))\n",
        "      cp_t_delta_bw = abs_time_to_delta(torch.flip(cp_t, [1]))    \n",
        "    \n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp_fw = self.embed_dp(dp)\n",
        "      embedded_cp_fw = self.embed_cp(cp)\n",
        "      embedded_dp_bw = torch.flip(embedded_dp_fw, [1])\n",
        "      embedded_cp_bw = torch.flip(embedded_cp_fw, [1])\n",
        "      ## Dropout\n",
        "      embedded_dp_fw = self.dropout(embedded_dp_fw)\n",
        "      embedded_cp_fw = self.dropout(embedded_cp_fw)\n",
        "      embedded_dp_bw = self.dropout(embedded_dp_bw)\n",
        "      embedded_cp_bw = self.dropout(embedded_cp_bw)\n",
        "      \n",
        "      # GRU\n",
        "      ## output dim rnn:        batch_size x seq_len x embedding_dim\n",
        "      rnn_dp_fw = self.gru_dp_fw(embedded_dp_fw, dp_t_delta_fw)\n",
        "      rnn_cp_fw = self.gru_cp_fw(embedded_cp_fw, cp_t_delta_fw)\n",
        "      rnn_dp_bw = self.gru_dp_bw(embedded_dp_bw, dp_t_delta_bw)\n",
        "      rnn_cp_bw = self.gru_cp_bw(embedded_cp_bw, cp_t_delta_bw)\n",
        "      # concatenate forward and backward\n",
        "      ## output dim: batch_size x seq_len x 2*embedding_dim\n",
        "      rnn_dp = torch.cat((rnn_dp_fw, torch.flip(rnn_dp_bw, [1])), dim=-1)\n",
        "      rnn_cp = torch.cat((rnn_cp_fw, torch.flip(rnn_cp_bw, [1])), dim=-1)\n",
        "\n",
        "      # Attention\n",
        "      ## output dim: batch_size x 2*embedding_dim\n",
        "      attended_dp, weights_dp = self.attention_dp(rnn_dp, (dp > 0).float())\n",
        "      attended_cp, weights_cp = self.attention_cp(rnn_cp, (cp > 0).float())\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(attended_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(attended_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []\n",
        "\n",
        "\n",
        "elif net_variant == 'mce_attention':\n",
        "  # Attention Only\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(2*np.ceil(num_dp_codes**0.25))\n",
        "      self.embed_cp_dim = int(2*np.ceil(num_cp_codes**0.25))\n",
        "\n",
        "      # Precomputed embedding weights\n",
        "      self.device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')\n",
        "      self.emb_weight_dp = torch.Tensor(np.load(hp.data_dir + 'emb_weight_dp_13.npy')).to(self.device)\n",
        "      self.emb_weight_cp = torch.Tensor(np.load(hp.data_dir + 'emb_weight_cp_11.npy')).to(self.device)\n",
        "      \n",
        "      # Attention layers\n",
        "      self.attention_dp = Attention(embedding_dim=self.embed_dp_dim+1) #+1 for the concatenated time\n",
        "      self.attention_cp = Attention(embedding_dim=self.embed_cp_dim+1)\n",
        "      \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(self.embed_dp_dim+1, 1)\n",
        "      self.fc_cp  = nn.Linear(self.embed_cp_dim+1, 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp = F.embedding(dp, self.emb_weight_dp, padding_idx=0)\n",
        "      embedded_cp = F.embedding(cp, self.emb_weight_cp, padding_idx=0)\n",
        "      ## Dropout\n",
        "      embedded_dp = self.dropout(embedded_dp)\n",
        "      embedded_cp = self.dropout(embedded_cp)\n",
        "      \n",
        "      # Attention\n",
        "      ## output dim: batch_size x (embedding_dim+1)\n",
        "      attended_dp, weights_dp = self.attention_dp(embedded_dp, (dp > 0).float())\n",
        "      attended_cp, weights_cp = self.attention_cp(embedded_cp, (cp > 0).float())\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(attended_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(attended_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []\n",
        "\n",
        "\n",
        "if net_variant == 'mce_birnn':\n",
        "  # GRU\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(np.ceil(num_dp_codes**0.25))\n",
        "      self.embed_cp_dim = int(np.ceil(num_cp_codes**0.25))\n",
        "\n",
        "      # Precomputed embedding weights\n",
        "      self.device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')\n",
        "      self.emb_weight_dp = torch.Tensor(np.load(hp.data_dir + 'emb_weight_dp_7.npy')).to(self.device)\n",
        "      self.emb_weight_cp = torch.Tensor(np.load(hp.data_dir + 'emb_weight_cp_6.npy')).to(self.device)\n",
        "\n",
        "      # GRU layers\n",
        "      self.gru_dp_fw = nn.GRU(input_size=self.embed_dp_dim+1, hidden_size=self.embed_dp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_fw = nn.GRU(input_size=self.embed_cp_dim+1, hidden_size=self.embed_cp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_dp_bw = nn.GRU(input_size=self.embed_dp_dim+1, hidden_size=self.embed_dp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_bw = nn.GRU(input_size=self.embed_cp_dim+1, hidden_size=self.embed_cp_dim+1, num_layers=1, batch_first=True)\n",
        "      \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(2*(self.embed_dp_dim+1), 1)\n",
        "      self.fc_cp  = nn.Linear(2*(self.embed_cp_dim+1), 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp_fw = F.embedding(dp, self.emb_weight_dp, padding_idx=0)\n",
        "      embedded_cp_fw = F.embedding(cp, self.emb_weight_cp, padding_idx=0)\n",
        "      embedded_dp_bw = torch.flip(embedded_dp_fw, [1])\n",
        "      embedded_cp_bw = torch.flip(embedded_cp_fw, [1])\n",
        "      ## Dropout\n",
        "      embedded_dp_fw = self.dropout(embedded_dp_fw)\n",
        "      embedded_cp_fw = self.dropout(embedded_cp_fw)\n",
        "      embedded_dp_bw = self.dropout(embedded_dp_bw)\n",
        "      embedded_cp_bw = self.dropout(embedded_cp_bw)\n",
        "      \n",
        "      # GRU\n",
        "      ## output dim rnn:        batch_size x seq_len x (embedding_dim+1)\n",
        "      ## output dim rnn_hidden: batch_size x 1 x (embedding_dim+1)\n",
        "      rnn_dp_fw, rnn_hidden_dp_fw = self.gru_dp_fw(embedded_dp_fw)\n",
        "      rnn_cp_fw, rnn_hidden_cp_fw = self.gru_cp_fw(embedded_cp_fw)\n",
        "      rnn_dp_bw, rnn_hidden_dp_bw = self.gru_dp_bw(embedded_dp_bw)\n",
        "      rnn_cp_bw, rnn_hidden_cp_bw = self.gru_cp_bw(embedded_cp_bw)      \n",
        "      ## output dim rnn_hidden: batch_size x (embedding_dim+1)\n",
        "      rnn_hidden_dp_fw = rnn_hidden_dp_fw.view(-1, self.embed_dp_dim+1)\n",
        "      rnn_hidden_cp_fw = rnn_hidden_cp_fw.view(-1, self.embed_cp_dim+1)\n",
        "      rnn_hidden_dp_bw = rnn_hidden_dp_bw.view(-1, self.embed_dp_dim+1)\n",
        "      rnn_hidden_cp_bw = rnn_hidden_cp_bw.view(-1, self.embed_cp_dim+1)\n",
        "      ## concatenate forward and backward: batch_size x 2*(embedding_dim+1)\n",
        "      rnn_hidden_dp = torch.cat((rnn_hidden_dp_fw, rnn_hidden_dp_bw), dim=-1)\n",
        "      rnn_hidden_cp = torch.cat((rnn_hidden_cp_fw, rnn_hidden_cp_bw), dim=-1)\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(rnn_hidden_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(rnn_hidden_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []\n",
        "\n",
        "\n",
        "elif net_variant == 'mce_birnn_attention':\n",
        "  # GRU\n",
        "  class Net(nn.Module):\n",
        "    def __init__(self, num_static, num_dp_codes, num_cp_codes):\n",
        "      super(Net, self).__init__()\n",
        "      \n",
        "      # Embedding dimensions\n",
        "      self.embed_dp_dim = int(np.ceil(num_dp_codes**0.25))\n",
        "      self.embed_cp_dim = int(np.ceil(num_cp_codes**0.25))\n",
        "\n",
        "      # Precomputed embedding weights\n",
        "      self.device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')\n",
        "      #self.device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')\n",
        "      self.emb_weight_dp = torch.Tensor(np.load(hp.data_dir + 'emb_weight_dp_7.npy')).to(self.device)\n",
        "      self.emb_weight_cp = torch.Tensor(np.load(hp.data_dir + 'emb_weight_cp_6.npy')).to(self.device)\n",
        "\n",
        "      # GRU layers\n",
        "      self.gru_dp_fw = nn.GRU(input_size=self.embed_dp_dim+1, hidden_size=self.embed_dp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_fw = nn.GRU(input_size=self.embed_cp_dim+1, hidden_size=self.embed_cp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_dp_bw = nn.GRU(input_size=self.embed_dp_dim+1, hidden_size=self.embed_dp_dim+1, num_layers=1, batch_first=True)\n",
        "      self.gru_cp_bw = nn.GRU(input_size=self.embed_cp_dim+1, hidden_size=self.embed_cp_dim+1, num_layers=1, batch_first=True)\n",
        "\n",
        "      # Attention layers\n",
        "      self.attention_dp = Attention(embedding_dim=2*(self.embed_dp_dim+1)) #+1 for the concatenated time\n",
        "      self.attention_cp = Attention(embedding_dim=2*(self.embed_cp_dim+1))\n",
        "            \n",
        "      # Fully connected output\n",
        "      self.fc_dp  = nn.Linear(2*(self.embed_dp_dim+1), 1)\n",
        "      self.fc_cp  = nn.Linear(2*(self.embed_cp_dim+1), 1)\n",
        "      self.fc_all = nn.Linear(num_static + 2, 1)\n",
        "      \n",
        "      # Others\n",
        "      self.dropout = nn.Dropout(p=0.5)\n",
        "\n",
        "    def forward(self, stat, dp, cp, dp_t, cp_t):\n",
        "      # Embedding\n",
        "      ## output dim: batch_size x seq_len x embedding_dim\n",
        "      embedded_dp_fw = F.embedding(dp, self.emb_weight_dp, padding_idx=0)\n",
        "      embedded_cp_fw = F.embedding(cp, self.emb_weight_cp, padding_idx=0)\n",
        "      embedded_dp_bw = torch.flip(embedded_dp_fw, [1])\n",
        "      embedded_cp_bw = torch.flip(embedded_cp_fw, [1])\n",
        "      ## Dropout\n",
        "      embedded_dp_fw = self.dropout(embedded_dp_fw)\n",
        "      embedded_cp_fw = self.dropout(embedded_cp_fw)\n",
        "      embedded_dp_bw = self.dropout(embedded_dp_bw)\n",
        "      embedded_cp_bw = self.dropout(embedded_cp_bw)\n",
        "      \n",
        "      # GRU\n",
        "      ## output dim rnn:        batch_size x seq_len x (embedding_dim+1)\n",
        "      ## output dim rnn_hidden: batch_size x 1 x (embedding_dim+1)\n",
        "      rnn_dp_fw, rnn_hidden_dp_fw = self.gru_dp_fw(embedded_dp_fw)\n",
        "      rnn_cp_fw, rnn_hidden_cp_fw = self.gru_cp_fw(embedded_cp_fw)\n",
        "      rnn_dp_bw, rnn_hidden_dp_bw = self.gru_dp_bw(embedded_dp_bw)\n",
        "      rnn_cp_bw, rnn_hidden_cp_bw = self.gru_cp_bw(embedded_cp_bw)   \n",
        "      # concatenate forward and backward\n",
        "      ## output dim: batch_size x seq_len x 2*(embedding_dim+1)\n",
        "      rnn_dp = torch.cat((rnn_dp_fw, torch.flip(rnn_dp_bw, [1])), dim=-1)\n",
        "      rnn_cp = torch.cat((rnn_cp_fw, torch.flip(rnn_cp_bw, [1])), dim=-1)\n",
        "\n",
        "      # Attention\n",
        "      ## output dim: batch_size x 2*(embedding_dim+1)\n",
        "      attended_dp, weights_dp = self.attention_dp(rnn_dp, (dp > 0).float())\n",
        "      attended_cp, weights_cp = self.attention_cp(rnn_cp, (cp > 0).float())\n",
        "      \n",
        "      # Scores\n",
        "      score_dp = self.fc_dp(self.dropout(attended_dp))\n",
        "      score_cp = self.fc_cp(self.dropout(attended_cp))\n",
        "\n",
        "      # Concatenate to variable collection\n",
        "      all = torch.cat((stat, score_dp, score_cp), dim=1)\n",
        "      \n",
        "      # Final linear projection\n",
        "      out = self.fc_all(self.dropout(all)).squeeze()\n",
        "\n",
        "      return out, []"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U7-at8g7P4n4"
      },
      "source": [
        "Train"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "background_save": true,
          "base_uri": "https://localhost:8080/"
        },
        "id": "N-hmm2W-P6Nc",
        "outputId": "37b3129c-afbe-480a-cc38-05b24e6edaa8"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Load data...\n",
            "data/\n",
            "-----------------------------------------\n",
            "Train...\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "  0%|          | 0/10 [00:00<?, ?it/s]\n",
            "  0%|          | 0/318 [00:00<?, ?it/s]\u001b[A/usr/local/lib/python3.10/dist-packages/torchdiffeq/_impl/misc.py:11: UserWarning: Euler: Unexpected arguments {'max_num_steps': 1000}\n",
            "  warnings.warn('{}: Unexpected arguments {}'.format(solver.__class__.__name__, unused_kwargs))\n",
            "\n",
            "  0%|          | 1/318 [00:03<18:35,  3.52s/it]\u001b[A\n",
            "  1%|          | 2/318 [00:03<08:29,  1.61s/it]\u001b[A\n",
            "  1%|          | 3/318 [00:04<05:18,  1.01s/it]\u001b[A\n",
            "  1%|▏         | 4/318 [00:04<03:48,  1.37it/s]\u001b[A\n",
            "  2%|▏         | 5/318 [00:04<03:06,  1.68it/s]\u001b[A\n",
            "  2%|▏         | 6/318 [00:05<02:36,  2.00it/s]\u001b[A\n",
            "  2%|▏         | 7/318 [00:05<02:30,  2.06it/s]\u001b[A\n",
            "  3%|▎         | 8/318 [00:05<02:10,  2.37it/s]\u001b[A\n",
            "  3%|▎         | 9/318 [00:06<02:28,  2.08it/s]\u001b[A\n",
            "  3%|▎         | 10/318 [00:06<02:09,  2.38it/s]\u001b[A\n",
            "  3%|▎         | 11/318 [00:06<01:57,  2.62it/s]\u001b[A\n",
            "  4%|▍         | 12/318 [00:07<02:16,  2.25it/s]\u001b[A\n",
            "  4%|▍         | 13/318 [00:07<01:58,  2.57it/s]\u001b[A\n",
            "  4%|▍         | 14/318 [00:08<02:08,  2.37it/s]\u001b[A\n",
            "  5%|▍         | 15/318 [00:08<01:52,  2.69it/s]\u001b[A\n",
            "  5%|▌         | 16/318 [00:08<01:52,  2.67it/s]\u001b[A\n",
            "  5%|▌         | 17/318 [00:09<01:39,  3.01it/s]\u001b[A\n",
            "  6%|▌         | 18/318 [00:09<01:34,  3.18it/s]\u001b[A\n",
            "  6%|▌         | 19/318 [00:09<01:31,  3.25it/s]\u001b[A\n",
            "  6%|▋         | 20/318 [00:10<01:25,  3.51it/s]\u001b[A\n",
            "  7%|▋         | 21/318 [00:10<01:20,  3.70it/s]\u001b[A\n",
            "  7%|▋         | 22/318 [00:10<01:16,  3.87it/s]\u001b[A\n",
            "  7%|▋         | 23/318 [00:10<01:20,  3.68it/s]\u001b[A\n",
            "  8%|▊         | 24/318 [00:11<01:19,  3.69it/s]\u001b[A\n",
            "  8%|▊         | 25/318 [00:11<01:13,  3.99it/s]\u001b[A\n",
            "  8%|▊         | 26/318 [00:11<01:07,  4.35it/s]\u001b[A\n",
            "  8%|▊         | 27/318 [00:11<01:09,  4.18it/s]\u001b[A\n",
            "  9%|▉         | 28/318 [00:11<01:08,  4.20it/s]\u001b[A\n",
            "  9%|▉         | 29/318 [00:12<01:10,  4.11it/s]\u001b[A\n",
            "  9%|▉         | 30/318 [00:12<01:10,  4.08it/s]\u001b[A\n",
            " 10%|▉         | 31/318 [00:12<01:11,  4.03it/s]\u001b[A\n",
            " 10%|█         | 32/318 [00:12<01:12,  3.95it/s]\u001b[A\n",
            " 10%|█         | 33/318 [00:13<01:13,  3.89it/s]\u001b[A\n",
            " 11%|█         | 34/318 [00:13<01:06,  4.27it/s]\u001b[A\n",
            " 11%|█         | 35/318 [00:13<01:03,  4.46it/s]\u001b[A\n",
            " 11%|█▏        | 36/318 [00:13<01:04,  4.36it/s]\u001b[A\n",
            " 12%|█▏        | 37/318 [00:14<01:04,  4.34it/s]\u001b[A\n",
            " 12%|█▏        | 38/318 [00:14<01:10,  3.99it/s]\u001b[A\n",
            " 12%|█▏        | 39/318 [00:14<01:08,  4.07it/s]\u001b[A\n",
            " 13%|█▎        | 40/318 [00:14<01:11,  3.88it/s]\u001b[A\n",
            " 13%|█▎        | 41/318 [00:15<01:08,  4.05it/s]\u001b[A\n",
            " 13%|█▎        | 42/318 [00:15<01:05,  4.23it/s]\u001b[A\n",
            " 14%|█▎        | 43/318 [00:15<01:08,  4.03it/s]\u001b[A\n",
            " 14%|█▍        | 44/318 [00:15<01:09,  3.94it/s]\u001b[A\n",
            " 14%|█▍        | 45/318 [00:16<01:07,  4.04it/s]\u001b[A\n",
            " 14%|█▍        | 46/318 [00:16<01:05,  4.13it/s]\u001b[A\n",
            " 15%|█▍        | 47/318 [00:16<01:07,  4.03it/s]\u001b[A\n",
            " 15%|█▌        | 48/318 [00:16<01:05,  4.13it/s]\u001b[A\n",
            " 15%|█▌        | 49/318 [00:17<01:04,  4.16it/s]\u001b[A\n",
            " 16%|█▌        | 50/318 [00:17<01:06,  4.03it/s]\u001b[A\n",
            " 16%|█▌        | 51/318 [00:17<01:08,  3.92it/s]\u001b[A\n",
            " 16%|█▋        | 52/318 [00:17<01:09,  3.84it/s]\u001b[A\n",
            " 17%|█▋        | 53/318 [00:18<01:12,  3.65it/s]\u001b[A\n",
            " 17%|█▋        | 54/318 [00:18<01:13,  3.58it/s]\u001b[A\n",
            " 17%|█▋        | 55/318 [00:18<01:11,  3.66it/s]\u001b[A\n",
            " 18%|█▊        | 56/318 [00:18<01:08,  3.83it/s]\u001b[A\n",
            " 18%|█▊        | 57/318 [00:19<01:06,  3.94it/s]\u001b[A\n",
            " 18%|█▊        | 58/318 [00:19<01:09,  3.73it/s]\u001b[A\n",
            " 19%|█▊        | 59/318 [00:19<01:07,  3.81it/s]\u001b[A\n",
            " 19%|█▉        | 60/318 [00:19<01:04,  4.02it/s]\u001b[A\n",
            " 19%|█▉        | 61/318 [00:20<01:02,  4.09it/s]\u001b[A\n",
            " 19%|█▉        | 62/318 [00:20<01:05,  3.89it/s]\u001b[A\n",
            " 20%|█▉        | 63/318 [00:20<01:07,  3.76it/s]\u001b[A\n",
            " 20%|██        | 64/318 [00:21<01:09,  3.66it/s]\u001b[A\n",
            " 20%|██        | 65/318 [00:21<01:06,  3.78it/s]\u001b[A\n",
            " 21%|██        | 66/318 [00:21<01:03,  3.94it/s]\u001b[A\n",
            " 21%|██        | 67/318 [00:21<01:03,  3.96it/s]\u001b[A\n",
            " 21%|██▏       | 68/318 [00:22<01:00,  4.11it/s]\u001b[A\n",
            " 22%|██▏       | 69/318 [00:22<01:03,  3.90it/s]\u001b[A\n",
            " 22%|██▏       | 70/318 [00:22<01:07,  3.70it/s]\u001b[A\n",
            " 22%|██▏       | 71/318 [00:22<01:03,  3.92it/s]\u001b[A\n",
            " 23%|██▎       | 72/318 [00:23<01:01,  3.98it/s]\u001b[A\n",
            " 23%|██▎       | 73/318 [00:23<00:58,  4.16it/s]\u001b[A\n",
            " 23%|██▎       | 74/318 [00:23<00:58,  4.14it/s]\u001b[A\n",
            " 24%|██▎       | 75/318 [00:23<00:57,  4.21it/s]\u001b[A\n",
            " 24%|██▍       | 76/318 [00:24<01:00,  4.03it/s]\u001b[A\n",
            " 24%|██▍       | 77/318 [00:24<01:01,  3.94it/s]\u001b[A\n",
            " 25%|██▍       | 78/318 [00:24<01:00,  3.96it/s]\u001b[A\n",
            " 25%|██▍       | 79/318 [00:24<00:57,  4.14it/s]\u001b[A\n",
            " 25%|██▌       | 80/318 [00:24<00:56,  4.25it/s]\u001b[A\n",
            " 25%|██▌       | 81/318 [00:25<00:57,  4.15it/s]\u001b[A\n",
            " 26%|██▌       | 82/318 [00:25<00:56,  4.16it/s]\u001b[A\n",
            " 26%|██▌       | 83/318 [00:25<00:57,  4.07it/s]\u001b[A\n",
            " 26%|██▋       | 84/318 [00:25<00:59,  3.94it/s]\u001b[A\n",
            " 27%|██▋       | 85/318 [00:26<00:58,  3.98it/s]\u001b[A\n",
            " 27%|██▋       | 86/318 [00:26<00:59,  3.93it/s]\u001b[A\n",
            " 27%|██▋       | 87/318 [00:26<00:58,  3.98it/s]\u001b[A\n",
            " 28%|██▊       | 88/318 [00:27<00:59,  3.88it/s]\u001b[A\n",
            " 28%|██▊       | 89/318 [00:27<00:55,  4.14it/s]\u001b[A\n",
            " 28%|██▊       | 90/318 [00:27<00:54,  4.16it/s]\u001b[A\n",
            " 29%|██▊       | 91/318 [00:27<00:55,  4.07it/s]\u001b[A\n",
            " 29%|██▉       | 92/318 [00:27<00:54,  4.16it/s]\u001b[A\n",
            " 29%|██▉       | 93/318 [00:28<00:52,  4.33it/s]\u001b[A\n",
            " 30%|██▉       | 94/318 [00:28<00:52,  4.24it/s]\u001b[A\n",
            " 30%|██▉       | 95/318 [00:28<00:52,  4.22it/s]\u001b[A\n",
            " 30%|███       | 96/318 [00:28<00:49,  4.49it/s]\u001b[A\n",
            " 31%|███       | 97/318 [00:29<00:51,  4.30it/s]\u001b[A\n",
            " 31%|███       | 98/318 [00:29<00:52,  4.16it/s]\u001b[A\n",
            " 31%|███       | 99/318 [00:29<00:56,  3.87it/s]\u001b[A\n",
            " 31%|███▏      | 100/318 [00:29<00:59,  3.67it/s]\u001b[A\n",
            " 32%|███▏      | 101/318 [00:30<00:55,  3.91it/s]\u001b[A\n",
            " 32%|███▏      | 102/318 [00:30<00:56,  3.79it/s]\u001b[A\n",
            " 32%|███▏      | 103/318 [00:30<00:54,  3.91it/s]\u001b[A\n",
            " 33%|███▎      | 104/318 [00:30<00:55,  3.85it/s]\u001b[A\n",
            " 33%|███▎      | 105/318 [00:31<00:53,  3.96it/s]\u001b[A\n",
            " 33%|███▎      | 106/318 [00:31<00:57,  3.66it/s]\u001b[A\n",
            " 34%|███▎      | 107/318 [00:31<01:00,  3.50it/s]\u001b[A\n",
            " 34%|███▍      | 108/318 [00:32<00:57,  3.63it/s]\u001b[A\n",
            " 34%|███▍      | 109/318 [00:32<00:54,  3.83it/s]\u001b[A\n",
            " 35%|███▍      | 110/318 [00:32<00:50,  4.14it/s]\u001b[A\n",
            " 35%|███▍      | 111/318 [00:32<00:54,  3.81it/s]\u001b[A\n",
            " 35%|███▌      | 112/318 [00:33<00:52,  3.92it/s]\u001b[A\n",
            " 36%|███▌      | 113/318 [00:33<00:51,  3.98it/s]\u001b[A\n",
            " 36%|███▌      | 114/318 [00:33<00:48,  4.19it/s]\u001b[A\n",
            " 36%|███▌      | 115/318 [00:33<00:50,  4.04it/s]\u001b[A\n",
            " 36%|███▋      | 116/318 [00:34<00:49,  4.10it/s]\u001b[A\n",
            " 37%|███▋      | 117/318 [00:34<00:47,  4.19it/s]\u001b[A\n",
            " 37%|███▋      | 118/318 [00:34<00:50,  3.97it/s]\u001b[A\n",
            " 37%|███▋      | 119/318 [00:34<00:50,  3.94it/s]\u001b[A\n",
            " 38%|███▊      | 120/318 [00:35<00:49,  4.02it/s]\u001b[A\n",
            " 38%|███▊      | 121/318 [00:35<00:52,  3.75it/s]\u001b[A\n",
            " 38%|███▊      | 122/318 [00:35<00:50,  3.92it/s]\u001b[A\n",
            " 39%|███▊      | 123/318 [00:35<00:50,  3.84it/s]\u001b[A\n",
            " 39%|███▉      | 124/318 [00:36<00:50,  3.84it/s]\u001b[A\n",
            " 39%|███▉      | 125/318 [00:36<00:48,  4.01it/s]\u001b[A\n",
            " 40%|███▉      | 126/318 [00:36<00:47,  4.04it/s]\u001b[A\n",
            " 40%|███▉      | 127/318 [00:36<00:45,  4.20it/s]\u001b[A\n",
            " 40%|████      | 128/318 [00:37<00:49,  3.87it/s]\u001b[A\n",
            " 41%|████      | 129/318 [00:37<00:50,  3.76it/s]\u001b[A\n",
            " 41%|████      | 130/318 [00:37<00:50,  3.74it/s]\u001b[A\n",
            " 41%|████      | 131/318 [00:37<00:52,  3.58it/s]\u001b[A\n",
            " 42%|████▏     | 132/318 [00:38<00:50,  3.68it/s]\u001b[A\n",
            " 42%|████▏     | 133/318 [00:38<00:46,  3.98it/s]\u001b[A\n",
            " 42%|████▏     | 134/318 [00:38<00:46,  3.92it/s]\u001b[A\n",
            " 42%|████▏     | 135/318 [00:38<00:45,  4.03it/s]\u001b[A\n",
            " 43%|████▎     | 136/318 [00:39<00:46,  3.93it/s]\u001b[A\n",
            " 43%|████▎     | 137/318 [00:39<00:44,  4.08it/s]\u001b[A\n",
            " 43%|████▎     | 138/318 [00:39<00:45,  4.00it/s]\u001b[A\n",
            " 44%|████▎     | 139/318 [00:39<00:42,  4.23it/s]\u001b[A\n",
            " 44%|████▍     | 140/318 [00:40<00:41,  4.31it/s]\u001b[A\n",
            " 44%|████▍     | 141/318 [00:40<00:42,  4.15it/s]\u001b[A\n",
            " 45%|████▍     | 142/318 [00:40<00:41,  4.25it/s]\u001b[A\n",
            " 45%|████▍     | 143/318 [00:40<00:42,  4.07it/s]\u001b[A\n",
            " 45%|████▌     | 144/318 [00:41<00:44,  3.88it/s]\u001b[A\n",
            " 46%|████▌     | 145/318 [00:41<00:44,  3.89it/s]\u001b[A\n",
            " 46%|████▌     | 146/318 [00:41<00:41,  4.17it/s]\u001b[A\n",
            " 46%|████▌     | 147/318 [00:41<00:43,  3.93it/s]\u001b[A\n",
            " 47%|████▋     | 148/318 [00:42<00:44,  3.85it/s]\u001b[A\n",
            " 47%|████▋     | 149/318 [00:42<00:40,  4.12it/s]\u001b[A\n",
            " 47%|████▋     | 150/318 [00:42<00:39,  4.24it/s]\u001b[A\n",
            " 47%|████▋     | 151/318 [00:42<00:41,  4.00it/s]\u001b[A\n",
            " 48%|████▊     | 152/318 [00:43<00:39,  4.24it/s]\u001b[A\n",
            " 48%|████▊     | 153/318 [00:43<00:38,  4.24it/s]\u001b[A\n",
            " 48%|████▊     | 154/318 [00:43<00:38,  4.29it/s]\u001b[A\n",
            " 49%|████▊     | 155/318 [00:43<00:38,  4.19it/s]\u001b[A\n",
            " 49%|████▉     | 156/318 [00:44<00:40,  3.96it/s]\u001b[A\n",
            " 49%|████▉     | 157/318 [00:44<00:42,  3.83it/s]\u001b[A\n",
            " 50%|████▉     | 158/318 [00:44<00:42,  3.79it/s]\u001b[A\n",
            " 50%|█████     | 159/318 [00:44<00:39,  3.99it/s]\u001b[A\n",
            " 50%|█████     | 160/318 [00:45<00:39,  3.97it/s]\u001b[A\n",
            " 51%|█████     | 161/318 [00:45<00:38,  4.09it/s]\u001b[A\n",
            " 51%|█████     | 162/318 [00:45<00:37,  4.13it/s]\u001b[A\n",
            " 51%|█████▏    | 163/318 [00:45<00:37,  4.10it/s]\u001b[A\n",
            " 52%|█████▏    | 164/318 [00:45<00:36,  4.25it/s]\u001b[A\n",
            " 52%|█████▏    | 165/318 [00:46<00:36,  4.14it/s]\u001b[A\n",
            " 52%|█████▏    | 166/318 [00:46<00:39,  3.84it/s]\u001b[A\n",
            " 53%|█████▎    | 167/318 [00:46<00:39,  3.80it/s]\u001b[A\n",
            " 53%|█████▎    | 168/318 [00:47<00:40,  3.72it/s]\u001b[A\n",
            " 53%|█████▎    | 169/318 [00:47<00:36,  4.04it/s]\u001b[A\n",
            " 53%|█████▎    | 170/318 [00:47<00:35,  4.20it/s]\u001b[A\n",
            " 54%|█████▍    | 171/318 [00:47<00:34,  4.29it/s]\u001b[A\n",
            " 54%|█████▍    | 172/318 [00:47<00:33,  4.36it/s]\u001b[A\n",
            " 54%|█████▍    | 173/318 [00:48<00:32,  4.40it/s]\u001b[A\n",
            " 55%|█████▍    | 174/318 [00:48<00:32,  4.43it/s]\u001b[A\n",
            " 55%|█████▌    | 175/318 [00:48<00:32,  4.40it/s]\u001b[A\n",
            " 55%|█████▌    | 176/318 [00:48<00:32,  4.33it/s]\u001b[A\n",
            " 56%|█████▌    | 177/318 [00:49<00:34,  4.14it/s]\u001b[A\n",
            " 56%|█████▌    | 178/318 [00:49<00:33,  4.21it/s]\u001b[A\n",
            " 56%|█████▋    | 179/318 [00:49<00:34,  4.04it/s]\u001b[A\n",
            " 57%|█████▋    | 180/318 [00:49<00:33,  4.13it/s]\u001b[A\n",
            " 57%|█████▋    | 181/318 [00:50<00:32,  4.22it/s]\u001b[A\n",
            " 57%|█████▋    | 182/318 [00:50<00:32,  4.21it/s]\u001b[A\n",
            " 58%|█████▊    | 183/318 [00:50<00:32,  4.10it/s]\u001b[A\n",
            " 58%|█████▊    | 184/318 [00:50<00:35,  3.82it/s]\u001b[A\n",
            " 58%|█████▊    | 185/318 [00:51<00:34,  3.85it/s]\u001b[A\n",
            " 58%|█████▊    | 186/318 [00:51<00:33,  3.89it/s]\u001b[A\n",
            " 59%|█████▉    | 187/318 [00:51<00:34,  3.84it/s]\u001b[A\n",
            " 59%|█████▉    | 188/318 [00:51<00:32,  4.00it/s]\u001b[A\n",
            " 59%|█████▉    | 189/318 [00:52<00:31,  4.03it/s]\u001b[A\n",
            " 60%|█████▉    | 190/318 [00:52<00:29,  4.32it/s]\u001b[A\n",
            " 60%|██████    | 191/318 [00:52<00:32,  3.95it/s]\u001b[A\n",
            " 60%|██████    | 192/318 [00:52<00:31,  4.01it/s]\u001b[A\n",
            " 61%|██████    | 193/318 [00:53<00:31,  3.92it/s]\u001b[A\n",
            " 61%|██████    | 194/318 [00:53<00:30,  4.05it/s]\u001b[A\n",
            " 61%|██████▏   | 195/318 [00:53<00:28,  4.27it/s]\u001b[A\n",
            " 62%|██████▏   | 196/318 [00:53<00:27,  4.48it/s]\u001b[A\n",
            " 62%|██████▏   | 197/318 [00:53<00:26,  4.51it/s]\u001b[A\n",
            " 62%|██████▏   | 198/318 [00:54<00:27,  4.36it/s]\u001b[A\n",
            " 63%|██████▎   | 199/318 [00:54<00:26,  4.56it/s]\u001b[A\n",
            " 63%|██████▎   | 200/318 [00:54<00:27,  4.35it/s]\u001b[A\n",
            " 63%|██████▎   | 201/318 [00:54<00:27,  4.20it/s]\u001b[A\n",
            " 64%|██████▎   | 202/318 [00:55<00:27,  4.24it/s]\u001b[A\n",
            " 64%|██████▍   | 203/318 [00:55<00:29,  3.96it/s]\u001b[A\n",
            " 64%|██████▍   | 204/318 [00:55<00:28,  4.03it/s]\u001b[A\n",
            " 64%|██████▍   | 205/318 [00:55<00:27,  4.10it/s]\u001b[A\n",
            " 65%|██████▍   | 206/318 [00:56<00:27,  4.14it/s]\u001b[A\n",
            " 65%|██████▌   | 207/318 [00:56<00:27,  4.03it/s]\u001b[A\n",
            " 65%|██████▌   | 208/318 [00:56<00:28,  3.91it/s]\u001b[A\n",
            " 66%|██████▌   | 209/318 [00:56<00:28,  3.87it/s]\u001b[A\n",
            " 66%|██████▌   | 210/318 [00:57<00:33,  3.23it/s]\u001b[A\n",
            " 66%|██████▋   | 211/318 [00:57<00:30,  3.53it/s]\u001b[A\n",
            " 67%|██████▋   | 212/318 [00:57<00:31,  3.42it/s]\u001b[A\n",
            " 67%|██████▋   | 213/318 [00:58<00:33,  3.10it/s]\u001b[A\n",
            " 67%|██████▋   | 214/318 [00:58<00:31,  3.35it/s]\u001b[A\n",
            " 68%|██████▊   | 215/318 [00:58<00:29,  3.47it/s]\u001b[A\n",
            " 68%|██████▊   | 216/318 [00:59<00:30,  3.39it/s]\u001b[A\n",
            " 68%|██████▊   | 217/318 [00:59<00:31,  3.24it/s]\u001b[A\n",
            " 69%|██████▊   | 218/318 [00:59<00:28,  3.49it/s]\u001b[A\n",
            " 69%|██████▉   | 219/318 [00:59<00:27,  3.55it/s]\u001b[A\n",
            " 69%|██████▉   | 220/318 [01:00<00:26,  3.76it/s]\u001b[A\n",
            " 69%|██████▉   | 221/318 [01:00<00:25,  3.84it/s]\u001b[A\n",
            " 70%|██████▉   | 222/318 [01:00<00:24,  3.88it/s]\u001b[A\n",
            " 70%|███████   | 223/318 [01:00<00:24,  3.91it/s]\u001b[A\n",
            " 70%|███████   | 224/318 [01:01<00:22,  4.20it/s]\u001b[A\n",
            " 71%|███████   | 225/318 [01:01<00:21,  4.25it/s]\u001b[A\n",
            " 71%|███████   | 226/318 [01:01<00:21,  4.20it/s]\u001b[A\n",
            " 71%|███████▏  | 227/318 [01:01<00:20,  4.33it/s]\u001b[A\n",
            " 72%|███████▏  | 228/318 [01:02<00:21,  4.14it/s]\u001b[A\n",
            " 72%|███████▏  | 229/318 [01:02<00:21,  4.19it/s]\u001b[A\n",
            " 72%|███████▏  | 230/318 [01:02<00:20,  4.33it/s]\u001b[A\n",
            " 73%|███████▎  | 231/318 [01:02<00:20,  4.25it/s]\u001b[A\n",
            " 73%|███████▎  | 232/318 [01:03<00:19,  4.35it/s]\u001b[A\n",
            " 73%|███████▎  | 233/318 [01:03<00:19,  4.43it/s]\u001b[A\n",
            " 74%|███████▎  | 234/318 [01:03<00:19,  4.34it/s]\u001b[A\n",
            " 74%|███████▍  | 235/318 [01:03<00:18,  4.49it/s]\u001b[A\n",
            " 74%|███████▍  | 236/318 [01:03<00:19,  4.19it/s]\u001b[A\n",
            " 75%|███████▍  | 237/318 [01:04<00:18,  4.32it/s]\u001b[A\n",
            " 75%|███████▍  | 238/318 [01:04<00:18,  4.34it/s]\u001b[A\n",
            " 75%|███████▌  | 239/318 [01:04<00:17,  4.42it/s]\u001b[A\n",
            " 75%|███████▌  | 240/318 [01:04<00:17,  4.37it/s]\u001b[A\n",
            " 76%|███████▌  | 241/318 [01:05<00:17,  4.37it/s]\u001b[A\n",
            " 76%|███████▌  | 242/318 [01:05<00:17,  4.39it/s]\u001b[A\n",
            " 76%|███████▋  | 243/318 [01:05<00:16,  4.56it/s]\u001b[A\n",
            " 77%|███████▋  | 244/318 [01:05<00:16,  4.43it/s]\u001b[A\n",
            " 77%|███████▋  | 245/318 [01:05<00:16,  4.43it/s]\u001b[A\n",
            " 77%|███████▋  | 246/318 [01:06<00:16,  4.49it/s]\u001b[A\n",
            " 78%|███████▊  | 247/318 [01:06<00:15,  4.49it/s]\u001b[A\n",
            " 78%|███████▊  | 248/318 [01:06<00:15,  4.56it/s]\u001b[A\n",
            " 78%|███████▊  | 249/318 [01:06<00:15,  4.33it/s]\u001b[A\n",
            " 79%|███████▊  | 250/318 [01:07<00:15,  4.31it/s]\u001b[A\n",
            " 79%|███████▉  | 251/318 [01:07<00:17,  3.87it/s]\u001b[A\n",
            " 79%|███████▉  | 252/318 [01:07<00:16,  3.90it/s]\u001b[A\n",
            " 80%|███████▉  | 253/318 [01:07<00:16,  3.98it/s]\u001b[A\n",
            " 80%|███████▉  | 254/318 [01:08<00:15,  4.14it/s]\u001b[A\n",
            " 80%|████████  | 255/318 [01:08<00:14,  4.43it/s]\u001b[A\n",
            " 81%|████████  | 256/318 [01:08<00:14,  4.18it/s]\u001b[A\n",
            " 81%|████████  | 257/318 [01:08<00:14,  4.10it/s]\u001b[A\n",
            " 81%|████████  | 258/318 [01:09<00:14,  4.11it/s]\u001b[A\n",
            " 81%|████████▏ | 259/318 [01:09<00:15,  3.91it/s]\u001b[A\n",
            " 82%|████████▏ | 260/318 [01:09<00:14,  3.99it/s]\u001b[A\n",
            " 82%|████████▏ | 261/318 [01:09<00:13,  4.08it/s]\u001b[A\n",
            " 82%|████████▏ | 262/318 [01:10<00:13,  4.10it/s]\u001b[A\n",
            " 83%|████████▎ | 263/318 [01:10<00:13,  4.16it/s]\u001b[A\n",
            " 83%|████████▎ | 264/318 [01:10<00:13,  4.12it/s]\u001b[A\n",
            " 83%|████████▎ | 265/318 [01:10<00:13,  3.97it/s]\u001b[A\n",
            " 84%|████████▎ | 266/318 [01:11<00:12,  4.04it/s]\u001b[A\n",
            " 84%|████████▍ | 267/318 [01:11<00:12,  4.09it/s]\u001b[A\n",
            " 84%|████████▍ | 268/318 [01:11<00:12,  3.97it/s]\u001b[A\n",
            " 85%|████████▍ | 269/318 [01:11<00:12,  4.04it/s]\u001b[A\n",
            " 85%|████████▍ | 270/318 [01:12<00:12,  3.96it/s]\u001b[A\n",
            " 85%|████████▌ | 271/318 [01:12<00:11,  4.02it/s]\u001b[A\n",
            " 86%|████████▌ | 272/318 [01:12<00:11,  4.12it/s]\u001b[A\n",
            " 86%|████████▌ | 273/318 [01:12<00:11,  3.89it/s]\u001b[A\n",
            " 86%|████████▌ | 274/318 [01:13<00:10,  4.16it/s]\u001b[A\n",
            " 86%|████████▋ | 275/318 [01:13<00:10,  4.12it/s]\u001b[A\n",
            " 87%|████████▋ | 276/318 [01:13<00:10,  4.11it/s]\u001b[A\n",
            " 87%|████████▋ | 277/318 [01:13<00:09,  4.10it/s]\u001b[A\n",
            " 87%|████████▋ | 278/318 [01:14<00:09,  4.16it/s]\u001b[A\n",
            " 88%|████████▊ | 279/318 [01:14<00:08,  4.34it/s]\u001b[A\n",
            " 88%|████████▊ | 280/318 [01:14<00:08,  4.53it/s]\u001b[A\n",
            " 88%|████████▊ | 281/318 [01:14<00:07,  4.64it/s]\u001b[A\n",
            " 89%|████████▊ | 282/318 [01:14<00:08,  4.41it/s]\u001b[A\n",
            " 89%|████████▉ | 283/318 [01:15<00:07,  4.41it/s]\u001b[A\n",
            " 89%|████████▉ | 284/318 [01:15<00:08,  4.25it/s]\u001b[A\n",
            " 90%|████████▉ | 285/318 [01:15<00:07,  4.24it/s]\u001b[A\n",
            " 90%|████████▉ | 286/318 [01:15<00:07,  4.41it/s]\u001b[A\n",
            " 90%|█████████ | 287/318 [01:16<00:07,  4.37it/s]\u001b[A\n",
            " 91%|█████████ | 288/318 [01:16<00:06,  4.36it/s]\u001b[A\n",
            " 91%|█████████ | 289/318 [01:16<00:06,  4.43it/s]\u001b[A\n",
            " 91%|█████████ | 290/318 [01:16<00:06,  4.61it/s]\u001b[A\n",
            " 92%|█████████▏| 291/318 [01:16<00:06,  4.33it/s]\u001b[A\n",
            " 92%|█████████▏| 292/318 [01:17<00:06,  4.10it/s]\u001b[A\n",
            " 92%|█████████▏| 293/318 [01:17<00:06,  4.01it/s]\u001b[A\n",
            " 92%|█████████▏| 294/318 [01:17<00:05,  4.16it/s]\u001b[A\n",
            " 93%|█████████▎| 295/318 [01:18<00:05,  3.89it/s]\u001b[A\n",
            " 93%|█████████▎| 296/318 [01:18<00:05,  3.89it/s]\u001b[A\n",
            " 93%|█████████▎| 297/318 [01:18<00:05,  3.94it/s]\u001b[A\n",
            " 94%|█████████▎| 298/318 [01:18<00:04,  4.23it/s]\u001b[A\n",
            " 94%|█████████▍| 299/318 [01:19<00:04,  3.93it/s]\u001b[A\n",
            " 94%|█████████▍| 300/318 [01:19<00:04,  4.01it/s]\u001b[A\n",
            " 95%|█████████▍| 301/318 [01:19<00:04,  3.78it/s]\u001b[A\n",
            " 95%|█████████▍| 302/318 [01:19<00:04,  3.94it/s]\u001b[A\n",
            " 95%|█████████▌| 303/318 [01:20<00:03,  3.83it/s]\u001b[A\n",
            " 96%|█████████▌| 304/318 [01:20<00:03,  4.01it/s]\u001b[A\n",
            " 96%|█████████▌| 305/318 [01:20<00:03,  4.08it/s]\u001b[A\n",
            " 96%|█████████▌| 306/318 [01:20<00:02,  4.17it/s]\u001b[A\n",
            " 97%|█████████▋| 307/318 [01:20<00:02,  4.12it/s]\u001b[A\n",
            " 97%|█████████▋| 308/318 [01:21<00:02,  4.15it/s]\u001b[A\n",
            " 97%|█████████▋| 309/318 [01:21<00:02,  4.17it/s]\u001b[A\n",
            " 97%|█████████▋| 310/318 [01:21<00:01,  4.23it/s]\u001b[A\n",
            " 98%|█████████▊| 311/318 [01:21<00:01,  4.00it/s]\u001b[A\n",
            " 98%|█████████▊| 312/318 [01:22<00:01,  3.91it/s]\u001b[A\n",
            " 98%|█████████▊| 313/318 [01:22<00:01,  4.05it/s]\u001b[A\n",
            " 99%|█████████▊| 314/318 [01:22<00:01,  3.93it/s]\u001b[A\n",
            " 99%|█████████▉| 315/318 [01:23<00:00,  3.89it/s]\u001b[A\n",
            " 99%|█████████▉| 316/318 [01:23<00:00,  3.80it/s]\u001b[A\n",
            "100%|█████████▉| 317/318 [01:23<00:00,  3.79it/s]\u001b[A\n",
            "100%|██████████| 318/318 [01:23<00:00,  3.79it/s]\n",
            " 10%|█         | 1/10 [01:23<12:34, 83.83s/it]\n",
            "  0%|          | 0/318 [00:00<?, ?it/s]\u001b[A\n",
            "  0%|          | 1/318 [00:00<02:12,  2.38it/s]\u001b[A\n",
            "  1%|          | 2/318 [00:00<01:51,  2.83it/s]\u001b[A\n",
            "  1%|          | 3/318 [00:01<01:41,  3.11it/s]\u001b[A\n",
            "  1%|▏         | 4/318 [00:01<01:42,  3.05it/s]\u001b[A\n",
            "  2%|▏         | 5/318 [00:01<01:28,  3.53it/s]\u001b[A\n",
            "  2%|▏         | 6/318 [00:01<01:31,  3.40it/s]\u001b[A\n",
            "  2%|▏         | 7/318 [00:02<01:29,  3.49it/s]\u001b[A\n",
            "  3%|▎         | 8/318 [00:02<01:33,  3.30it/s]\u001b[A\n",
            "  3%|▎         | 9/318 [00:02<01:27,  3.54it/s]\u001b[A\n",
            "  3%|▎         | 10/318 [00:02<01:24,  3.65it/s]\u001b[A\n",
            "  3%|▎         | 11/318 [00:03<01:23,  3.70it/s]\u001b[A\n",
            "  4%|▍         | 12/318 [00:03<01:13,  4.15it/s]\u001b[A\n",
            "  4%|▍         | 13/318 [00:03<01:12,  4.21it/s]\u001b[A\n",
            "  4%|▍         | 14/318 [00:03<01:15,  4.04it/s]\u001b[A\n",
            "  5%|▍         | 15/318 [00:04<01:11,  4.25it/s]\u001b[A\n",
            "  5%|▌         | 16/318 [00:04<01:09,  4.32it/s]\u001b[A\n",
            "  5%|▌         | 17/318 [00:04<01:06,  4.50it/s]\u001b[A\n",
            "  6%|▌         | 18/318 [00:04<01:12,  4.15it/s]\u001b[A\n",
            "  6%|▌         | 19/318 [00:05<01:11,  4.21it/s]\u001b[A\n",
            "  6%|▋         | 20/318 [00:05<01:11,  4.19it/s]\u001b[A\n",
            "  7%|▋         | 21/318 [00:05<01:10,  4.23it/s]\u001b[A\n",
            "  7%|▋         | 22/318 [00:05<01:10,  4.22it/s]\u001b[A\n",
            "  7%|▋         | 23/318 [00:06<01:13,  4.01it/s]\u001b[A\n",
            "  8%|▊         | 24/318 [00:06<01:11,  4.09it/s]\u001b[A\n",
            "  8%|▊         | 25/318 [00:06<01:12,  4.06it/s]\u001b[A\n",
            "  8%|▊         | 26/318 [00:06<01:11,  4.10it/s]\u001b[A\n",
            "  8%|▊         | 27/318 [00:06<01:09,  4.20it/s]\u001b[A\n",
            "  9%|▉         | 28/318 [00:07<01:07,  4.30it/s]\u001b[A\n",
            "  9%|▉         | 29/318 [00:07<01:07,  4.29it/s]\u001b[A\n",
            "  9%|▉         | 30/318 [00:07<01:04,  4.45it/s]\u001b[A\n",
            " 10%|▉         | 31/318 [00:07<01:04,  4.46it/s]\u001b[A\n",
            " 10%|█         | 32/318 [00:08<01:04,  4.42it/s]\u001b[A\n",
            " 10%|█         | 33/318 [00:08<01:06,  4.31it/s]\u001b[A\n",
            " 11%|█         | 34/318 [00:08<01:04,  4.42it/s]\u001b[A\n",
            " 11%|█         | 35/318 [00:08<01:06,  4.23it/s]\u001b[A\n",
            " 11%|█▏        | 36/318 [00:09<01:16,  3.71it/s]\u001b[A\n",
            " 12%|█▏        | 37/318 [00:09<01:12,  3.89it/s]\u001b[A\n",
            " 12%|█▏        | 38/318 [00:09<01:06,  4.24it/s]\u001b[A\n",
            " 12%|█▏        | 39/318 [00:09<01:07,  4.16it/s]\u001b[A\n",
            " 13%|█▎        | 40/318 [00:10<01:05,  4.26it/s]\u001b[A\n",
            " 13%|█▎        | 41/318 [00:10<01:04,  4.31it/s]\u001b[A\n",
            " 13%|█▎        | 42/318 [00:10<01:04,  4.28it/s]\u001b[A\n",
            " 14%|█▎        | 43/318 [00:10<01:03,  4.36it/s]\u001b[A\n",
            " 14%|█▍        | 44/318 [00:10<01:04,  4.26it/s]\u001b[A\n",
            " 14%|█▍        | 45/318 [00:11<01:07,  4.02it/s]\u001b[A\n",
            " 14%|█▍        | 46/318 [00:11<01:06,  4.10it/s]\u001b[A\n",
            " 15%|█▍        | 47/318 [00:11<01:08,  3.97it/s]\u001b[A\n",
            " 15%|█▌        | 48/318 [00:11<01:04,  4.19it/s]\u001b[A\n",
            " 15%|█▌        | 49/318 [00:12<01:01,  4.35it/s]\u001b[A\n",
            " 16%|█▌        | 50/318 [00:12<01:08,  3.92it/s]\u001b[A\n",
            " 16%|█▌        | 51/318 [00:12<01:05,  4.08it/s]\u001b[A\n",
            " 16%|█▋        | 52/318 [00:13<01:11,  3.74it/s]\u001b[A\n",
            " 17%|█▋        | 53/318 [00:13<01:13,  3.63it/s]\u001b[A\n",
            " 17%|█▋        | 54/318 [00:13<01:14,  3.54it/s]\u001b[A\n",
            " 17%|█▋        | 55/318 [00:13<01:08,  3.87it/s]\u001b[A\n",
            " 18%|█▊        | 56/318 [00:14<01:08,  3.84it/s]\u001b[A\n",
            " 18%|█▊        | 57/318 [00:14<01:03,  4.08it/s]\u001b[A\n",
            " 18%|█▊        | 58/318 [00:14<01:04,  4.02it/s]\u001b[A\n",
            " 19%|█▊        | 59/318 [00:14<01:01,  4.22it/s]\u001b[A\n",
            " 19%|█▉        | 60/318 [00:15<01:03,  4.07it/s]\u001b[A\n",
            " 19%|█▉        | 61/318 [00:15<01:02,  4.14it/s]\u001b[A\n",
            " 19%|█▉        | 62/318 [00:15<01:01,  4.17it/s]\u001b[A\n",
            " 20%|█▉        | 63/318 [00:15<01:02,  4.07it/s]\u001b[A\n",
            " 20%|██        | 64/318 [00:15<00:59,  4.25it/s]\u001b[A\n",
            " 20%|██        | 65/318 [00:16<01:01,  4.11it/s]\u001b[A\n",
            " 21%|██        | 66/318 [00:16<01:04,  3.92it/s]\u001b[A\n",
            " 21%|██        | 67/318 [00:16<01:03,  3.94it/s]\u001b[A\n",
            " 21%|██▏       | 68/318 [00:17<01:04,  3.85it/s]\u001b[A\n",
            " 22%|██▏       | 69/318 [00:17<01:03,  3.94it/s]\u001b[A\n",
            " 22%|██▏       | 70/318 [00:17<01:06,  3.74it/s]\u001b[A\n",
            " 22%|██▏       | 71/318 [00:17<01:13,  3.38it/s]\u001b[A\n",
            " 23%|██▎       | 72/318 [00:18<01:08,  3.61it/s]\u001b[A\n",
            " 23%|██▎       | 73/318 [00:18<01:08,  3.60it/s]\u001b[A\n",
            " 23%|██▎       | 74/318 [00:18<01:02,  3.88it/s]\u001b[A\n",
            " 24%|██▎       | 75/318 [00:18<01:04,  3.76it/s]\u001b[A\n",
            " 24%|██▍       | 76/318 [00:19<01:01,  3.93it/s]\u001b[A\n",
            " 24%|██▍       | 77/318 [00:19<00:59,  4.05it/s]\u001b[A\n",
            " 25%|██▍       | 78/318 [00:19<00:58,  4.12it/s]\u001b[A\n",
            " 25%|██▍       | 79/318 [00:19<00:56,  4.27it/s]\u001b[A\n",
            " 25%|██▌       | 80/318 [00:20<00:59,  4.01it/s]\u001b[A\n",
            " 25%|██▌       | 81/318 [00:20<00:58,  4.02it/s]\u001b[A\n",
            " 26%|██▌       | 82/318 [00:20<00:55,  4.27it/s]\u001b[A\n",
            " 26%|██▌       | 83/318 [00:20<00:52,  4.51it/s]\u001b[A\n",
            " 26%|██▋       | 84/318 [00:21<00:51,  4.54it/s]\u001b[A\n",
            " 27%|██▋       | 85/318 [00:21<00:51,  4.50it/s]\u001b[A\n",
            " 27%|██▋       | 86/318 [00:21<00:50,  4.58it/s]\u001b[A\n",
            " 27%|██▋       | 87/318 [00:21<00:50,  4.54it/s]\u001b[A\n",
            " 28%|██▊       | 88/318 [00:21<00:49,  4.61it/s]\u001b[A\n",
            " 28%|██▊       | 89/318 [00:22<00:49,  4.63it/s]\u001b[A\n",
            " 28%|██▊       | 90/318 [00:22<00:48,  4.66it/s]\u001b[A\n",
            " 29%|██▊       | 91/318 [00:22<00:48,  4.65it/s]\u001b[A\n",
            " 29%|██▉       | 92/318 [00:22<00:47,  4.75it/s]\u001b[A\n",
            " 29%|██▉       | 93/318 [00:22<00:45,  4.97it/s]\u001b[A\n",
            " 30%|██▉       | 94/318 [00:23<00:46,  4.83it/s]\u001b[A\n",
            " 30%|██▉       | 95/318 [00:23<00:46,  4.79it/s]\u001b[A\n",
            " 30%|███       | 96/318 [00:23<00:48,  4.59it/s]\u001b[A\n",
            " 31%|███       | 97/318 [00:23<00:50,  4.37it/s]\u001b[A\n",
            " 31%|███       | 98/318 [00:24<00:58,  3.79it/s]\u001b[A\n",
            " 31%|███       | 99/318 [00:24<00:58,  3.72it/s]\u001b[A\n",
            " 31%|███▏      | 100/318 [00:24<00:58,  3.70it/s]\u001b[A\n",
            " 32%|███▏      | 101/318 [00:25<01:02,  3.46it/s]\u001b[A\n",
            " 32%|███▏      | 102/318 [00:25<01:02,  3.44it/s]\u001b[A\n",
            " 32%|███▏      | 103/318 [00:25<01:01,  3.49it/s]\u001b[A\n",
            " 33%|███▎      | 104/318 [00:25<00:59,  3.59it/s]\u001b[A\n",
            " 33%|███▎      | 105/318 [00:26<00:58,  3.63it/s]\u001b[A\n",
            " 33%|███▎      | 106/318 [00:26<00:56,  3.76it/s]\u001b[A\n",
            " 34%|███▎      | 107/318 [00:26<00:53,  3.98it/s]\u001b[A\n",
            " 34%|███▍      | 108/318 [00:26<00:50,  4.13it/s]\u001b[A\n",
            " 34%|███▍      | 109/318 [00:27<00:55,  3.75it/s]\u001b[A\n",
            " 35%|███▍      | 110/318 [00:27<00:55,  3.72it/s]\u001b[A\n",
            " 35%|███▍      | 111/318 [00:27<00:57,  3.62it/s]\u001b[A\n",
            " 35%|███▌      | 112/318 [00:27<00:54,  3.77it/s]\u001b[A\n",
            " 36%|███▌      | 113/318 [00:28<00:53,  3.83it/s]\u001b[A\n",
            " 36%|███▌      | 114/318 [00:28<00:55,  3.70it/s]\u001b[A\n",
            " 36%|███▌      | 115/318 [00:28<00:48,  4.19it/s]\u001b[A\n",
            " 36%|███▋      | 116/318 [00:28<00:50,  4.00it/s]\u001b[A\n",
            " 37%|███▋      | 117/318 [00:29<00:46,  4.29it/s]\u001b[A\n",
            " 37%|███▋      | 118/318 [00:29<00:44,  4.52it/s]\u001b[A\n",
            " 37%|███▋      | 119/318 [00:29<00:45,  4.35it/s]\u001b[A\n",
            " 38%|███▊      | 120/318 [00:29<00:44,  4.46it/s]\u001b[A\n",
            " 38%|███▊      | 121/318 [00:30<00:44,  4.40it/s]\u001b[A\n",
            " 38%|███▊      | 122/318 [00:30<00:44,  4.37it/s]\u001b[A\n",
            " 39%|███▊      | 123/318 [00:30<00:48,  4.03it/s]\u001b[A\n",
            " 39%|███▉      | 124/318 [00:30<00:48,  3.99it/s]\u001b[A\n",
            " 39%|███▉      | 125/318 [00:31<00:46,  4.12it/s]\u001b[A\n",
            " 40%|███▉      | 126/318 [00:31<00:48,  3.96it/s]\u001b[A\n",
            " 40%|███▉      | 127/318 [00:31<00:46,  4.08it/s]\u001b[A\n",
            " 40%|████      | 128/318 [00:31<00:43,  4.33it/s]\u001b[A\n",
            " 41%|████      | 129/318 [00:31<00:44,  4.28it/s]\u001b[A\n",
            " 41%|████      | 130/318 [00:32<00:43,  4.27it/s]\u001b[A\n",
            " 41%|████      | 131/318 [00:32<00:43,  4.26it/s]\u001b[A\n",
            " 42%|████▏     | 132/318 [00:32<00:42,  4.37it/s]\u001b[A\n",
            " 42%|████▏     | 133/318 [00:32<00:44,  4.19it/s]\u001b[A\n",
            " 42%|████▏     | 134/318 [00:33<00:47,  3.89it/s]\u001b[A\n",
            " 42%|████▏     | 135/318 [00:33<00:46,  3.94it/s]\u001b[A\n",
            " 43%|████▎     | 136/318 [00:33<00:43,  4.14it/s]\u001b[A\n",
            " 43%|████▎     | 137/318 [00:33<00:43,  4.14it/s]\u001b[A\n",
            " 43%|████▎     | 138/318 [00:34<00:44,  4.02it/s]\u001b[A\n",
            " 44%|████▎     | 139/318 [00:34<00:40,  4.39it/s]\u001b[A\n",
            " 44%|████▍     | 140/318 [00:34<00:42,  4.22it/s]\u001b[A\n",
            " 44%|████▍     | 141/318 [00:34<00:47,  3.77it/s]\u001b[A\n",
            " 45%|████▍     | 142/318 [00:35<00:46,  3.81it/s]\u001b[A\n",
            " 45%|████▍     | 143/318 [00:35<00:45,  3.86it/s]\u001b[A\n",
            " 45%|████▌     | 144/318 [00:35<00:44,  3.90it/s]\u001b[A\n",
            " 46%|████▌     | 145/318 [00:35<00:43,  3.96it/s]\u001b[A\n",
            " 46%|████▌     | 146/318 [00:36<00:46,  3.71it/s]\u001b[A\n",
            " 46%|████▌     | 147/318 [00:36<00:45,  3.77it/s]\u001b[A\n",
            " 47%|████▋     | 148/318 [00:36<00:44,  3.80it/s]\u001b[A\n",
            " 47%|████▋     | 149/318 [00:37<00:43,  3.85it/s]\u001b[A\n",
            " 47%|████▋     | 150/318 [00:37<00:42,  3.98it/s]\u001b[A\n",
            " 47%|████▋     | 151/318 [00:37<00:42,  3.94it/s]\u001b[A\n",
            " 48%|████▊     | 152/318 [00:37<00:39,  4.17it/s]\u001b[A\n",
            " 48%|████▊     | 153/318 [00:37<00:36,  4.49it/s]\u001b[A\n",
            " 48%|████▊     | 154/318 [00:38<00:37,  4.39it/s]\u001b[A\n",
            " 49%|████▊     | 155/318 [00:38<00:38,  4.22it/s]\u001b[A\n",
            " 49%|████▉     | 156/318 [00:38<00:39,  4.13it/s]\u001b[A\n",
            " 49%|████▉     | 157/318 [00:38<00:39,  4.09it/s]\u001b[A\n",
            " 50%|████▉     | 158/318 [00:39<00:38,  4.11it/s]\u001b[A\n",
            " 50%|█████     | 159/318 [00:39<00:37,  4.19it/s]\u001b[A\n",
            " 50%|█████     | 160/318 [00:39<00:37,  4.17it/s]\u001b[A\n",
            " 51%|█████     | 161/318 [00:39<00:38,  4.11it/s]\u001b[A\n",
            " 51%|█████     | 162/318 [00:40<00:39,  3.99it/s]\u001b[A\n",
            " 51%|█████▏    | 163/318 [00:40<00:39,  3.95it/s]\u001b[A\n",
            " 52%|█████▏    | 164/318 [00:40<00:38,  4.03it/s]\u001b[A\n",
            " 52%|█████▏    | 165/318 [00:40<00:37,  4.11it/s]\u001b[A\n",
            " 52%|█████▏    | 166/318 [00:41<00:34,  4.37it/s]\u001b[A\n",
            " 53%|█████▎    | 167/318 [00:41<00:36,  4.14it/s]\u001b[A\n",
            " 53%|█████▎    | 168/318 [00:41<00:35,  4.21it/s]\u001b[A\n",
            " 53%|█████▎    | 169/318 [00:41<00:34,  4.31it/s]\u001b[A\n",
            " 53%|█████▎    | 170/318 [00:42<00:36,  4.04it/s]\u001b[A\n",
            " 54%|█████▍    | 171/318 [00:42<00:35,  4.15it/s]\u001b[A\n",
            " 54%|█████▍    | 172/318 [00:42<00:37,  3.94it/s]\u001b[A\n",
            " 54%|█████▍    | 173/318 [00:42<00:34,  4.18it/s]\u001b[A\n",
            " 55%|█████▍    | 174/318 [00:43<00:34,  4.22it/s]\u001b[A\n",
            " 55%|█████▌    | 175/318 [00:43<00:35,  4.02it/s]\u001b[A\n",
            " 55%|█████▌    | 176/318 [00:43<00:34,  4.11it/s]\u001b[A\n",
            " 56%|█████▌    | 177/318 [00:43<00:35,  3.98it/s]\u001b[A\n",
            " 56%|█████▌    | 178/318 [00:44<00:35,  3.92it/s]\u001b[A\n",
            " 56%|█████▋    | 179/318 [00:44<00:36,  3.79it/s]\u001b[A\n",
            " 57%|█████▋    | 180/318 [00:44<00:33,  4.11it/s]\u001b[A\n",
            " 57%|█████▋    | 181/318 [00:44<00:33,  4.15it/s]\u001b[A\n",
            " 57%|█████▋    | 182/318 [00:45<00:35,  3.86it/s]\u001b[A\n",
            " 58%|█████▊    | 183/318 [00:45<00:35,  3.82it/s]\u001b[A\n",
            " 58%|█████▊    | 184/318 [00:45<00:36,  3.67it/s]\u001b[A\n",
            " 58%|█████▊    | 185/318 [00:45<00:34,  3.83it/s]\u001b[A\n",
            " 58%|█████▊    | 186/318 [00:46<00:34,  3.80it/s]\u001b[A\n",
            " 59%|█████▉    | 187/318 [00:46<00:32,  4.08it/s]\u001b[A\n",
            " 59%|█████▉    | 188/318 [00:46<00:31,  4.14it/s]\u001b[A\n",
            " 59%|█████▉    | 189/318 [00:46<00:32,  4.00it/s]\u001b[A\n",
            " 60%|█████▉    | 190/318 [00:47<00:32,  3.97it/s]\u001b[A\n",
            " 60%|██████    | 191/318 [00:47<00:30,  4.19it/s]\u001b[A\n",
            " 60%|██████    | 192/318 [00:47<00:31,  3.97it/s]\u001b[A\n",
            " 61%|██████    | 193/318 [00:47<00:31,  3.96it/s]\u001b[A\n",
            " 61%|██████    | 194/318 [00:48<00:30,  4.01it/s]\u001b[A\n",
            " 61%|██████▏   | 195/318 [00:48<00:29,  4.12it/s]\u001b[A\n",
            " 62%|██████▏   | 196/318 [00:48<00:29,  4.09it/s]\u001b[A\n",
            " 62%|██████▏   | 197/318 [00:48<00:31,  3.85it/s]\u001b[A\n",
            " 62%|██████▏   | 198/318 [00:49<00:28,  4.25it/s]\u001b[A\n",
            " 63%|██████▎   | 199/318 [00:49<00:28,  4.14it/s]\u001b[A\n",
            " 63%|██████▎   | 200/318 [00:49<00:28,  4.15it/s]\u001b[A\n",
            " 63%|██████▎   | 201/318 [00:49<00:28,  4.16it/s]\u001b[A\n",
            " 64%|██████▎   | 202/318 [00:50<00:27,  4.18it/s]\u001b[A\n",
            " 64%|██████▍   | 203/318 [00:50<00:27,  4.24it/s]\u001b[A\n",
            " 64%|██████▍   | 204/318 [00:50<00:25,  4.39it/s]\u001b[A\n",
            " 64%|██████▍   | 205/318 [00:50<00:28,  4.03it/s]\u001b[A\n",
            " 65%|██████▍   | 206/318 [00:50<00:26,  4.22it/s]\u001b[A\n",
            " 65%|██████▌   | 207/318 [00:51<00:28,  3.96it/s]\u001b[A\n",
            " 65%|██████▌   | 208/318 [00:51<00:28,  3.84it/s]\u001b[A\n",
            " 66%|██████▌   | 209/318 [00:51<00:27,  3.97it/s]\u001b[A\n",
            " 66%|██████▌   | 210/318 [00:52<00:27,  3.93it/s]\u001b[A\n",
            " 66%|██████▋   | 211/318 [00:52<00:26,  4.01it/s]\u001b[A\n",
            " 67%|██████▋   | 212/318 [00:52<00:25,  4.16it/s]\u001b[A\n",
            " 67%|██████▋   | 213/318 [00:52<00:24,  4.31it/s]\u001b[A\n",
            " 67%|██████▋   | 214/318 [00:52<00:24,  4.24it/s]\u001b[A\n",
            " 68%|██████▊   | 215/318 [00:53<00:25,  4.02it/s]\u001b[A\n",
            " 68%|██████▊   | 216/318 [00:53<00:23,  4.30it/s]\u001b[A\n",
            " 68%|██████▊   | 217/318 [00:53<00:22,  4.46it/s]\u001b[A\n",
            " 69%|██████▊   | 218/318 [00:53<00:24,  4.12it/s]\u001b[A\n",
            " 69%|██████▉   | 219/318 [00:54<00:23,  4.20it/s]\u001b[A\n",
            " 69%|██████▉   | 220/318 [00:54<00:22,  4.37it/s]\u001b[A\n",
            " 69%|██████▉   | 221/318 [00:54<00:21,  4.46it/s]\u001b[A\n",
            " 70%|██████▉   | 222/318 [00:54<00:21,  4.41it/s]\u001b[A\n",
            " 70%|███████   | 223/318 [00:55<00:21,  4.44it/s]\u001b[A\n",
            " 70%|███████   | 224/318 [00:55<00:22,  4.24it/s]\u001b[A\n",
            " 71%|███████   | 225/318 [00:55<00:21,  4.41it/s]\u001b[A\n",
            " 71%|███████   | 226/318 [00:55<00:21,  4.22it/s]\u001b[A\n",
            " 71%|███████▏  | 227/318 [00:55<00:21,  4.18it/s]\u001b[A\n",
            " 72%|███████▏  | 228/318 [00:56<00:22,  4.03it/s]\u001b[A\n",
            " 72%|███████▏  | 229/318 [00:56<00:21,  4.13it/s]\u001b[A\n",
            " 72%|███████▏  | 230/318 [00:56<00:20,  4.24it/s]\u001b[A\n",
            " 73%|███████▎  | 231/318 [00:56<00:21,  3.97it/s]\u001b[A\n",
            " 73%|███████▎  | 232/318 [00:57<00:21,  4.02it/s]\u001b[A\n",
            " 73%|███████▎  | 233/318 [00:57<00:23,  3.67it/s]\u001b[A\n",
            " 74%|███████▎  | 234/318 [00:57<00:22,  3.72it/s]\u001b[A\n",
            " 74%|███████▍  | 235/318 [00:58<00:21,  3.80it/s]\u001b[A\n",
            " 74%|███████▍  | 236/318 [00:58<00:20,  4.00it/s]\u001b[A\n",
            " 75%|███████▍  | 237/318 [00:58<00:19,  4.13it/s]\u001b[A\n",
            " 75%|███████▍  | 238/318 [00:58<00:20,  3.94it/s]\u001b[A\n",
            " 75%|███████▌  | 239/318 [00:59<00:19,  4.03it/s]\u001b[A\n",
            " 75%|███████▌  | 240/318 [00:59<00:18,  4.13it/s]\u001b[A\n",
            " 76%|███████▌  | 241/318 [00:59<00:18,  4.11it/s]\u001b[A\n",
            " 76%|███████▌  | 242/318 [00:59<00:18,  4.04it/s]\u001b[A\n",
            " 76%|███████▋  | 243/318 [01:00<00:19,  3.94it/s]\u001b[A\n",
            " 77%|███████▋  | 244/318 [01:00<00:18,  4.08it/s]\u001b[A\n",
            " 77%|███████▋  | 245/318 [01:00<00:18,  4.04it/s]\u001b[A\n",
            " 77%|███████▋  | 246/318 [01:00<00:18,  3.91it/s]\u001b[A\n",
            " 78%|███████▊  | 247/318 [01:01<00:18,  3.79it/s]\u001b[A\n",
            " 78%|███████▊  | 248/318 [01:01<00:18,  3.74it/s]\u001b[A\n",
            " 78%|███████▊  | 249/318 [01:01<00:18,  3.80it/s]\u001b[A\n",
            " 79%|███████▊  | 250/318 [01:01<00:18,  3.62it/s]\u001b[A\n",
            " 79%|███████▉  | 251/318 [01:02<00:17,  3.84it/s]\u001b[A\n",
            " 79%|███████▉  | 252/318 [01:02<00:18,  3.66it/s]\u001b[A\n",
            " 80%|███████▉  | 253/318 [01:02<00:16,  3.84it/s]\u001b[A\n",
            " 80%|███████▉  | 254/318 [01:02<00:16,  3.82it/s]\u001b[A\n",
            " 80%|████████  | 255/318 [01:03<00:15,  4.02it/s]\u001b[A\n",
            " 81%|████████  | 256/318 [01:03<00:14,  4.33it/s]\u001b[A\n",
            " 81%|████████  | 257/318 [01:03<00:14,  4.31it/s]\u001b[A\n",
            " 81%|████████  | 258/318 [01:03<00:14,  4.19it/s]\u001b[A\n",
            " 81%|████████▏ | 259/318 [01:04<00:14,  4.05it/s]\u001b[A\n",
            " 82%|████████▏ | 260/318 [01:04<00:13,  4.27it/s]\u001b[A\n",
            " 82%|████████▏ | 261/318 [01:04<00:13,  4.17it/s]\u001b[A\n",
            " 82%|████████▏ | 262/318 [01:04<00:13,  4.16it/s]\u001b[A\n",
            " 83%|████████▎ | 263/318 [01:05<00:13,  3.96it/s]\u001b[A\n",
            " 83%|████████▎ | 264/318 [01:05<00:13,  4.03it/s]\u001b[A\n",
            " 83%|████████▎ | 265/318 [01:05<00:12,  4.10it/s]\u001b[A\n",
            " 84%|████████▎ | 266/318 [01:05<00:12,  4.02it/s]\u001b[A\n",
            " 84%|████████▍ | 267/318 [01:06<00:13,  3.92it/s]\u001b[A\n",
            " 84%|████████▍ | 268/318 [01:06<00:12,  4.15it/s]\u001b[A\n",
            " 85%|████████▍ | 269/318 [01:06<00:11,  4.18it/s]\u001b[A\n",
            " 85%|████████▍ | 270/318 [01:06<00:10,  4.39it/s]\u001b[A\n",
            " 85%|████████▌ | 271/318 [01:06<00:10,  4.36it/s]\u001b[A\n",
            " 86%|████████▌ | 272/318 [01:07<00:10,  4.32it/s]\u001b[A\n",
            " 86%|████████▌ | 273/318 [01:07<00:10,  4.19it/s]\u001b[A\n",
            " 86%|████████▌ | 274/318 [01:07<00:09,  4.46it/s]\u001b[A\n",
            " 86%|████████▋ | 275/318 [01:07<00:09,  4.47it/s]\u001b[A\n",
            " 87%|████████▋ | 276/318 [01:08<00:09,  4.45it/s]\u001b[A\n",
            " 87%|████████▋ | 277/318 [01:08<00:09,  4.55it/s]\u001b[A\n",
            " 87%|████████▋ | 278/318 [01:08<00:08,  4.53it/s]\u001b[A\n",
            " 88%|████████▊ | 279/318 [01:08<00:08,  4.46it/s]\u001b[A\n",
            " 88%|████████▊ | 280/318 [01:09<00:09,  4.21it/s]\u001b[A\n",
            " 88%|████████▊ | 281/318 [01:09<00:09,  4.00it/s]\u001b[A\n",
            " 89%|████████▊ | 282/318 [01:09<00:09,  3.92it/s]\u001b[A\n",
            " 89%|████████▉ | 283/318 [01:09<00:09,  3.82it/s]\u001b[A\n",
            " 89%|████████▉ | 284/318 [01:10<00:08,  3.88it/s]\u001b[A\n",
            " 90%|████████▉ | 285/318 [01:10<00:08,  3.74it/s]\u001b[A\n",
            " 90%|████████▉ | 286/318 [01:10<00:07,  4.02it/s]\u001b[A\n",
            " 90%|█████████ | 287/318 [01:10<00:07,  3.99it/s]\u001b[A\n",
            " 91%|█████████ | 288/318 [01:11<00:07,  4.04it/s]\u001b[A\n",
            " 91%|█████████ | 289/318 [01:11<00:06,  4.31it/s]\u001b[A\n",
            " 91%|█████████ | 290/318 [01:11<00:06,  4.27it/s]\u001b[A\n",
            " 92%|█████████▏| 291/318 [01:11<00:06,  4.09it/s]\u001b[A\n",
            " 92%|█████████▏| 292/318 [01:12<00:06,  4.01it/s]\u001b[A\n",
            " 92%|█████████▏| 293/318 [01:12<00:06,  3.96it/s]\u001b[A\n",
            " 92%|█████████▏| 294/318 [01:12<00:06,  3.95it/s]\u001b[A\n",
            " 93%|█████████▎| 295/318 [01:12<00:06,  3.67it/s]\u001b[A\n",
            " 93%|█████████▎| 296/318 [01:13<00:05,  3.72it/s]\u001b[A\n",
            " 93%|█████████▎| 297/318 [01:13<00:05,  3.96it/s]\u001b[A\n",
            " 94%|█████████▎| 298/318 [01:13<00:04,  4.03it/s]\u001b[A\n",
            " 94%|█████████▍| 299/318 [01:13<00:04,  3.92it/s]\u001b[A\n",
            " 94%|█████████▍| 300/318 [01:14<00:04,  4.12it/s]\u001b[A\n",
            " 95%|█████████▍| 301/318 [01:14<00:04,  4.14it/s]\u001b[A\n",
            " 95%|█████████▍| 302/318 [01:14<00:04,  3.90it/s]\u001b[A\n",
            " 95%|█████████▌| 303/318 [01:14<00:03,  4.15it/s]\u001b[A\n",
            " 96%|█████████▌| 304/318 [01:15<00:03,  4.22it/s]\u001b[A\n",
            " 96%|█████████▌| 305/318 [01:15<00:02,  4.34it/s]\u001b[A\n",
            " 96%|█████████▌| 306/318 [01:15<00:02,  4.23it/s]\u001b[A\n",
            " 97%|█████████▋| 307/318 [01:15<00:02,  4.38it/s]\u001b[A\n",
            " 97%|█████████▋| 308/318 [01:15<00:02,  4.49it/s]\u001b[A\n",
            " 97%|█████████▋| 309/318 [01:16<00:02,  4.41it/s]\u001b[A\n",
            " 97%|█████████▋| 310/318 [01:16<00:01,  4.43it/s]\u001b[A\n",
            " 98%|█████████▊| 311/318 [01:16<00:01,  4.49it/s]\u001b[A\n",
            " 98%|█████████▊| 312/318 [01:16<00:01,  4.17it/s]\u001b[A\n",
            " 98%|█████████▊| 313/318 [01:17<00:01,  4.08it/s]\u001b[A\n",
            " 99%|█████████▊| 314/318 [01:17<00:00,  4.16it/s]\u001b[A\n",
            " 99%|█████████▉| 315/318 [01:17<00:00,  4.15it/s]\u001b[A\n",
            " 99%|█████████▉| 316/318 [01:17<00:00,  3.93it/s]\u001b[A\n",
            "100%|█████████▉| 317/318 [01:18<00:00,  4.00it/s]\u001b[A\n",
            "100%|██████████| 318/318 [01:18<00:00,  4.06it/s]\n",
            " 20%|██        | 2/10 [02:42<10:45, 80.65s/it]\n",
            "  0%|          | 0/318 [00:00<?, ?it/s]\u001b[A\n",
            "  0%|          | 1/318 [00:00<01:34,  3.35it/s]\u001b[A\n",
            "  1%|          | 2/318 [00:00<01:27,  3.62it/s]\u001b[A\n",
            "  1%|          | 3/318 [00:00<01:30,  3.48it/s]\u001b[A\n",
            "  1%|▏         | 4/318 [00:01<01:26,  3.63it/s]\u001b[A\n",
            "  2%|▏         | 5/318 [00:01<01:17,  4.05it/s]\u001b[A\n",
            "  2%|▏         | 6/318 [00:01<01:16,  4.06it/s]\u001b[A\n",
            "  2%|▏         | 7/318 [00:01<01:15,  4.12it/s]\u001b[A\n",
            "  3%|▎         | 8/318 [00:02<01:11,  4.32it/s]\u001b[A\n",
            "  3%|▎         | 9/318 [00:02<01:10,  4.41it/s]\u001b[A\n",
            "  3%|▎         | 10/318 [00:02<01:13,  4.22it/s]\u001b[A\n",
            "  3%|▎         | 11/318 [00:02<01:11,  4.29it/s]\u001b[A\n",
            "  4%|▍         | 12/318 [00:02<01:11,  4.26it/s]\u001b[A\n",
            "  4%|▍         | 13/318 [00:03<01:09,  4.36it/s]\u001b[A\n",
            "  4%|▍         | 14/318 [00:03<01:09,  4.40it/s]\u001b[A\n",
            "  5%|▍         | 15/318 [00:03<01:06,  4.53it/s]\u001b[A\n",
            "  5%|▌         | 16/318 [00:03<01:14,  4.08it/s]\u001b[A\n",
            "  5%|▌         | 17/318 [00:04<01:16,  3.96it/s]\u001b[A\n",
            "  6%|▌         | 18/318 [00:04<01:12,  4.13it/s]\u001b[A\n",
            "  6%|▌         | 19/318 [00:04<01:15,  3.95it/s]\u001b[A\n",
            "  6%|▋         | 20/318 [00:04<01:15,  3.95it/s]\u001b[A\n",
            "  7%|▋         | 21/318 [00:05<01:10,  4.19it/s]\u001b[A\n",
            "  7%|▋         | 22/318 [00:05<01:12,  4.08it/s]\u001b[A\n",
            "  7%|▋         | 23/318 [00:05<01:12,  4.09it/s]\u001b[A\n",
            "  8%|▊         | 24/318 [00:05<01:08,  4.27it/s]\u001b[A\n",
            "  8%|▊         | 25/318 [00:06<01:11,  4.12it/s]\u001b[A\n",
            "  8%|▊         | 26/318 [00:06<01:14,  3.91it/s]\u001b[A\n",
            "  8%|▊         | 27/318 [00:06<01:18,  3.71it/s]\u001b[A\n",
            "  9%|▉         | 28/318 [00:06<01:14,  3.89it/s]\u001b[A\n",
            "  9%|▉         | 29/318 [00:07<01:17,  3.71it/s]\u001b[A\n",
            "  9%|▉         | 30/318 [00:07<01:15,  3.82it/s]\u001b[A\n",
            " 10%|▉         | 31/318 [00:07<01:11,  4.01it/s]\u001b[A\n",
            " 10%|█         | 32/318 [00:07<01:14,  3.85it/s]\u001b[A\n",
            " 10%|█         | 33/318 [00:08<01:10,  4.05it/s]\u001b[A\n",
            " 11%|█         | 34/318 [00:08<01:10,  4.01it/s]\u001b[A\n",
            " 11%|█         | 35/318 [00:08<01:04,  4.40it/s]\u001b[A\n",
            " 11%|█▏        | 36/318 [00:08<01:06,  4.25it/s]\u001b[A\n",
            " 12%|█▏        | 37/318 [00:09<01:07,  4.15it/s]\u001b[A\n",
            " 12%|█▏        | 38/318 [00:09<01:13,  3.82it/s]\u001b[A\n",
            " 12%|█▏        | 39/318 [00:09<01:10,  3.98it/s]\u001b[A\n",
            " 13%|█▎        | 40/318 [00:09<01:07,  4.12it/s]\u001b[A\n",
            " 13%|█▎        | 41/318 [00:10<01:08,  4.03it/s]\u001b[A\n",
            " 13%|█▎        | 42/318 [00:10<01:08,  4.05it/s]\u001b[A\n",
            " 14%|█▎        | 43/318 [00:10<01:09,  3.95it/s]\u001b[A\n",
            " 14%|█▍        | 44/318 [00:10<01:14,  3.65it/s]\u001b[A\n",
            " 14%|█▍        | 45/318 [00:11<01:14,  3.67it/s]\u001b[A\n",
            " 14%|█▍        | 46/318 [00:11<01:12,  3.77it/s]\u001b[A\n",
            " 15%|█▍        | 47/318 [00:11<01:09,  3.90it/s]\u001b[A\n",
            " 15%|█▌        | 48/318 [00:11<01:05,  4.11it/s]\u001b[A\n",
            " 15%|█▌        | 49/318 [00:12<01:07,  3.96it/s]\u001b[A\n",
            " 16%|█▌        | 50/318 [00:12<01:03,  4.23it/s]\u001b[A\n",
            " 16%|█▌        | 51/318 [00:12<01:02,  4.29it/s]\u001b[A\n",
            " 16%|█▋        | 52/318 [00:12<01:02,  4.26it/s]\u001b[A\n",
            " 17%|█▋        | 53/318 [00:13<01:09,  3.83it/s]\u001b[A\n",
            " 17%|█▋        | 54/318 [00:13<01:07,  3.93it/s]\u001b[A\n",
            " 17%|█▋        | 55/318 [00:13<01:09,  3.81it/s]\u001b[A\n",
            " 18%|█▊        | 56/318 [00:13<01:04,  4.05it/s]\u001b[A\n",
            " 18%|█▊        | 57/318 [00:14<01:03,  4.08it/s]\u001b[A\n",
            " 18%|█▊        | 58/318 [00:14<01:03,  4.12it/s]\u001b[A\n",
            " 19%|█▊        | 59/318 [00:14<01:01,  4.20it/s]\u001b[A\n",
            " 19%|█▉        | 60/318 [00:14<01:00,  4.26it/s]\u001b[A\n",
            " 19%|█▉        | 61/318 [00:15<01:02,  4.12it/s]\u001b[A\n",
            " 19%|█▉        | 62/318 [00:15<01:02,  4.09it/s]\u001b[A\n",
            " 20%|█▉        | 63/318 [00:15<01:01,  4.18it/s]\u001b[A\n",
            " 20%|██        | 64/318 [00:15<00:57,  4.43it/s]\u001b[A\n",
            " 20%|██        | 65/318 [00:15<00:55,  4.52it/s]\u001b[A\n",
            " 21%|██        | 66/318 [00:16<00:56,  4.46it/s]\u001b[A\n",
            " 21%|██        | 67/318 [00:16<00:58,  4.28it/s]\u001b[A\n",
            " 21%|██▏       | 68/318 [00:16<01:02,  3.97it/s]\u001b[A\n",
            " 22%|██▏       | 69/318 [00:17<01:07,  3.70it/s]\u001b[A\n",
            " 22%|██▏       | 70/318 [00:17<01:06,  3.74it/s]\u001b[A\n",
            " 22%|██▏       | 71/318 [00:17<01:12,  3.38it/s]\u001b[A\n",
            " 23%|██▎       | 72/318 [00:18<01:14,  3.32it/s]\u001b[A\n",
            " 23%|██▎       | 73/318 [00:18<01:12,  3.36it/s]\u001b[A\n",
            " 23%|██▎       | 74/318 [00:18<01:09,  3.50it/s]\u001b[A\n",
            " 24%|██▎       | 75/318 [00:18<01:09,  3.51it/s]\u001b[A\n",
            " 24%|██▍       | 76/318 [00:19<01:11,  3.39it/s]\u001b[A\n",
            " 24%|██▍       | 77/318 [00:19<01:10,  3.43it/s]\u001b[A\n",
            " 25%|██▍       | 78/318 [00:19<01:06,  3.62it/s]\u001b[A\n",
            " 25%|██▍       | 79/318 [00:19<01:02,  3.83it/s]\u001b[A\n",
            " 25%|██▌       | 80/318 [00:20<01:00,  3.92it/s]\u001b[A\n",
            " 25%|██▌       | 81/318 [00:20<00:59,  3.97it/s]\u001b[A\n",
            " 26%|██▌       | 82/318 [00:20<00:56,  4.19it/s]\u001b[A\n",
            " 26%|██▌       | 83/318 [00:20<00:58,  4.03it/s]\u001b[A\n",
            " 26%|██▋       | 84/318 [00:21<00:57,  4.04it/s]\u001b[A\n",
            " 27%|██▋       | 85/318 [00:21<01:00,  3.84it/s]\u001b[A\n",
            " 27%|██▋       | 86/318 [00:21<01:05,  3.57it/s]\u001b[A\n",
            " 27%|██▋       | 87/318 [00:21<01:00,  3.85it/s]\u001b[A\n",
            " 28%|██▊       | 88/318 [00:22<00:59,  3.87it/s]\u001b[A\n",
            " 28%|██▊       | 89/318 [00:22<00:57,  3.97it/s]\u001b[A\n",
            " 28%|██▊       | 90/318 [00:22<00:53,  4.28it/s]\u001b[A\n",
            " 29%|██▊       | 91/318 [00:22<00:55,  4.12it/s]\u001b[A\n",
            " 29%|██▉       | 92/318 [00:23<00:57,  3.95it/s]\u001b[A\n",
            " 29%|██▉       | 93/318 [00:23<00:59,  3.76it/s]\u001b[A\n",
            " 30%|██▉       | 94/318 [00:23<01:00,  3.71it/s]\u001b[A\n",
            " 30%|██▉       | 95/318 [00:24<01:00,  3.67it/s]\u001b[A\n",
            " 30%|███       | 96/318 [00:24<00:59,  3.72it/s]\u001b[A\n",
            " 31%|███       | 97/318 [00:24<01:00,  3.64it/s]\u001b[A\n",
            " 31%|███       | 98/318 [00:24<00:57,  3.84it/s]\u001b[A\n",
            " 31%|███       | 99/318 [00:25<00:56,  3.87it/s]\u001b[A\n",
            " 31%|███▏      | 100/318 [00:25<00:53,  4.10it/s]\u001b[A\n",
            " 32%|███▏      | 101/318 [00:25<00:53,  4.09it/s]\u001b[A\n",
            " 32%|███▏      | 102/318 [00:25<00:51,  4.22it/s]\u001b[A\n",
            " 32%|███▏      | 103/318 [00:25<00:49,  4.38it/s]\u001b[A\n",
            " 33%|███▎      | 104/318 [00:26<00:50,  4.23it/s]\u001b[A\n",
            " 33%|███▎      | 105/318 [00:26<00:55,  3.84it/s]\u001b[A\n",
            " 33%|███▎      | 106/318 [00:26<00:53,  3.93it/s]\u001b[A\n",
            " 34%|███▎      | 107/318 [00:26<00:49,  4.30it/s]\u001b[A\n",
            " 34%|███▍      | 108/318 [00:27<00:54,  3.86it/s]\u001b[A\n",
            " 34%|███▍      | 109/318 [00:27<00:56,  3.72it/s]\u001b[A\n",
            " 35%|███▍      | 110/318 [00:27<00:53,  3.89it/s]\u001b[A\n",
            " 35%|███▍      | 111/318 [00:28<00:51,  3.99it/s]\u001b[A\n",
            " 35%|███▌      | 112/318 [00:28<00:48,  4.25it/s]\u001b[A\n",
            " 36%|███▌      | 113/318 [00:28<00:48,  4.20it/s]\u001b[A\n",
            " 36%|███▌      | 114/318 [00:28<00:48,  4.24it/s]\u001b[A\n",
            " 36%|███▌      | 115/318 [00:28<00:50,  4.03it/s]\u001b[A\n",
            " 36%|███▋      | 116/318 [00:29<00:51,  3.89it/s]\u001b[A\n",
            " 37%|███▋      | 117/318 [00:29<00:52,  3.83it/s]\u001b[A\n",
            " 37%|███▋      | 118/318 [00:29<00:52,  3.82it/s]\u001b[A\n",
            " 37%|███▋      | 119/318 [00:29<00:48,  4.09it/s]\u001b[A\n",
            " 38%|███▊      | 120/318 [00:30<00:51,  3.83it/s]\u001b[A\n",
            " 38%|███▊      | 121/318 [00:30<00:50,  3.87it/s]\u001b[A\n",
            " 38%|███▊      | 122/318 [00:30<00:48,  4.04it/s]\u001b[A\n",
            " 39%|███▊      | 123/318 [00:30<00:47,  4.11it/s]\u001b[A\n",
            " 39%|███▉      | 124/318 [00:31<00:48,  4.01it/s]\u001b[A\n",
            " 39%|███▉      | 125/318 [00:31<00:52,  3.65it/s]\u001b[A\n",
            " 40%|███▉      | 126/318 [00:31<00:50,  3.79it/s]\u001b[A\n",
            " 40%|███▉      | 127/318 [00:32<00:48,  3.96it/s]\u001b[A\n",
            " 40%|████      | 128/318 [00:32<00:50,  3.78it/s]\u001b[A\n",
            " 41%|████      | 129/318 [00:32<00:49,  3.85it/s]\u001b[A\n",
            " 41%|████      | 130/318 [00:32<00:49,  3.84it/s]\u001b[A\n",
            " 41%|████      | 131/318 [00:33<00:45,  4.08it/s]\u001b[A\n",
            " 42%|████▏     | 132/318 [00:33<00:43,  4.26it/s]\u001b[A\n",
            " 42%|████▏     | 133/318 [00:33<00:46,  3.97it/s]\u001b[A\n",
            " 42%|████▏     | 134/318 [00:33<00:48,  3.80it/s]\u001b[A\n",
            " 42%|████▏     | 135/318 [00:34<00:47,  3.87it/s]\u001b[A\n",
            " 43%|████▎     | 136/318 [00:34<00:45,  4.00it/s]\u001b[A\n",
            " 43%|████▎     | 137/318 [00:34<00:45,  3.99it/s]\u001b[A\n",
            " 43%|████▎     | 138/318 [00:34<00:45,  4.00it/s]\u001b[A\n",
            " 44%|████▎     | 139/318 [00:35<00:45,  3.92it/s]\u001b[A\n",
            " 44%|████▍     | 140/318 [00:35<00:44,  3.97it/s]\u001b[A\n",
            " 44%|████▍     | 141/318 [00:35<00:49,  3.55it/s]\u001b[A\n",
            " 45%|████▍     | 142/318 [00:35<00:50,  3.52it/s]\u001b[A\n",
            " 45%|████▍     | 143/318 [00:36<00:48,  3.64it/s]\u001b[A\n",
            " 45%|████▌     | 144/318 [00:36<00:45,  3.85it/s]\u001b[A\n",
            " 46%|████▌     | 145/318 [00:36<00:43,  4.00it/s]\u001b[A\n",
            " 46%|████▌     | 146/318 [00:36<00:44,  3.86it/s]\u001b[A\n",
            " 46%|████▌     | 147/318 [00:37<00:43,  3.96it/s]\u001b[A\n",
            " 47%|████▋     | 148/318 [00:37<00:41,  4.14it/s]\u001b[A\n",
            " 47%|████▋     | 149/318 [00:37<00:41,  4.10it/s]\u001b[A\n",
            " 47%|████▋     | 150/318 [00:37<00:39,  4.22it/s]\u001b[A\n",
            " 47%|████▋     | 151/318 [00:38<00:39,  4.21it/s]\u001b[A\n",
            " 48%|████▊     | 152/318 [00:38<00:37,  4.42it/s]\u001b[A\n",
            " 48%|████▊     | 153/318 [00:38<00:39,  4.23it/s]\u001b[A\n",
            " 48%|████▊     | 154/318 [00:38<00:40,  4.08it/s]\u001b[A\n",
            " 49%|████▊     | 155/318 [00:39<00:41,  3.91it/s]\u001b[A\n",
            " 49%|████▉     | 156/318 [00:39<00:37,  4.30it/s]\u001b[A\n",
            " 49%|████▉     | 157/318 [00:39<00:36,  4.38it/s]\u001b[A\n",
            " 50%|████▉     | 158/318 [00:39<00:35,  4.45it/s]\u001b[A\n",
            " 50%|█████     | 159/318 [00:40<00:39,  4.08it/s]\u001b[A\n",
            " 50%|█████     | 160/318 [00:40<00:39,  3.99it/s]\u001b[A\n",
            " 51%|█████     | 161/318 [00:40<00:41,  3.78it/s]\u001b[A\n",
            " 51%|█████     | 162/318 [00:40<00:39,  3.99it/s]\u001b[A\n",
            " 51%|█████▏    | 163/318 [00:41<00:36,  4.29it/s]\u001b[A\n",
            " 52%|█████▏    | 164/318 [00:41<00:35,  4.28it/s]\u001b[A\n",
            " 52%|█████▏    | 165/318 [00:41<00:38,  3.99it/s]\u001b[A\n",
            " 52%|█████▏    | 166/318 [00:41<00:38,  3.95it/s]\u001b[A\n",
            " 53%|█████▎    | 167/318 [00:42<00:38,  3.93it/s]\u001b[A\n",
            " 53%|█████▎    | 168/318 [00:42<00:37,  4.04it/s]\u001b[A\n",
            " 53%|█████▎    | 169/318 [00:42<00:39,  3.81it/s]\u001b[A\n",
            " 53%|█████▎    | 170/318 [00:42<00:38,  3.88it/s]\u001b[A\n",
            " 54%|█████▍    | 171/318 [00:43<00:36,  4.06it/s]\u001b[A\n",
            " 54%|█████▍    | 172/318 [00:43<00:34,  4.20it/s]\u001b[A\n",
            " 54%|█████▍    | 173/318 [00:43<00:36,  4.00it/s]\u001b[A\n",
            " 55%|█████▍    | 174/318 [00:43<00:36,  3.97it/s]\u001b[A\n",
            " 55%|█████▌    | 175/318 [00:44<00:36,  3.97it/s]\u001b[A\n",
            " 55%|█████▌    | 176/318 [00:44<00:36,  3.92it/s]\u001b[A\n",
            " 56%|█████▌    | 177/318 [00:44<00:36,  3.90it/s]\u001b[A\n",
            " 56%|█████▌    | 178/318 [00:44<00:33,  4.18it/s]\u001b[A\n",
            " 56%|█████▋    | 179/318 [00:45<00:32,  4.27it/s]\u001b[A\n",
            " 57%|█████▋    | 180/318 [00:45<00:31,  4.35it/s]\u001b[A\n",
            " 57%|█████▋    | 181/318 [00:45<00:32,  4.22it/s]\u001b[A\n",
            " 57%|█████▋    | 182/318 [00:45<00:32,  4.15it/s]\u001b[A\n",
            " 58%|█████▊    | 183/318 [00:45<00:32,  4.17it/s]\u001b[A\n",
            " 58%|█████▊    | 184/318 [00:46<00:31,  4.23it/s]\u001b[A\n",
            " 58%|█████▊    | 185/318 [00:46<00:33,  4.02it/s]\u001b[A\n",
            " 58%|█████▊    | 186/318 [00:46<00:33,  3.91it/s]\u001b[A\n",
            " 59%|█████▉    | 187/318 [00:47<00:34,  3.78it/s]\u001b[A\n",
            " 59%|█████▉    | 188/318 [00:47<00:33,  3.92it/s]\u001b[A\n",
            " 59%|█████▉    | 189/318 [00:47<00:32,  3.99it/s]\u001b[A\n",
            " 60%|█████▉    | 190/318 [00:47<00:31,  4.06it/s]\u001b[A\n",
            " 60%|██████    | 191/318 [00:47<00:31,  4.02it/s]\u001b[A\n",
            " 60%|██████    | 192/318 [00:48<00:29,  4.21it/s]\u001b[A\n",
            " 61%|██████    | 193/318 [00:48<00:28,  4.37it/s]\u001b[A\n",
            " 61%|██████    | 194/318 [00:48<00:28,  4.38it/s]\u001b[A\n",
            " 61%|██████▏   | 195/318 [00:48<00:27,  4.42it/s]\u001b[A\n",
            " 62%|██████▏   | 196/318 [00:49<00:26,  4.63it/s]\u001b[A\n",
            " 62%|██████▏   | 197/318 [00:49<00:28,  4.31it/s]\u001b[A\n",
            " 62%|██████▏   | 198/318 [00:49<00:27,  4.31it/s]\u001b[A\n",
            " 63%|██████▎   | 199/318 [00:49<00:27,  4.27it/s]\u001b[A\n",
            " 63%|██████▎   | 200/318 [00:50<00:29,  4.01it/s]\u001b[A\n",
            " 63%|██████▎   | 201/318 [00:50<00:30,  3.80it/s]\u001b[A\n",
            " 64%|██████▎   | 202/318 [00:50<00:28,  4.00it/s]\u001b[A\n",
            " 64%|██████▍   | 203/318 [00:50<00:27,  4.13it/s]\u001b[A\n",
            " 64%|██████▍   | 204/318 [00:51<00:26,  4.24it/s]\u001b[A\n",
            " 64%|██████▍   | 205/318 [00:51<00:24,  4.55it/s]\u001b[A\n",
            " 65%|██████▍   | 206/318 [00:51<00:24,  4.54it/s]\u001b[A\n",
            " 65%|██████▌   | 207/318 [00:51<00:24,  4.48it/s]\u001b[A\n",
            " 65%|██████▌   | 208/318 [00:51<00:25,  4.34it/s]\u001b[A\n",
            " 66%|██████▌   | 209/318 [00:52<00:26,  4.18it/s]\u001b[A\n",
            " 66%|██████▌   | 210/318 [00:52<00:28,  3.85it/s]\u001b[A\n",
            " 66%|██████▋   | 211/318 [00:52<00:29,  3.66it/s]\u001b[A\n",
            " 67%|██████▋   | 212/318 [00:53<00:29,  3.58it/s]\u001b[A\n",
            " 67%|██████▋   | 213/318 [00:53<00:26,  3.94it/s]\u001b[A\n",
            " 67%|██████▋   | 214/318 [00:53<00:26,  3.92it/s]\u001b[A\n",
            " 68%|██████▊   | 215/318 [00:53<00:26,  3.86it/s]\u001b[A\n",
            " 68%|██████▊   | 216/318 [00:54<00:24,  4.15it/s]\u001b[A\n",
            " 68%|██████▊   | 217/318 [00:54<00:24,  4.09it/s]\u001b[A\n",
            " 69%|██████▊   | 218/318 [00:54<00:25,  3.90it/s]\u001b[A\n",
            " 69%|██████▉   | 219/318 [00:54<00:25,  3.92it/s]\u001b[A\n",
            " 69%|██████▉   | 220/318 [00:55<00:25,  3.85it/s]\u001b[A\n",
            " 69%|██████▉   | 221/318 [00:55<00:23,  4.11it/s]\u001b[A\n",
            " 70%|██████▉   | 222/318 [00:55<00:21,  4.37it/s]\u001b[A\n",
            " 70%|███████   | 223/318 [00:55<00:21,  4.51it/s]\u001b[A\n",
            " 70%|███████   | 224/318 [00:55<00:22,  4.22it/s]\u001b[A\n",
            " 71%|███████   | 225/318 [00:56<00:22,  4.14it/s]\u001b[A\n",
            " 71%|███████   | 226/318 [00:56<00:21,  4.25it/s]\u001b[A\n",
            " 71%|███████▏  | 227/318 [00:56<00:22,  4.12it/s]\u001b[A\n",
            " 72%|███████▏  | 228/318 [00:56<00:21,  4.25it/s]\u001b[A\n",
            " 72%|███████▏  | 229/318 [00:57<00:21,  4.13it/s]\u001b[A\n",
            " 72%|███████▏  | 230/318 [00:57<00:20,  4.20it/s]\u001b[A\n",
            " 73%|███████▎  | 231/318 [00:57<00:20,  4.34it/s]\u001b[A\n",
            " 73%|███████▎  | 232/318 [00:57<00:19,  4.44it/s]\u001b[A\n",
            " 73%|███████▎  | 233/318 [00:58<00:18,  4.64it/s]\u001b[A\n",
            " 74%|███████▎  | 234/318 [00:58<00:19,  4.40it/s]\u001b[A\n",
            " 74%|███████▍  | 235/318 [00:58<00:18,  4.57it/s]\u001b[A\n",
            " 74%|███████▍  | 236/318 [00:58<00:19,  4.24it/s]\u001b[A\n",
            " 75%|███████▍  | 237/318 [00:58<00:19,  4.20it/s]\u001b[A\n",
            " 75%|███████▍  | 238/318 [00:59<00:18,  4.28it/s]\u001b[A\n",
            " 75%|███████▌  | 239/318 [00:59<00:17,  4.40it/s]\u001b[A\n",
            " 75%|███████▌  | 240/318 [00:59<00:18,  4.23it/s]\u001b[A\n",
            " 76%|███████▌  | 241/318 [00:59<00:19,  3.90it/s]\u001b[A\n",
            " 76%|███████▌  | 242/318 [01:00<00:18,  4.16it/s]\u001b[A\n",
            " 76%|███████▋  | 243/318 [01:00<00:19,  3.79it/s]\u001b[A\n",
            " 77%|███████▋  | 244/318 [01:00<00:17,  4.11it/s]\u001b[A\n",
            " 77%|███████▋  | 245/318 [01:00<00:17,  4.17it/s]\u001b[A\n",
            " 77%|███████▋  | 246/318 [01:01<00:16,  4.24it/s]\u001b[A\n",
            " 78%|███████▊  | 247/318 [01:01<00:17,  4.10it/s]\u001b[A\n",
            " 78%|███████▊  | 248/318 [01:01<00:16,  4.33it/s]\u001b[A\n",
            " 78%|███████▊  | 249/318 [01:01<00:16,  4.30it/s]\u001b[A\n",
            " 79%|███████▊  | 250/318 [01:02<00:16,  4.20it/s]\u001b[A\n",
            " 79%|███████▉  | 251/318 [01:02<00:16,  4.10it/s]\u001b[A\n",
            " 79%|███████▉  | 252/318 [01:02<00:17,  3.75it/s]\u001b[A\n",
            " 80%|███████▉  | 253/318 [01:02<00:17,  3.76it/s]\u001b[A\n",
            " 80%|███████▉  | 254/318 [01:03<00:16,  3.84it/s]\u001b[A\n",
            " 80%|████████  | 255/318 [01:03<00:16,  3.84it/s]\u001b[A\n",
            " 81%|████████  | 256/318 [01:03<00:16,  3.75it/s]\u001b[A\n",
            " 81%|████████  | 257/318 [01:03<00:14,  4.08it/s]\u001b[A\n",
            " 81%|████████  | 258/318 [01:04<00:14,  4.00it/s]\u001b[A\n",
            " 81%|████████▏ | 259/318 [01:04<00:14,  4.15it/s]\u001b[A\n",
            " 82%|████████▏ | 260/318 [01:04<00:14,  3.91it/s]\u001b[A\n",
            " 82%|████████▏ | 261/318 [01:04<00:13,  4.20it/s]\u001b[A\n",
            " 82%|████████▏ | 262/318 [01:05<00:13,  4.31it/s]\u001b[A\n",
            " 83%|████████▎ | 263/318 [01:05<00:12,  4.50it/s]\u001b[A\n",
            " 83%|████████▎ | 264/318 [01:05<00:12,  4.28it/s]\u001b[A\n",
            " 83%|████████▎ | 265/318 [01:05<00:13,  3.87it/s]\u001b[A\n",
            " 84%|████████▎ | 266/318 [01:06<00:13,  3.94it/s]\u001b[A\n",
            " 84%|████████▍ | 267/318 [01:06<00:12,  4.12it/s]\u001b[A\n",
            " 84%|████████▍ | 268/318 [01:06<00:12,  4.01it/s]\u001b[A\n",
            " 85%|████████▍ | 269/318 [01:06<00:12,  3.92it/s]\u001b[A\n",
            " 85%|████████▍ | 270/318 [01:07<00:12,  3.94it/s]\u001b[A\n",
            " 85%|████████▌ | 271/318 [01:07<00:11,  3.94it/s]\u001b[A\n",
            " 86%|████████▌ | 272/318 [01:07<00:11,  4.03it/s]\u001b[A\n",
            " 86%|████████▌ | 273/318 [01:07<00:11,  3.94it/s]\u001b[A\n",
            " 86%|████████▌ | 274/318 [01:08<00:11,  3.91it/s]\u001b[A\n",
            " 86%|████████▋ | 275/318 [01:08<00:10,  4.01it/s]\u001b[A\n",
            " 87%|████████▋ | 276/318 [01:08<00:10,  4.16it/s]\u001b[A\n",
            " 87%|████████▋ | 277/318 [01:08<00:09,  4.37it/s]\u001b[A\n",
            " 87%|████████▋ | 278/318 [01:09<00:09,  4.23it/s]\u001b[A\n",
            " 88%|████████▊ | 279/318 [01:09<00:09,  4.26it/s]\u001b[A\n",
            " 88%|████████▊ | 280/318 [01:09<00:08,  4.53it/s]\u001b[A\n",
            " 88%|████████▊ | 281/318 [01:09<00:08,  4.50it/s]\u001b[A\n",
            " 89%|████████▊ | 282/318 [01:09<00:08,  4.46it/s]\u001b[A\n",
            " 89%|████████▉ | 283/318 [01:10<00:07,  4.58it/s]\u001b[A\n",
            " 89%|████████▉ | 284/318 [01:10<00:07,  4.45it/s]\u001b[A\n",
            " 90%|████████▉ | 285/318 [01:10<00:07,  4.52it/s]\u001b[A\n",
            " 90%|████████▉ | 286/318 [01:10<00:07,  4.48it/s]\u001b[A\n",
            " 90%|█████████ | 287/318 [01:11<00:06,  4.65it/s]\u001b[A\n",
            " 91%|█████████ | 288/318 [01:11<00:06,  4.39it/s]\u001b[A\n",
            " 91%|█████████ | 289/318 [01:11<00:06,  4.43it/s]\u001b[A\n",
            " 91%|█████████ | 290/318 [01:11<00:06,  4.19it/s]\u001b[A\n",
            " 92%|█████████▏| 291/318 [01:12<00:06,  4.07it/s]\u001b[A\n",
            " 92%|█████████▏| 292/318 [01:12<00:06,  4.10it/s]\u001b[A\n",
            " 92%|█████████▏| 293/318 [01:12<00:06,  3.96it/s]\u001b[A\n",
            " 92%|█████████▏| 294/318 [01:12<00:06,  3.99it/s]\u001b[A\n",
            " 93%|█████████▎| 295/318 [01:13<00:05,  3.85it/s]\u001b[A\n",
            " 93%|█████████▎| 296/318 [01:13<00:05,  3.93it/s]\u001b[A\n",
            " 93%|█████████▎| 297/318 [01:13<00:06,  3.42it/s]\u001b[A\n",
            " 94%|█████████▎| 298/318 [01:13<00:05,  3.64it/s]\u001b[A\n",
            " 94%|█████████▍| 299/318 [01:14<00:05,  3.48it/s]\u001b[A\n",
            " 94%|█████████▍| 300/318 [01:14<00:04,  3.66it/s]\u001b[A\n",
            " 95%|█████████▍| 301/318 [01:14<00:04,  3.63it/s]\u001b[A\n",
            " 95%|█████████▍| 302/318 [01:15<00:04,  3.66it/s]\u001b[A\n",
            " 95%|█████████▌| 303/318 [01:15<00:03,  3.95it/s]\u001b[A\n",
            " 96%|█████████▌| 304/318 [01:15<00:03,  4.11it/s]\u001b[A\n",
            " 96%|█████████▌| 305/318 [01:15<00:03,  3.79it/s]\u001b[A\n",
            " 96%|█████████▌| 306/318 [01:15<00:03,  3.93it/s]\u001b[A\n",
            " 97%|█████████▋| 307/318 [01:16<00:02,  4.20it/s]\u001b[A\n",
            " 97%|█████████▋| 308/318 [01:16<00:02,  4.50it/s]\u001b[A\n",
            " 97%|█████████▋| 309/318 [01:16<00:02,  4.36it/s]\u001b[A\n",
            " 97%|█████████▋| 310/318 [01:16<00:01,  4.34it/s]\u001b[A\n",
            " 98%|█████████▊| 311/318 [01:17<00:01,  4.24it/s]\u001b[A\n",
            " 98%|█████████▊| 312/318 [01:17<00:01,  4.09it/s]\u001b[A\n",
            " 98%|█████████▊| 313/318 [01:17<00:01,  4.12it/s]\u001b[A\n",
            " 99%|█████████▊| 314/318 [01:17<00:00,  4.22it/s]\u001b[A\n",
            " 99%|█████████▉| 315/318 [01:18<00:00,  4.32it/s]\u001b[A\n",
            " 99%|█████████▉| 316/318 [01:18<00:00,  4.34it/s]\u001b[A\n",
            "100%|█████████▉| 317/318 [01:18<00:00,  4.31it/s]\u001b[A\n",
            "100%|██████████| 318/318 [01:18<00:00,  4.03it/s]\n",
            " 30%|███       | 3/10 [04:01<09:19, 79.87s/it]\n",
            "  0%|          | 0/318 [00:00<?, ?it/s]\u001b[A\n",
            "  0%|          | 1/318 [00:00<02:05,  2.52it/s]\u001b[A\n",
            "  1%|          | 2/318 [00:00<01:39,  3.18it/s]\u001b[A\n",
            "  1%|          | 3/318 [00:00<01:29,  3.50it/s]\u001b[A\n",
            "  1%|▏         | 4/318 [00:01<01:26,  3.65it/s]\u001b[A\n",
            "  2%|▏         | 5/318 [00:01<01:25,  3.67it/s]\u001b[A\n",
            "  2%|▏         | 6/318 [00:01<01:19,  3.93it/s]\u001b[A\n",
            "  2%|▏         | 7/318 [00:01<01:15,  4.12it/s]\u001b[A\n",
            "  3%|▎         | 8/318 [00:02<01:10,  4.37it/s]\u001b[A\n",
            "  3%|▎         | 9/318 [00:02<01:08,  4.51it/s]\u001b[A\n",
            "  3%|▎         | 10/318 [00:02<01:13,  4.21it/s]\u001b[A\n",
            "  3%|▎         | 11/318 [00:02<01:10,  4.36it/s]\u001b[A\n",
            "  4%|▍         | 12/318 [00:02<01:09,  4.38it/s]\u001b[A\n",
            "  4%|▍         | 13/318 [00:03<01:11,  4.28it/s]\u001b[A\n",
            "  4%|▍         | 14/318 [00:03<01:16,  3.99it/s]\u001b[A\n",
            "  5%|▍         | 15/318 [00:03<01:11,  4.26it/s]\u001b[A\n",
            "  5%|▌         | 16/318 [00:04<01:15,  4.01it/s]\u001b[A\n",
            "  5%|▌         | 17/318 [00:04<01:12,  4.17it/s]\u001b[A\n",
            "  6%|▌         | 18/318 [00:04<01:11,  4.20it/s]\u001b[A\n",
            "  6%|▌         | 19/318 [00:04<01:10,  4.27it/s]\u001b[A\n",
            "  6%|▋         | 20/318 [00:04<01:11,  4.19it/s]\u001b[A\n",
            "  7%|▋         | 21/318 [00:05<01:12,  4.07it/s]\u001b[A\n",
            "  7%|▋         | 22/318 [00:05<01:12,  4.07it/s]\u001b[A\n",
            "  7%|▋         | 23/318 [00:05<01:12,  4.09it/s]\u001b[A\n",
            "  8%|▊         | 24/318 [00:05<01:08,  4.28it/s]\u001b[A\n",
            "  8%|▊         | 25/318 [00:06<01:15,  3.90it/s]\u001b[A\n",
            "  8%|▊         | 26/318 [00:06<01:12,  4.01it/s]\u001b[A\n",
            "  8%|▊         | 27/318 [00:06<01:11,  4.08it/s]\u001b[A\n",
            "  9%|▉         | 28/318 [00:06<01:12,  3.99it/s]\u001b[A\n",
            "  9%|▉         | 29/318 [00:07<01:13,  3.93it/s]\u001b[A\n",
            "  9%|▉         | 30/318 [00:07<01:14,  3.87it/s]\u001b[A\n",
            " 10%|▉         | 31/318 [00:07<01:16,  3.77it/s]\u001b[A\n",
            " 10%|█         | 32/318 [00:07<01:14,  3.84it/s]\u001b[A\n",
            " 10%|█         | 33/318 [00:08<01:18,  3.63it/s]\u001b[A\n",
            " 11%|█         | 34/318 [00:08<01:17,  3.67it/s]\u001b[A\n",
            " 11%|█         | 35/318 [00:08<01:14,  3.80it/s]\u001b[A\n",
            " 11%|█▏        | 36/318 [00:09<01:10,  3.98it/s]\u001b[A\n",
            " 12%|█▏        | 37/318 [00:09<01:15,  3.71it/s]\u001b[A\n",
            " 12%|█▏        | 38/318 [00:09<01:17,  3.63it/s]\u001b[A\n",
            " 12%|█▏        | 39/318 [00:09<01:11,  3.90it/s]\u001b[A\n",
            " 13%|█▎        | 40/318 [00:10<01:12,  3.81it/s]\u001b[A\n",
            " 13%|█▎        | 41/318 [00:10<01:10,  3.92it/s]\u001b[A\n",
            " 13%|█▎        | 42/318 [00:10<01:13,  3.78it/s]\u001b[A\n",
            " 14%|█▎        | 43/318 [00:10<01:12,  3.80it/s]\u001b[A\n",
            " 14%|█▍        | 44/318 [00:11<01:07,  4.06it/s]\u001b[A\n",
            " 14%|█▍        | 45/318 [00:11<01:08,  4.01it/s]\u001b[A\n",
            " 14%|█▍        | 46/318 [00:11<01:03,  4.30it/s]\u001b[A\n",
            " 15%|█▍        | 47/318 [00:11<01:06,  4.09it/s]\u001b[A\n",
            " 15%|█▌        | 48/318 [00:12<01:09,  3.91it/s]\u001b[A\n",
            " 15%|█▌        | 49/318 [00:12<01:08,  3.95it/s]\u001b[A\n",
            " 16%|█▌        | 50/318 [00:12<01:08,  3.92it/s]\u001b[A\n",
            " 16%|█▌        | 51/318 [00:12<01:09,  3.85it/s]\u001b[A\n",
            " 16%|█▋        | 52/318 [00:13<01:11,  3.72it/s]\u001b[A\n",
            " 17%|█▋        | 53/318 [00:13<01:10,  3.77it/s]\u001b[A\n",
            " 17%|█▋        | 54/318 [00:13<01:06,  4.00it/s]\u001b[A\n",
            " 17%|█▋        | 55/318 [00:13<01:08,  3.86it/s]\u001b[A\n",
            " 18%|█▊        | 56/318 [00:14<01:05,  4.02it/s]\u001b[A\n",
            " 18%|█▊        | 57/318 [00:14<01:06,  3.92it/s]\u001b[A\n",
            " 18%|█▊        | 58/318 [00:14<01:03,  4.11it/s]\u001b[A\n",
            " 19%|█▊        | 59/318 [00:14<01:01,  4.24it/s]\u001b[A\n",
            " 19%|█▉        | 60/318 [00:15<00:59,  4.33it/s]\u001b[A\n",
            " 19%|█▉        | 61/318 [00:15<01:01,  4.21it/s]\u001b[A\n",
            " 19%|█▉        | 62/318 [00:15<00:59,  4.27it/s]\u001b[A\n",
            " 20%|█▉        | 63/318 [00:15<01:01,  4.16it/s]\u001b[A\n",
            " 20%|██        | 64/318 [00:16<01:02,  4.07it/s]\u001b[A\n",
            " 20%|██        | 65/318 [00:16<01:00,  4.19it/s]\u001b[A\n",
            " 21%|██        | 66/318 [00:16<01:01,  4.08it/s]\u001b[A\n",
            " 21%|██        | 67/318 [00:16<00:57,  4.35it/s]\u001b[A\n",
            " 21%|██▏       | 68/318 [00:16<00:56,  4.45it/s]\u001b[A\n",
            " 22%|██▏       | 69/318 [00:17<00:53,  4.66it/s]\u001b[A\n",
            " 22%|██▏       | 70/318 [00:17<00:56,  4.43it/s]\u001b[A\n",
            " 22%|██▏       | 71/318 [00:17<00:56,  4.39it/s]\u001b[A\n",
            " 23%|██▎       | 72/318 [00:17<00:58,  4.17it/s]\u001b[A\n",
            " 23%|██▎       | 73/318 [00:18<01:00,  4.03it/s]\u001b[A\n",
            " 23%|██▎       | 74/318 [00:18<01:00,  4.06it/s]\u001b[A\n",
            " 24%|██▎       | 75/318 [00:18<01:00,  3.99it/s]\u001b[A\n",
            " 24%|██▍       | 76/318 [00:18<00:59,  4.06it/s]\u001b[A\n",
            " 24%|██▍       | 77/318 [00:19<01:00,  3.99it/s]\u001b[A\n",
            " 25%|██▍       | 78/318 [00:19<00:58,  4.12it/s]\u001b[A\n",
            " 25%|██▍       | 79/318 [00:19<00:53,  4.46it/s]\u001b[A\n",
            " 25%|██▌       | 80/318 [00:19<00:51,  4.58it/s]\u001b[A\n",
            " 25%|██▌       | 81/318 [00:20<00:53,  4.44it/s]\u001b[A\n",
            " 26%|██▌       | 82/318 [00:20<00:53,  4.44it/s]\u001b[A\n",
            " 26%|██▌       | 83/318 [00:20<00:51,  4.60it/s]\u001b[A\n",
            " 26%|██▋       | 84/318 [00:20<00:52,  4.46it/s]\u001b[A\n",
            " 27%|██▋       | 85/318 [00:20<00:54,  4.28it/s]\u001b[A\n",
            " 27%|██▋       | 86/318 [00:21<00:53,  4.36it/s]\u001b[A\n",
            " 27%|██▋       | 87/318 [00:21<00:50,  4.58it/s]\u001b[A\n",
            " 28%|██▊       | 88/318 [00:21<00:53,  4.33it/s]\u001b[A\n",
            " 28%|██▊       | 89/318 [00:21<00:53,  4.27it/s]\u001b[A\n",
            " 28%|██▊       | 90/318 [00:22<00:54,  4.20it/s]\u001b[A\n",
            " 29%|██▊       | 91/318 [00:22<00:57,  3.98it/s]\u001b[A\n",
            " 29%|██▉       | 92/318 [00:22<00:56,  4.02it/s]\u001b[A\n",
            " 29%|██▉       | 93/318 [00:22<00:55,  4.07it/s]\u001b[A\n",
            " 30%|██▉       | 94/318 [00:23<01:00,  3.71it/s]\u001b[A\n",
            " 30%|██▉       | 95/318 [00:23<00:59,  3.73it/s]\u001b[A\n",
            " 30%|███       | 96/318 [00:23<00:59,  3.72it/s]\u001b[A\n",
            " 31%|███       | 97/318 [00:24<01:00,  3.63it/s]\u001b[A\n",
            " 31%|███       | 98/318 [00:24<00:57,  3.84it/s]\u001b[A\n",
            " 31%|███       | 99/318 [00:24<00:54,  4.03it/s]\u001b[A\n",
            " 31%|███▏      | 100/318 [00:24<00:54,  3.98it/s]\u001b[A\n",
            " 32%|███▏      | 101/318 [00:24<00:53,  4.08it/s]\u001b[A\n",
            " 32%|███▏      | 102/318 [00:25<00:54,  3.98it/s]\u001b[A\n",
            " 32%|███▏      | 103/318 [00:25<00:58,  3.69it/s]\u001b[A\n",
            " 33%|███▎      | 104/318 [00:25<00:55,  3.83it/s]\u001b[A\n",
            " 33%|███▎      | 105/318 [00:25<00:50,  4.22it/s]\u001b[A\n",
            " 33%|███▎      | 106/318 [00:26<00:49,  4.27it/s]\u001b[A\n",
            " 34%|███▎      | 107/318 [00:26<00:48,  4.35it/s]\u001b[A\n",
            " 34%|███▍      | 108/318 [00:26<00:51,  4.07it/s]\u001b[A\n",
            " 34%|███▍      | 109/318 [00:26<00:53,  3.90it/s]\u001b[A\n",
            " 35%|███▍      | 110/318 [00:27<00:50,  4.10it/s]\u001b[A\n",
            " 35%|███▍      | 111/318 [00:27<00:49,  4.19it/s]\u001b[A\n",
            " 35%|███▌      | 112/318 [00:27<00:54,  3.79it/s]\u001b[A\n",
            " 36%|███▌      | 113/318 [00:27<00:54,  3.79it/s]\u001b[A\n",
            " 36%|███▌      | 114/318 [00:28<00:53,  3.84it/s]\u001b[A\n",
            " 36%|███▌      | 115/318 [00:28<00:51,  3.96it/s]\u001b[A\n",
            " 36%|███▋      | 116/318 [00:28<00:53,  3.77it/s]\u001b[A\n",
            " 37%|███▋      | 117/318 [00:29<00:51,  3.90it/s]\u001b[A\n",
            " 37%|███▋      | 118/318 [00:29<00:47,  4.23it/s]\u001b[A\n",
            " 37%|███▋      | 119/318 [00:29<00:49,  4.00it/s]\u001b[A\n",
            " 38%|███▊      | 120/318 [00:29<00:49,  4.04it/s]\u001b[A\n",
            " 38%|███▊      | 121/318 [00:29<00:47,  4.11it/s]\u001b[A\n",
            " 38%|███▊      | 122/318 [00:30<00:47,  4.12it/s]\u001b[A\n",
            " 39%|███▊      | 123/318 [00:30<00:46,  4.20it/s]\u001b[A\n",
            " 39%|███▉      | 124/318 [00:30<00:47,  4.10it/s]\u001b[A\n",
            " 39%|███▉      | 125/318 [00:30<00:47,  4.07it/s]\u001b[A\n",
            " 40%|███▉      | 126/318 [00:31<00:48,  3.96it/s]\u001b[A\n",
            " 40%|███▉      | 127/318 [00:31<00:46,  4.11it/s]\u001b[A\n",
            " 40%|████      | 128/318 [00:31<00:47,  4.04it/s]\u001b[A\n",
            " 41%|████      | 129/318 [00:31<00:47,  3.96it/s]\u001b[A\n",
            " 41%|████      | 130/318 [00:32<00:46,  4.03it/s]\u001b[A\n",
            " 41%|████      | 131/318 [00:32<00:44,  4.17it/s]\u001b[A\n",
            " 42%|████▏     | 132/318 [00:32<00:44,  4.13it/s]\u001b[A\n",
            " 42%|████▏     | 133/318 [00:32<00:48,  3.85it/s]\u001b[A\n",
            " 42%|████▏     | 134/318 [00:33<00:47,  3.83it/s]\u001b[A\n",
            " 42%|████▏     | 135/318 [00:33<00:45,  4.04it/s]\u001b[A\n",
            " 43%|████▎     | 136/318 [00:33<00:44,  4.07it/s]\u001b[A\n",
            " 43%|████▎     | 137/318 [00:33<00:46,  3.87it/s]\u001b[A\n",
            " 43%|████▎     | 138/318 [00:34<00:45,  3.92it/s]\u001b[A\n",
            " 44%|████▎     | 139/318 [00:34<00:44,  4.07it/s]\u001b[A\n",
            " 44%|████▍     | 140/318 [00:34<00:42,  4.22it/s]\u001b[A\n",
            " 44%|████▍     | 141/318 [00:34<00:41,  4.25it/s]\u001b[A\n",
            " 45%|████▍     | 142/318 [00:35<00:42,  4.17it/s]\u001b[A\n",
            " 45%|████▍     | 143/318 [00:35<00:40,  4.32it/s]\u001b[A\n",
            " 45%|████▌     | 144/318 [00:35<00:40,  4.26it/s]\u001b[A\n",
            " 46%|████▌     | 145/318 [00:35<00:46,  3.70it/s]\u001b[A\n",
            " 46%|████▌     | 146/318 [00:36<00:46,  3.67it/s]\u001b[A\n",
            " 46%|████▌     | 147/318 [00:36<00:48,  3.55it/s]\u001b[A\n",
            " 47%|████▋     | 148/318 [00:36<00:46,  3.69it/s]\u001b[A\n",
            " 47%|████▋     | 149/318 [00:37<00:43,  3.87it/s]\u001b[A\n",
            " 47%|████▋     | 150/318 [00:37<00:42,  3.96it/s]\u001b[A\n",
            " 47%|████▋     | 151/318 [00:37<00:42,  3.97it/s]\u001b[A\n",
            " 48%|████▊     | 152/318 [00:37<00:42,  3.93it/s]\u001b[A\n",
            " 48%|████▊     | 153/318 [00:37<00:38,  4.24it/s]\u001b[A\n",
            " 48%|████▊     | 154/318 [00:38<00:38,  4.29it/s]\u001b[A\n",
            " 49%|████▊     | 155/318 [00:38<00:40,  4.06it/s]\u001b[A\n",
            " 49%|████▉     | 156/318 [00:38<00:40,  3.99it/s]\u001b[A\n",
            " 49%|████▉     | 157/318 [00:38<00:41,  3.84it/s]\u001b[A\n",
            " 50%|████▉     | 158/318 [00:39<00:41,  3.87it/s]\u001b[A\n",
            " 50%|█████     | 159/318 [00:39<00:38,  4.15it/s]\u001b[A\n",
            " 50%|█████     | 160/318 [00:39<00:41,  3.77it/s]\u001b[A\n",
            " 51%|█████     | 161/318 [00:39<00:38,  4.08it/s]\u001b[A\n",
            " 51%|█████     | 162/318 [00:40<00:38,  4.07it/s]\u001b[A\n",
            " 51%|█████▏    | 163/318 [00:40<00:38,  4.00it/s]\u001b[A\n",
            " 52%|█████▏    | 164/318 [00:40<00:35,  4.35it/s]\u001b[A\n",
            " 52%|█████▏    | 165/318 [00:40<00:34,  4.41it/s]\u001b[A\n",
            " 52%|█████▏    | 166/318 [00:41<00:36,  4.21it/s]\u001b[A\n",
            " 53%|█████▎    | 167/318 [00:41<00:34,  4.36it/s]\u001b[A\n",
            " 53%|█████▎    | 168/318 [00:41<00:38,  3.94it/s]\u001b[A\n",
            " 53%|█████▎    | 169/318 [00:41<00:36,  4.12it/s]\u001b[A\n",
            " 53%|█████▎    | 170/318 [00:42<00:36,  4.01it/s]\u001b[A\n",
            " 54%|█████▍    | 171/318 [00:42<00:37,  3.93it/s]\u001b[A\n",
            " 54%|█████▍    | 172/318 [00:42<00:35,  4.07it/s]\u001b[A\n",
            " 54%|█████▍    | 173/318 [00:42<00:34,  4.15it/s]\u001b[A\n",
            " 55%|█████▍    | 174/318 [00:43<00:35,  4.02it/s]\u001b[A\n",
            " 55%|█████▌    | 175/318 [00:43<00:34,  4.15it/s]\u001b[A\n",
            " 55%|█████▌    | 176/318 [00:43<00:33,  4.25it/s]\u001b[A\n",
            " 56%|█████▌    | 177/318 [00:43<00:33,  4.24it/s]\u001b[A\n",
            " 56%|█████▌    | 178/318 [00:44<00:32,  4.25it/s]\u001b[A\n",
            " 56%|█████▋    | 179/318 [00:44<00:32,  4.27it/s]\u001b[A\n",
            " 57%|█████▋    | 180/318 [00:44<00:32,  4.20it/s]\u001b[A\n",
            " 57%|█████▋    | 181/318 [00:44<00:32,  4.21it/s]\u001b[A\n",
            " 57%|█████▋    | 182/318 [00:44<00:31,  4.29it/s]\u001b[A\n",
            " 58%|█████▊    | 183/318 [00:45<00:31,  4.25it/s]\u001b[A\n",
            " 58%|█████▊    | 184/318 [00:45<00:33,  4.03it/s]\u001b[A\n",
            " 58%|█████▊    | 185/318 [00:45<00:31,  4.17it/s]\u001b[A\n",
            " 58%|█████▊    | 186/318 [00:45<00:32,  4.05it/s]\u001b[A\n",
            " 59%|█████▉    | 187/318 [00:46<00:33,  3.86it/s]\u001b[A\n",
            " 59%|█████▉    | 188/318 [00:46<00:33,  3.83it/s]\u001b[A\n",
            " 59%|█████▉    | 189/318 [00:46<00:32,  3.92it/s]\u001b[A\n",
            " 60%|█████▉    | 190/318 [00:47<00:33,  3.82it/s]\u001b[A\n",
            " 60%|██████    | 191/318 [00:47<00:33,  3.78it/s]\u001b[A\n",
            " 60%|██████    | 192/318 [00:47<00:32,  3.87it/s]\u001b[A\n",
            " 61%|██████    | 193/318 [00:47<00:33,  3.75it/s]\u001b[A\n",
            " 61%|██████    | 194/318 [00:48<00:32,  3.79it/s]\u001b[A\n",
            " 61%|██████▏   | 195/318 [00:48<00:31,  3.86it/s]\u001b[A\n",
            " 62%|██████▏   | 196/318 [00:48<00:36,  3.34it/s]\u001b[A\n",
            " 62%|██████▏   | 197/318 [00:49<00:34,  3.51it/s]\u001b[A\n",
            " 62%|██████▏   | 198/318 [00:49<00:32,  3.69it/s]\u001b[A\n",
            " 63%|██████▎   | 199/318 [00:49<00:31,  3.73it/s]\u001b[A\n",
            " 63%|██████▎   | 200/318 [00:49<00:31,  3.79it/s]\u001b[A\n",
            " 63%|██████▎   | 201/318 [00:49<00:29,  3.95it/s]\u001b[A\n",
            " 64%|██████▎   | 202/318 [00:50<00:29,  3.90it/s]\u001b[A\n",
            " 64%|██████▍   | 203/318 [00:50<00:30,  3.81it/s]\u001b[A\n",
            " 64%|██████▍   | 204/318 [00:50<00:29,  3.86it/s]\u001b[A\n",
            " 64%|██████▍   | 205/318 [00:51<00:30,  3.69it/s]\u001b[A\n",
            " 65%|██████▍   | 206/318 [00:51<00:29,  3.84it/s]\u001b[A\n",
            " 65%|██████▌   | 207/318 [00:51<00:28,  3.94it/s]\u001b[A\n",
            " 65%|██████▌   | 208/318 [00:51<00:27,  3.99it/s]\u001b[A\n",
            " 66%|██████▌   | 209/318 [00:52<00:27,  4.03it/s]\u001b[A\n",
            " 66%|██████▌   | 210/318 [00:52<00:25,  4.20it/s]\u001b[A\n",
            " 66%|██████▋   | 211/318 [00:52<00:25,  4.12it/s]\u001b[A\n",
            " 67%|██████▋   | 212/318 [00:52<00:25,  4.13it/s]\u001b[A\n",
            " 67%|██████▋   | 213/318 [00:53<00:26,  3.99it/s]\u001b[A\n",
            " 67%|██████▋   | 214/318 [00:53<00:25,  4.13it/s]\u001b[A\n",
            " 68%|██████▊   | 215/318 [00:53<00:26,  3.95it/s]\u001b[A\n",
            " 68%|██████▊   | 216/318 [00:53<00:26,  3.87it/s]\u001b[A\n",
            " 68%|██████▊   | 217/318 [00:54<00:26,  3.74it/s]\u001b[A\n",
            " 69%|██████▊   | 218/318 [00:54<00:24,  4.03it/s]\u001b[A\n",
            " 69%|██████▉   | 219/318 [00:54<00:23,  4.18it/s]\u001b[A\n",
            " 69%|██████▉   | 220/318 [00:54<00:22,  4.39it/s]\u001b[A\n",
            " 69%|██████▉   | 221/318 [00:54<00:20,  4.69it/s]\u001b[A\n",
            " 70%|██████▉   | 222/318 [00:55<00:22,  4.19it/s]\u001b[A\n",
            " 70%|███████   | 223/318 [00:55<00:21,  4.35it/s]\u001b[A\n",
            " 70%|███████   | 224/318 [00:55<00:23,  4.03it/s]\u001b[A\n",
            " 71%|███████   | 225/318 [00:55<00:23,  4.01it/s]\u001b[A\n",
            " 71%|███████   | 226/318 [00:56<00:21,  4.26it/s]\u001b[A\n",
            " 71%|███████▏  | 227/318 [00:56<00:20,  4.34it/s]\u001b[A\n",
            " 72%|███████▏  | 228/318 [00:56<00:20,  4.46it/s]\u001b[A\n",
            " 72%|███████▏  | 229/318 [00:56<00:20,  4.33it/s]\u001b[A\n",
            " 72%|███████▏  | 230/318 [00:57<00:19,  4.53it/s]\u001b[A\n",
            " 73%|███████▎  | 231/318 [00:57<00:18,  4.69it/s]\u001b[A\n",
            " 73%|███████▎  | 232/318 [00:57<00:20,  4.25it/s]\u001b[A\n",
            " 73%|███████▎  | 233/318 [00:57<00:19,  4.26it/s]\u001b[A\n",
            " 74%|███████▎  | 234/318 [00:57<00:19,  4.34it/s]\u001b[A\n",
            " 74%|███████▍  | 235/318 [00:58<00:20,  4.13it/s]\u001b[A\n",
            " 74%|███████▍  | 236/318 [00:58<00:19,  4.21it/s]\u001b[A\n",
            " 75%|███████▍  | 237/318 [00:58<00:19,  4.14it/s]\u001b[A\n",
            " 75%|███████▍  | 238/318 [00:58<00:20,  3.88it/s]\u001b[A\n",
            " 75%|███████▌  | 239/318 [00:59<00:20,  3.82it/s]\u001b[A\n",
            " 75%|███████▌  | 240/318 [00:59<00:21,  3.66it/s]\u001b[A\n",
            " 76%|███████▌  | 241/318 [00:59<00:19,  3.92it/s]\u001b[A\n",
            " 76%|███████▌  | 242/318 [00:59<00:18,  4.13it/s]\u001b[A\n",
            " 76%|███████▋  | 243/318 [01:00<00:17,  4.18it/s]\u001b[A\n",
            " 77%|███████▋  | 244/318 [01:00<00:18,  3.90it/s]\u001b[A\n",
            " 77%|███████▋  | 245/318 [01:00<00:19,  3.74it/s]\u001b[A\n",
            " 77%|███████▋  | 246/318 [01:01<00:18,  3.82it/s]\u001b[A\n",
            " 78%|███████▊  | 247/318 [01:01<00:18,  3.93it/s]\u001b[A\n",
            " 78%|███████▊  | 248/318 [01:01<00:17,  3.96it/s]\u001b[A\n",
            " 78%|███████▊  | 249/318 [01:01<00:18,  3.77it/s]\u001b[A\n",
            " 79%|███████▊  | 250/318 [01:02<00:18,  3.65it/s]\u001b[A\n",
            " 79%|███████▉  | 251/318 [01:02<00:17,  3.84it/s]\u001b[A\n",
            " 79%|███████▉  | 252/318 [01:02<00:17,  3.74it/s]\u001b[A\n",
            " 80%|███████▉  | 253/318 [01:02<00:17,  3.77it/s]\u001b[A\n",
            " 80%|███████▉  | 254/318 [01:03<00:15,  4.04it/s]\u001b[A\n",
            " 80%|████████  | 255/318 [01:03<00:15,  3.94it/s]\u001b[A\n",
            " 81%|████████  | 256/318 [01:03<00:14,  4.22it/s]\u001b[A\n",
            " 81%|████████  | 257/318 [01:03<00:14,  4.29it/s]\u001b[A\n",
            " 81%|████████  | 258/318 [01:04<00:13,  4.34it/s]\u001b[A\n",
            " 81%|████████▏ | 259/318 [01:04<00:14,  4.21it/s]\u001b[A\n",
            " 82%|████████▏ | 260/318 [01:04<00:13,  4.19it/s]\u001b[A\n",
            " 82%|████████▏ | 261/318 [01:04<00:13,  4.21it/s]\u001b[A\n",
            " 82%|████████▏ | 262/318 [01:04<00:13,  4.26it/s]\u001b[A\n",
            " 83%|████████▎ | 263/318 [01:05<00:13,  3.95it/s]\u001b[A\n",
            " 83%|████████▎ | 264/318 [01:05<00:13,  4.08it/s]\u001b[A\n",
            " 83%|████████▎ | 265/318 [01:05<00:13,  4.02it/s]\u001b[A\n",
            " 84%|████████▎ | 266/318 [01:06<00:13,  3.99it/s]\u001b[A\n",
            " 84%|████████▍ | 267/318 [01:06<00:13,  3.86it/s]\u001b[A\n",
            " 84%|████████▍ | 268/318 [01:06<00:12,  4.08it/s]\u001b[A\n",
            " 85%|████████▍ | 269/318 [01:06<00:11,  4.21it/s]\u001b[A\n",
            " 85%|████████▍ | 270/318 [01:06<00:11,  4.20it/s]\u001b[A\n",
            " 85%|████████▌ | 271/318 [01:07<00:11,  4.21it/s]\u001b[A\n",
            " 86%|████████▌ | 272/318 [01:07<00:11,  3.97it/s]\u001b[A\n",
            " 86%|████████▌ | 273/318 [01:07<00:11,  3.99it/s]\u001b[A\n",
            " 86%|████████▌ | 274/318 [01:07<00:10,  4.14it/s]\u001b[A\n",
            " 86%|████████▋ | 275/318 [01:08<00:10,  4.15it/s]\u001b[A\n",
            " 87%|████████▋ | 276/318 [01:08<00:09,  4.43it/s]\u001b[A\n",
            " 87%|████████▋ | 277/318 [01:08<00:08,  4.72it/s]\u001b[A\n",
            " 87%|████████▋ | 278/318 [01:08<00:08,  4.56it/s]\u001b[A\n",
            " 88%|████████▊ | 279/318 [01:09<00:08,  4.46it/s]\u001b[A\n",
            " 88%|████████▊ | 280/318 [01:09<00:08,  4.55it/s]\u001b[A\n",
            " 88%|████████▊ | 281/318 [01:09<00:08,  4.55it/s]\u001b[A\n",
            " 89%|████████▊ | 282/318 [01:09<00:08,  4.28it/s]\u001b[A\n",
            " 89%|████████▉ | 283/318 [01:09<00:08,  4.22it/s]\u001b[A\n",
            " 89%|████████▉ | 284/318 [01:10<00:08,  4.09it/s]\u001b[A\n",
            " 90%|████████▉ | 285/318 [01:10<00:07,  4.33it/s]\u001b[A\n",
            " 90%|████████▉ | 286/318 [01:10<00:07,  4.15it/s]\u001b[A\n",
            " 90%|█████████ | 287/318 [01:10<00:07,  4.09it/s]\u001b[A\n",
            " 91%|█████████ | 288/318 [01:11<00:07,  4.16it/s]\u001b[A\n",
            " 91%|█████████ | 289/318 [01:11<00:06,  4.19it/s]\u001b[A\n",
            " 91%|█████████ | 290/318 [01:11<00:06,  4.04it/s]\u001b[A\n",
            " 92%|█████████▏| 291/318 [01:11<00:06,  3.93it/s]\u001b[A\n",
            " 92%|█████████▏| 292/318 [01:12<00:07,  3.70it/s]\u001b[A\n",
            " 92%|█████████▏| 293/318 [01:12<00:06,  3.70it/s]\u001b[A\n",
            " 92%|█████████▏| 294/318 [01:12<00:06,  3.44it/s]\u001b[A\n",
            " 93%|█████████▎| 295/318 [01:13<00:06,  3.67it/s]\u001b[A\n",
            " 93%|█████████▎| 296/318 [01:13<00:06,  3.62it/s]\u001b[A\n",
            " 93%|█████████▎| 297/318 [01:13<00:06,  3.50it/s]\u001b[A\n",
            " 94%|█████████▎| 298/318 [01:13<00:05,  3.52it/s]\u001b[A\n",
            " 94%|█████████▍| 299/318 [01:14<00:05,  3.68it/s]\u001b[A\n",
            " 94%|█████████▍| 300/318 [01:14<00:04,  3.92it/s]\u001b[A\n",
            " 95%|█████████▍| 301/318 [01:14<00:04,  3.97it/s]\u001b[A\n",
            " 95%|█████████▍| 302/318 [01:14<00:03,  4.05it/s]\u001b[A\n",
            " 95%|█████████▌| 303/318 [01:15<00:03,  4.16it/s]\u001b[A\n",
            " 96%|█████████▌| 304/318 [01:15<00:03,  4.03it/s]\u001b[A\n",
            " 96%|█████████▌| 305/318 [01:15<00:03,  3.91it/s]\u001b[A\n",
            " 96%|█████████▌| 306/318 [01:15<00:02,  4.21it/s]\u001b[A\n",
            " 97%|█████████▋| 307/318 [01:16<00:02,  4.02it/s]\u001b[A\n",
            " 97%|█████████▋| 308/318 [01:16<00:02,  4.17it/s]\u001b[A\n",
            " 97%|█████████▋| 309/318 [01:16<00:02,  4.39it/s]\u001b[A\n",
            " 97%|█████████▋| 310/318 [01:16<00:01,  4.25it/s]\u001b[A\n",
            " 98%|█████████▊| 311/318 [01:17<00:01,  3.69it/s]\u001b[A\n",
            " 98%|█████████▊| 312/318 [01:17<00:01,  3.85it/s]\u001b[A\n",
            " 98%|█████████▊| 313/318 [01:17<00:01,  4.05it/s]\u001b[A\n",
            " 99%|█████████▊| 314/318 [01:17<00:01,  3.98it/s]\u001b[A\n",
            " 99%|█████████▉| 315/318 [01:18<00:00,  4.22it/s]\u001b[A\n",
            " 99%|█████████▉| 316/318 [01:18<00:00,  4.07it/s]\u001b[A\n",
            "100%|█████████▉| 317/318 [01:18<00:00,  4.14it/s]\u001b[A\n",
            "100%|██████████| 318/318 [01:18<00:00,  4.03it/s]\n",
            " 40%|████      | 4/10 [05:20<07:56, 79.50s/it]\n",
            "  0%|          | 0/318 [00:00<?, ?it/s]\u001b[A\n",
            "  0%|          | 1/318 [00:00<02:02,  2.59it/s]\u001b[A\n",
            "  1%|          | 2/318 [00:00<01:47,  2.95it/s]\u001b[A\n",
            "  1%|          | 3/318 [00:00<01:27,  3.61it/s]\u001b[A\n",
            "  1%|▏         | 4/318 [00:01<01:23,  3.77it/s]\u001b[A\n",
            "  2%|▏         | 5/318 [00:01<01:13,  4.23it/s]\u001b[A\n",
            "  2%|▏         | 6/318 [00:01<01:12,  4.31it/s]\u001b[A\n",
            "  2%|▏         | 7/318 [00:01<01:16,  4.04it/s]\u001b[A\n",
            "  3%|▎         | 8/318 [00:02<01:14,  4.14it/s]\u001b[A\n",
            "  3%|▎         | 9/318 [00:02<01:13,  4.18it/s]\u001b[A\n",
            "  3%|▎         | 10/318 [00:02<01:08,  4.50it/s]\u001b[A\n",
            "  3%|▎         | 11/318 [00:02<01:12,  4.22it/s]\u001b[A\n",
            "  4%|▍         | 12/318 [00:03<01:16,  3.98it/s]\u001b[A\n",
            "  4%|▍         | 13/318 [00:03<01:13,  4.13it/s]\u001b[A\n",
            "  4%|▍         | 14/318 [00:03<01:20,  3.76it/s]\u001b[A\n",
            "  5%|▍         | 15/318 [00:03<01:21,  3.74it/s]\u001b[A\n",
            "  5%|▌         | 16/318 [00:04<01:14,  4.08it/s]\u001b[A\n",
            "  5%|▌         | 17/318 [00:04<01:13,  4.09it/s]\u001b[A\n",
            "  6%|▌         | 18/318 [00:04<01:12,  4.15it/s]\u001b[A\n",
            "  6%|▌         | 19/318 [00:04<01:08,  4.34it/s]\u001b[A\n",
            "  6%|▋         | 20/318 [00:04<01:08,  4.32it/s]\u001b[A\n",
            "  7%|▋         | 21/318 [00:05<01:11,  4.18it/s]\u001b[A\n",
            "  7%|▋         | 22/318 [00:05<01:11,  4.17it/s]\u001b[A\n",
            "  7%|▋         | 23/318 [00:05<01:16,  3.87it/s]\u001b[A\n",
            "  8%|▊         | 24/318 [00:06<01:26,  3.40it/s]\u001b[A\n",
            "  8%|▊         | 25/318 [00:06<01:23,  3.52it/s]\u001b[A\n",
            "  8%|▊         | 26/318 [00:06<01:17,  3.75it/s]\u001b[A\n",
            "  8%|▊         | 27/318 [00:06<01:21,  3.55it/s]\u001b[A\n",
            "  9%|▉         | 28/318 [00:07<01:17,  3.73it/s]\u001b[A\n",
            "  9%|▉         | 29/318 [00:07<01:17,  3.74it/s]\u001b[A\n",
            "  9%|▉         | 30/318 [00:07<01:17,  3.71it/s]\u001b[A\n",
            " 10%|▉         | 31/318 [00:07<01:17,  3.70it/s]\u001b[A\n",
            " 10%|█         | 32/318 [00:08<01:12,  3.92it/s]\u001b[A\n",
            " 10%|█         | 33/318 [00:08<01:16,  3.74it/s]\u001b[A\n",
            " 11%|█         | 34/318 [00:08<01:13,  3.85it/s]\u001b[A\n",
            " 11%|█         | 35/318 [00:08<01:11,  3.94it/s]\u001b[A\n",
            " 11%|█▏        | 36/318 [00:09<01:09,  4.08it/s]\u001b[A\n",
            " 12%|█▏        | 37/318 [00:09<01:05,  4.29it/s]\u001b[A\n",
            " 12%|█▏        | 38/318 [00:09<01:06,  4.22it/s]\u001b[A\n",
            " 12%|█▏        | 39/318 [00:09<01:07,  4.14it/s]\u001b[A\n",
            " 13%|█▎        | 40/318 [00:10<01:03,  4.36it/s]\u001b[A\n",
            " 13%|█▎        | 41/318 [00:10<01:06,  4.17it/s]\u001b[A\n",
            " 13%|█▎        | 42/318 [00:10<01:02,  4.43it/s]\u001b[A\n",
            " 14%|█▎        | 43/318 [00:10<01:06,  4.12it/s]\u001b[A\n",
            " 14%|█▍        | 44/318 [00:11<01:04,  4.23it/s]\u001b[A\n",
            " 14%|█▍        | 45/318 [00:11<01:01,  4.40it/s]\u001b[A\n",
            " 14%|█▍        | 46/318 [00:11<01:03,  4.28it/s]\u001b[A\n",
            " 15%|█▍        | 47/318 [00:11<01:02,  4.33it/s]\u001b[A\n",
            " 15%|█▌        | 48/318 [00:11<01:00,  4.45it/s]\u001b[A\n",
            " 15%|█▌        | 49/318 [00:12<01:00,  4.45it/s]\u001b[A\n",
            " 16%|█▌        | 50/318 [00:12<00:58,  4.55it/s]\u001b[A\n",
            " 16%|█▌        | 51/318 [00:12<00:59,  4.48it/s]\u001b[A\n",
            " 16%|█▋        | 52/318 [00:12<01:02,  4.26it/s]\u001b[A\n",
            " 17%|█▋        | 53/318 [00:13<01:02,  4.25it/s]\u001b[A\n",
            " 17%|█▋        | 54/318 [00:13<01:00,  4.35it/s]\u001b[A\n",
            " 17%|█▋        | 55/318 [00:13<00:58,  4.46it/s]\u001b[A\n",
            " 18%|█▊        | 56/318 [00:13<00:58,  4.52it/s]\u001b[A\n",
            " 18%|█▊        | 57/318 [00:14<01:02,  4.16it/s]\u001b[A\n",
            " 18%|█▊        | 58/318 [00:14<01:00,  4.27it/s]\u001b[A\n",
            " 19%|█▊        | 59/318 [00:14<01:01,  4.19it/s]\u001b[A\n",
            " 19%|█▉        | 60/318 [00:14<01:03,  4.05it/s]\u001b[A\n",
            " 19%|█▉        | 61/318 [00:15<01:01,  4.18it/s]\u001b[A\n",
            " 19%|█▉        | 62/318 [00:15<01:06,  3.84it/s]\u001b[A\n",
            " 20%|█▉        | 63/318 [00:15<01:10,  3.64it/s]\u001b[A\n",
            " 20%|██        | 64/318 [00:15<01:15,  3.38it/s]\u001b[A\n",
            " 20%|██        | 65/318 [00:16<01:08,  3.72it/s]\u001b[A\n",
            " 21%|██        | 66/318 [00:16<01:00,  4.14it/s]\u001b[A\n",
            " 21%|██        | 67/318 [00:16<01:01,  4.10it/s]\u001b[A\n",
            " 21%|██▏       | 68/318 [00:16<01:04,  3.89it/s]\u001b[A\n",
            " 22%|██▏       | 69/318 [00:17<01:00,  4.14it/s]\u001b[A\n",
            " 22%|██▏       | 70/318 [00:17<01:04,  3.86it/s]\u001b[A\n",
            " 22%|██▏       | 71/318 [00:17<00:58,  4.19it/s]\u001b[A\n",
            " 23%|██▎       | 72/318 [00:17<00:57,  4.28it/s]\u001b[A\n",
            " 23%|██▎       | 73/318 [00:18<00:56,  4.30it/s]\u001b[A\n",
            " 23%|██▎       | 74/318 [00:18<00:58,  4.18it/s]\u001b[A\n",
            " 24%|██▎       | 75/318 [00:18<01:03,  3.83it/s]\u001b[A\n",
            " 24%|██▍       | 76/318 [00:18<01:05,  3.71it/s]\u001b[A\n",
            " 24%|██▍       | 77/318 [00:19<01:05,  3.69it/s]\u001b[A\n",
            " 25%|██▍       | 78/318 [00:19<01:03,  3.76it/s]\u001b[A\n",
            " 25%|██▍       | 79/318 [00:19<00:57,  4.14it/s]\u001b[A\n",
            " 25%|██▌       | 80/318 [00:19<00:57,  4.12it/s]\u001b[A\n",
            " 25%|██▌       | 81/318 [00:20<00:59,  3.96it/s]\u001b[A\n",
            " 26%|██▌       | 82/318 [00:20<00:58,  4.06it/s]\u001b[A\n",
            " 26%|██▌       | 83/318 [00:20<01:01,  3.83it/s]\u001b[A\n",
            " 26%|██▋       | 84/318 [00:20<00:59,  3.92it/s]\u001b[A\n",
            " 27%|██▋       | 85/318 [00:21<00:58,  3.99it/s]\u001b[A\n",
            " 27%|██▋       | 86/318 [00:21<00:57,  4.06it/s]\u001b[A\n",
            " 27%|██▋       | 87/318 [00:21<00:56,  4.07it/s]\u001b[A\n",
            " 28%|██▊       | 88/318 [00:21<00:59,  3.87it/s]\u001b[A\n",
            " 28%|██▊       | 89/318 [00:22<01:00,  3.79it/s]\u001b[A\n",
            " 28%|██▊       | 90/318 [00:22<01:00,  3.79it/s]\u001b[A\n",
            " 29%|██▊       | 91/318 [00:22<00:55,  4.06it/s]\u001b[A\n",
            " 29%|██▉       | 92/318 [00:22<00:56,  3.97it/s]\u001b[A\n",
            " 29%|██▉       | 93/318 [00:23<00:54,  4.14it/s]\u001b[A\n",
            " 30%|██▉       | 94/318 [00:23<00:53,  4.22it/s]\u001b[A\n",
            " 30%|██▉       | 95/318 [00:23<00:51,  4.32it/s]\u001b[A\n",
            " 30%|███       | 96/318 [00:23<00:53,  4.18it/s]\u001b[A\n",
            " 31%|███       | 97/318 [00:24<00:52,  4.18it/s]\u001b[A\n",
            " 31%|███       | 98/318 [00:24<00:53,  4.14it/s]\u001b[A\n",
            " 31%|███       | 99/318 [00:24<00:53,  4.07it/s]\u001b[A\n",
            " 31%|███▏      | 100/318 [00:24<00:52,  4.17it/s]\u001b[A\n",
            " 32%|███▏      | 101/318 [00:25<00:52,  4.14it/s]\u001b[A\n",
            " 32%|███▏      | 102/318 [00:25<00:52,  4.10it/s]\u001b[A\n",
            " 32%|███▏      | 103/318 [00:25<00:50,  4.28it/s]\u001b[A\n",
            " 33%|███▎      | 104/318 [00:25<00:54,  3.96it/s]\u001b[A\n",
            " 33%|███▎      | 105/318 [00:26<00:50,  4.23it/s]\u001b[A\n",
            " 33%|███▎      | 106/318 [00:26<00:50,  4.23it/s]\u001b[A\n",
            " 34%|███▎      | 107/318 [00:26<00:54,  3.85it/s]\u001b[A\n",
            " 34%|███▍      | 108/318 [00:26<00:54,  3.89it/s]\u001b[A\n",
            " 34%|███▍      | 109/318 [00:27<00:52,  3.98it/s]\u001b[A\n",
            " 35%|███▍      | 110/318 [00:27<00:51,  4.04it/s]\u001b[A\n",
            " 35%|███▍      | 111/318 [00:27<00:49,  4.16it/s]\u001b[A\n",
            " 35%|███▌      | 112/318 [00:27<00:48,  4.27it/s]\u001b[A\n",
            " 36%|███▌      | 113/318 [00:28<00:54,  3.74it/s]\u001b[A\n",
            " 36%|███▌      | 114/318 [00:28<00:50,  4.05it/s]\u001b[A\n",
            " 36%|███▌      | 115/318 [00:28<00:50,  4.03it/s]\u001b[A\n",
            " 36%|███▋      | 116/318 [00:28<00:50,  3.97it/s]\u001b[A\n",
            " 37%|███▋      | 117/318 [00:29<00:50,  3.95it/s]\u001b[A\n",
            " 37%|███▋      | 118/318 [00:29<00:52,  3.79it/s]\u001b[A\n",
            " 37%|███▋      | 119/318 [00:29<00:52,  3.80it/s]\u001b[A\n",
            " 38%|███▊      | 120/318 [00:29<00:53,  3.68it/s]\u001b[A\n",
            " 38%|███▊      | 121/318 [00:30<00:52,  3.74it/s]\u001b[A\n",
            " 38%|███▊      | 122/318 [00:30<00:51,  3.79it/s]\u001b[A\n",
            " 39%|███▊      | 123/318 [00:30<00:52,  3.71it/s]\u001b[A\n",
            " 39%|███▉      | 124/318 [00:30<00:50,  3.84it/s]\u001b[A\n",
            " 39%|███▉      | 125/318 [00:31<00:48,  3.95it/s]\u001b[A\n",
            " 40%|███▉      | 126/318 [00:31<00:48,  3.98it/s]\u001b[A\n",
            " 40%|███▉      | 127/318 [00:31<00:47,  4.05it/s]\u001b[A\n",
            " 40%|████      | 128/318 [00:31<00:46,  4.08it/s]\u001b[A\n",
            " 41%|████      | 129/318 [00:32<00:45,  4.12it/s]\u001b[A\n",
            " 41%|████      | 130/318 [00:32<00:50,  3.75it/s]\u001b[A\n",
            " 41%|████      | 131/318 [00:32<00:48,  3.84it/s]\u001b[A\n",
            " 42%|████▏     | 132/318 [00:32<00:47,  3.93it/s]\u001b[A\n",
            " 42%|████▏     | 133/318 [00:33<00:46,  3.99it/s]\u001b[A\n",
            " 42%|████▏     | 134/318 [00:33<00:43,  4.23it/s]\u001b[A\n",
            " 42%|████▏     | 135/318 [00:33<00:45,  4.01it/s]\u001b[A\n",
            " 43%|████▎     | 136/318 [00:33<00:45,  4.03it/s]\u001b[A\n",
            " 43%|████▎     | 137/318 [00:34<00:44,  4.10it/s]\u001b[A\n",
            " 43%|████▎     | 138/318 [00:34<00:46,  3.91it/s]\u001b[A\n",
            " 44%|████▎     | 139/318 [00:34<00:45,  3.95it/s]\u001b[A\n",
            " 44%|████▍     | 140/318 [00:34<00:47,  3.73it/s]\u001b[A\n",
            " 44%|████▍     | 141/318 [00:35<00:47,  3.73it/s]\u001b[A\n",
            " 45%|████▍     | 142/318 [00:35<00:45,  3.88it/s]\u001b[A\n",
            " 45%|████▍     | 143/318 [00:35<00:45,  3.82it/s]\u001b[A\n",
            " 45%|████▌     | 144/318 [00:35<00:45,  3.85it/s]\u001b[A\n",
            " 46%|████▌     | 145/318 [00:36<00:43,  3.96it/s]\u001b[A\n",
            " 46%|████▌     | 146/318 [00:36<00:44,  3.83it/s]\u001b[A\n",
            " 46%|████▌     | 147/318 [00:36<00:42,  4.00it/s]\u001b[A\n",
            " 47%|████▋     | 148/318 [00:36<00:40,  4.15it/s]\u001b[A\n",
            " 47%|████▋     | 149/318 [00:37<00:41,  4.11it/s]\u001b[A\n",
            " 47%|████▋     | 150/318 [00:37<00:39,  4.21it/s]\u001b[A\n",
            " 47%|████▋     | 151/318 [00:37<00:38,  4.31it/s]\u001b[A\n",
            " 48%|████▊     | 152/318 [00:37<00:38,  4.28it/s]\u001b[A\n",
            " 48%|████▊     | 153/318 [00:38<00:41,  4.01it/s]\u001b[A\n",
            " 48%|████▊     | 154/318 [00:38<00:43,  3.80it/s]\u001b[A\n",
            " 49%|████▊     | 155/318 [00:38<00:40,  4.06it/s]\u001b[A\n",
            " 49%|████▉     | 156/318 [00:38<00:38,  4.21it/s]\u001b[A\n",
            " 49%|████▉     | 157/318 [00:39<00:38,  4.23it/s]\u001b[A\n",
            " 50%|████▉     | 158/318 [00:39<00:36,  4.37it/s]\u001b[A\n",
            " 50%|█████     | 159/318 [00:39<00:38,  4.18it/s]\u001b[A\n",
            " 50%|█████     | 160/318 [00:39<00:42,  3.70it/s]\u001b[A\n",
            " 51%|█████     | 161/318 [00:40<00:39,  3.98it/s]\u001b[A\n",
            " 51%|█████     | 162/318 [00:40<00:41,  3.80it/s]\u001b[A\n",
            " 51%|█████▏    | 163/318 [00:40<00:38,  4.03it/s]\u001b[A\n",
            " 52%|█████▏    | 164/318 [00:40<00:37,  4.06it/s]\u001b[A\n",
            " 52%|█████▏    | 165/318 [00:41<00:37,  4.13it/s]\u001b[A\n",
            " 52%|█████▏    | 166/318 [00:41<00:38,  3.94it/s]\u001b[A\n",
            " 53%|█████▎    | 167/318 [00:41<00:37,  4.06it/s]\u001b[A\n",
            " 53%|█████▎    | 168/318 [00:41<00:38,  3.94it/s]\u001b[A\n",
            " 53%|█████▎    | 169/318 [00:42<00:36,  4.13it/s]\u001b[A\n",
            " 53%|█████▎    | 170/318 [00:42<00:35,  4.23it/s]\u001b[A\n",
            " 54%|█████▍    | 171/318 [00:42<00:36,  4.04it/s]\u001b[A\n",
            " 54%|█████▍    | 172/318 [00:42<00:34,  4.28it/s]\u001b[A\n",
            " 54%|█████▍    | 173/318 [00:43<00:35,  4.09it/s]\u001b[A\n",
            " 55%|█████▍    | 174/318 [00:43<00:37,  3.88it/s]\u001b[A\n",
            " 55%|█████▌    | 175/318 [00:43<00:34,  4.09it/s]\u001b[A\n",
            " 55%|█████▌    | 176/318 [00:43<00:34,  4.15it/s]\u001b[A\n",
            " 56%|█████▌    | 177/318 [00:44<00:33,  4.27it/s]\u001b[A\n",
            " 56%|█████▌    | 178/318 [00:44<00:33,  4.20it/s]\u001b[A\n",
            " 56%|█████▋    | 179/318 [00:44<00:34,  4.06it/s]\u001b[A\n",
            " 57%|█████▋    | 180/318 [00:44<00:33,  4.14it/s]\u001b[A\n",
            " 57%|█████▋    | 181/318 [00:45<00:33,  4.04it/s]\u001b[A\n",
            " 57%|█████▋    | 182/318 [00:45<00:32,  4.15it/s]\u001b[A\n",
            " 58%|█████▊    | 183/318 [00:45<00:34,  3.88it/s]\u001b[A\n",
            " 58%|█████▊    | 184/318 [00:45<00:32,  4.13it/s]\u001b[A\n",
            " 58%|█████▊    | 185/318 [00:46<00:31,  4.18it/s]\u001b[A\n",
            " 58%|█████▊    | 186/318 [00:46<00:31,  4.15it/s]\u001b[A\n",
            " 59%|█████▉    | 187/318 [00:46<00:32,  4.02it/s]\u001b[A\n",
            " 59%|█████▉    | 188/318 [00:46<00:34,  3.81it/s]\u001b[A\n",
            " 59%|█████▉    | 189/318 [00:47<00:32,  3.92it/s]\u001b[A\n",
            " 60%|█████▉    | 190/318 [00:47<00:32,  3.92it/s]\u001b[A\n",
            " 60%|██████    | 191/318 [00:47<00:31,  4.05it/s]\u001b[A\n",
            " 60%|██████    | 192/318 [00:47<00:32,  3.86it/s]\u001b[A\n",
            " 61%|██████    | 193/318 [00:48<00:31,  3.99it/s]\u001b[A\n",
            " 61%|██████    | 194/318 [00:48<00:30,  4.04it/s]\u001b[A\n",
            " 61%|██████▏   | 195/318 [00:48<00:30,  4.09it/s]\u001b[A\n",
            " 62%|██████▏   | 196/318 [00:48<00:29,  4.10it/s]\u001b[A\n",
            " 62%|██████▏   | 197/318 [00:49<00:29,  4.07it/s]\u001b[A\n",
            " 62%|██████▏   | 198/318 [00:49<00:28,  4.22it/s]\u001b[A\n",
            " 63%|██████▎   | 199/318 [00:49<00:28,  4.20it/s]\u001b[A\n",
            " 63%|██████▎   | 200/318 [00:49<00:29,  3.96it/s]\u001b[A\n",
            " 63%|██████▎   | 201/318 [00:50<00:29,  3.95it/s]\u001b[A\n",
            " 64%|██████▎   | 202/318 [00:50<00:27,  4.18it/s]\u001b[A\n",
            " 64%|██████▍   | 203/318 [00:50<00:27,  4.11it/s]\u001b[A\n",
            " 64%|██████▍   | 204/318 [00:50<00:26,  4.23it/s]\u001b[A\n",
            " 64%|██████▍   | 205/318 [00:50<00:27,  4.12it/s]\u001b[A\n",
            " 65%|██████▍   | 206/318 [00:51<00:26,  4.15it/s]\u001b[A\n",
            " 65%|██████▌   | 207/318 [00:51<00:24,  4.47it/s]\u001b[A\n",
            " 65%|██████▌   | 208/318 [00:51<00:24,  4.47it/s]\u001b[A\n",
            " 66%|██████▌   | 209/318 [00:51<00:25,  4.34it/s]\u001b[A\n",
            " 66%|██████▌   | 210/318 [00:52<00:24,  4.38it/s]\u001b[A\n",
            " 66%|██████▋   | 211/318 [00:52<00:23,  4.53it/s]\u001b[A\n",
            " 67%|██████▋   | 212/318 [00:52<00:23,  4.45it/s]\u001b[A\n",
            " 67%|██████▋   | 213/318 [00:52<00:22,  4.59it/s]\u001b[A\n",
            " 67%|██████▋   | 214/318 [00:52<00:22,  4.58it/s]\u001b[A\n",
            " 68%|██████▊   | 215/318 [00:53<00:23,  4.33it/s]\u001b[A\n",
            " 68%|██████▊   | 216/318 [00:53<00:25,  3.98it/s]\u001b[A\n",
            " 68%|██████▊   | 217/318 [00:53<00:25,  3.97it/s]\u001b[A\n",
            " 69%|██████▊   | 218/318 [00:54<00:25,  3.87it/s]\u001b[A\n",
            " 69%|██████▉   | 219/318 [00:54<00:24,  4.11it/s]\u001b[A\n",
            " 69%|██████▉   | 220/318 [00:54<00:24,  4.06it/s]\u001b[A\n",
            " 69%|██████▉   | 221/318 [00:54<00:26,  3.60it/s]\u001b[A\n",
            " 70%|██████▉   | 222/318 [00:55<00:27,  3.52it/s]\u001b[A\n",
            " 70%|███████   | 223/318 [00:55<00:26,  3.59it/s]\u001b[A\n",
            " 70%|███████   | 224/318 [00:55<00:25,  3.66it/s]\u001b[A\n",
            " 71%|███████   | 225/318 [00:55<00:24,  3.80it/s]\u001b[A\n",
            " 71%|███████   | 226/318 [00:56<00:24,  3.72it/s]\u001b[A\n",
            " 71%|███████▏  | 227/318 [00:56<00:23,  3.94it/s]\u001b[A\n",
            " 72%|███████▏  | 228/318 [00:56<00:24,  3.75it/s]\u001b[A\n",
            " 72%|███████▏  | 229/318 [00:56<00:23,  3.82it/s]\u001b[A\n",
            " 72%|███████▏  | 230/318 [00:57<00:22,  3.95it/s]\u001b[A\n",
            " 73%|███████▎  | 231/318 [00:57<00:21,  4.02it/s]\u001b[A\n",
            " 73%|███████▎  | 232/318 [00:57<00:21,  4.01it/s]\u001b[A\n",
            " 73%|███████▎  | 233/318 [00:57<00:21,  3.89it/s]\u001b[A\n",
            " 74%|███████▎  | 234/318 [00:58<00:22,  3.78it/s]\u001b[A\n",
            " 74%|███████▍  | 235/318 [00:58<00:20,  4.04it/s]\u001b[A\n",
            " 74%|███████▍  | 236/318 [00:58<00:19,  4.25it/s]\u001b[A\n",
            " 75%|███████▍  | 237/318 [00:58<00:18,  4.38it/s]\u001b[A\n",
            " 75%|███████▍  | 238/318 [00:59<00:17,  4.58it/s]\u001b[A\n",
            " 75%|███████▌  | 239/318 [00:59<00:18,  4.21it/s]\u001b[A\n",
            " 75%|███████▌  | 240/318 [00:59<00:18,  4.25it/s]\u001b[A\n",
            " 76%|███████▌  | 241/318 [00:59<00:18,  4.24it/s]\u001b[A\n",
            " 76%|███████▌  | 242/318 [01:00<00:18,  4.15it/s]\u001b[A\n",
            " 76%|███████▋  | 243/318 [01:00<00:17,  4.31it/s]\u001b[A\n",
            " 77%|███████▋  | 244/318 [01:00<00:17,  4.24it/s]\u001b[A\n",
            " 77%|███████▋  | 245/318 [01:00<00:17,  4.19it/s]\u001b[A\n",
            " 77%|███████▋  | 246/318 [01:00<00:17,  4.16it/s]\u001b[A\n",
            " 78%|███████▊  | 247/318 [01:01<00:16,  4.33it/s]\u001b[A\n",
            " 78%|███████▊  | 248/318 [01:01<00:16,  4.23it/s]\u001b[A\n",
            " 78%|███████▊  | 249/318 [01:01<00:16,  4.29it/s]\u001b[A\n",
            " 79%|███████▊  | 250/318 [01:01<00:15,  4.36it/s]\u001b[A\n",
            " 79%|███████▉  | 251/318 [01:02<00:15,  4.31it/s]\u001b[A\n",
            " 79%|███████▉  | 252/318 [01:02<00:14,  4.41it/s]\u001b[A\n",
            " 80%|███████▉  | 253/318 [01:02<00:15,  4.17it/s]\u001b[A\n",
            " 80%|███████▉  | 254/318 [01:02<00:14,  4.28it/s]\u001b[A\n",
            " 80%|████████  | 255/318 [01:03<00:15,  4.19it/s]\u001b[A\n",
            " 81%|████████  | 256/318 [01:03<00:14,  4.14it/s]\u001b[A\n",
            " 81%|████████  | 257/318 [01:03<00:14,  4.17it/s]\u001b[A\n",
            " 81%|████████  | 258/318 [01:03<00:13,  4.32it/s]\u001b[A\n",
            " 81%|████████▏ | 259/318 [01:04<00:14,  4.21it/s]\u001b[A\n",
            " 82%|████████▏ | 260/318 [01:04<00:13,  4.25it/s]\u001b[A\n",
            " 82%|████████▏ | 261/318 [01:04<00:13,  4.08it/s]\u001b[A\n",
            " 82%|████████▏ | 262/318 [01:04<00:13,  4.10it/s]\u001b[A\n",
            " 83%|████████▎ | 263/318 [01:05<00:13,  4.03it/s]\u001b[A\n",
            " 83%|████████▎ | 264/318 [01:05<00:13,  4.14it/s]\u001b[A\n",
            " 83%|████████▎ | 265/318 [01:05<00:13,  3.96it/s]\u001b[A\n",
            " 84%|████████▎ | 266/318 [01:05<00:13,  3.90it/s]\u001b[A\n",
            " 84%|████████▍ | 267/318 [01:06<00:12,  3.99it/s]\u001b[A\n",
            " 84%|████████▍ | 268/318 [01:06<00:12,  3.92it/s]\u001b[A\n",
            " 85%|████████▍ | 269/318 [01:06<00:12,  4.07it/s]\u001b[A\n",
            " 85%|████████▍ | 270/318 [01:06<00:12,  3.89it/s]\u001b[A\n",
            " 85%|████████▌ | 271/318 [01:07<00:12,  3.85it/s]\u001b[A\n",
            " 86%|████████▌ | 272/318 [01:07<00:12,  3.75it/s]\u001b[A\n",
            " 86%|████████▌ | 273/318 [01:07<00:12,  3.53it/s]\u001b[A\n",
            " 86%|████████▌ | 274/318 [01:07<00:11,  3.72it/s]\u001b[A\n",
            " 86%|████████▋ | 275/318 [01:08<00:11,  3.84it/s]\u001b[A\n",
            " 87%|████████▋ | 276/318 [01:08<00:10,  3.96it/s]\u001b[A\n",
            " 87%|████████▋ | 277/318 [01:08<00:10,  4.06it/s]\u001b[A\n",
            " 87%|████████▋ | 278/318 [01:08<00:10,  3.96it/s]\u001b[A\n",
            " 88%|████████▊ | 279/318 [01:09<00:09,  4.03it/s]\u001b[A\n",
            " 88%|████████▊ | 280/318 [01:09<00:08,  4.24it/s]\u001b[A\n",
            " 88%|████████▊ | 281/318 [01:09<00:08,  4.16it/s]\u001b[A\n",
            " 89%|████████▊ | 282/318 [01:09<00:09,  3.89it/s]\u001b[A\n",
            " 89%|████████▉ | 283/318 [01:10<00:08,  4.11it/s]\u001b[A\n",
            " 89%|████████▉ | 284/318 [01:10<00:08,  3.87it/s]\u001b[A\n",
            " 90%|████████▉ | 285/318 [01:10<00:09,  3.66it/s]\u001b[A\n",
            " 90%|████████▉ | 286/318 [01:10<00:08,  3.73it/s]\u001b[A\n",
            " 90%|█████████ | 287/318 [01:11<00:08,  3.74it/s]\u001b[A\n",
            " 91%|█████████ | 288/318 [01:11<00:07,  3.85it/s]\u001b[A\n",
            " 91%|█████████ | 289/318 [01:11<00:07,  3.91it/s]\u001b[A\n",
            " 91%|█████████ | 290/318 [01:11<00:07,  3.85it/s]\u001b[A\n",
            " 92%|█████████▏| 291/318 [01:12<00:06,  4.04it/s]\u001b[A\n",
            " 92%|█████████▏| 292/318 [01:12<00:06,  4.01it/s]\u001b[A\n",
            " 92%|█████████▏| 293/318 [01:12<00:06,  4.03it/s]\u001b[A\n",
            " 92%|█████████▏| 294/318 [01:12<00:06,  3.85it/s]\u001b[A\n",
            " 93%|█████████▎| 295/318 [01:13<00:06,  3.73it/s]\u001b[A\n",
            " 93%|█████████▎| 296/318 [01:13<00:05,  3.97it/s]\u001b[A\n",
            " 93%|█████████▎| 297/318 [01:13<00:05,  4.05it/s]\u001b[A\n",
            " 94%|█████████▎| 298/318 [01:13<00:04,  4.11it/s]\u001b[A\n",
            " 94%|█████████▍| 299/318 [01:14<00:04,  3.90it/s]\u001b[A\n",
            " 94%|█████████▍| 300/318 [01:14<00:04,  4.05it/s]\u001b[A\n",
            " 95%|█████████▍| 301/318 [01:14<00:04,  4.13it/s]\u001b[A\n",
            " 95%|█████████▍| 302/318 [01:14<00:04,  3.95it/s]\u001b[A\n",
            " 95%|█████████▌| 303/318 [01:15<00:03,  3.82it/s]\u001b[A\n",
            " 96%|█████████▌| 304/318 [01:15<00:03,  4.05it/s]\u001b[A\n",
            " 96%|█████████▌| 305/318 [01:15<00:03,  3.90it/s]\u001b[A\n",
            " 96%|█████████▌| 306/318 [01:15<00:02,  4.17it/s]\u001b[A\n",
            " 97%|█████████▋| 307/318 [01:16<00:02,  4.12it/s]\u001b[A\n",
            " 97%|█████████▋| 308/318 [01:16<00:02,  3.89it/s]\u001b[A\n",
            " 97%|█████████▋| 309/318 [01:16<00:02,  3.74it/s]\u001b[A\n",
            " 97%|█████████▋| 310/318 [01:17<00:02,  3.69it/s]\u001b[A\n",
            " 98%|█████████▊| 311/318 [01:17<00:01,  3.64it/s]\u001b[A\n",
            " 98%|█████████▊| 312/318 [01:17<00:01,  3.74it/s]\u001b[A\n",
            " 98%|█████████▊| 313/318 [01:17<00:01,  3.79it/s]\u001b[A\n",
            " 99%|█████████▊| 314/318 [01:18<00:01,  3.67it/s]\u001b[A\n",
            " 99%|█████████▉| 315/318 [01:18<00:00,  3.85it/s]\u001b[A\n",
            " 99%|█████████▉| 316/318 [01:18<00:00,  3.68it/s]\u001b[A\n",
            "100%|█████████▉| 317/318 [01:18<00:00,  3.62it/s]\u001b[A\n",
            "100%|██████████| 318/318 [01:19<00:00,  4.01it/s]\n",
            " 50%|█████     | 5/10 [06:39<06:37, 79.41s/it]\n",
            "  0%|          | 0/318 [00:00<?, ?it/s]\u001b[A\n",
            "  0%|          | 1/318 [00:00<01:53,  2.80it/s]\u001b[A\n",
            "  1%|          | 2/318 [00:00<01:33,  3.37it/s]\u001b[A\n",
            "  1%|          | 3/318 [00:00<01:17,  4.06it/s]\u001b[A\n",
            "  1%|▏         | 4/318 [00:01<01:15,  4.14it/s]\u001b[A\n",
            "  2%|▏         | 5/318 [00:01<01:20,  3.91it/s]\u001b[A\n",
            "  2%|▏         | 6/318 [00:01<01:15,  4.14it/s]\u001b[A\n",
            "  2%|▏         | 7/318 [00:01<01:15,  4.13it/s]\u001b[A\n",
            "  3%|▎         | 8/318 [00:02<01:18,  3.95it/s]\u001b[A\n",
            "  3%|▎         | 9/318 [00:02<01:17,  3.97it/s]\u001b[A\n",
            "  3%|▎         | 10/318 [00:02<01:20,  3.80it/s]\u001b[A\n",
            "  3%|▎         | 11/318 [00:02<01:15,  4.05it/s]\u001b[A\n",
            "  4%|▍         | 12/318 [00:03<01:17,  3.97it/s]\u001b[A\n",
            "  4%|▍         | 13/318 [00:03<01:14,  4.09it/s]\u001b[A\n",
            "  4%|▍         | 14/318 [00:03<01:16,  3.97it/s]\u001b[A\n",
            "  5%|▍         | 15/318 [00:03<01:14,  4.06it/s]\u001b[A\n",
            "  5%|▌         | 16/318 [00:04<01:12,  4.17it/s]\u001b[A\n",
            "  5%|▌         | 17/318 [00:04<01:15,  3.98it/s]\u001b[A\n",
            "  6%|▌         | 18/318 [00:04<01:14,  4.05it/s]\u001b[A\n",
            "  6%|▌         | 19/318 [00:04<01:13,  4.07it/s]\u001b[A\n",
            "  6%|▋         | 20/318 [00:04<01:08,  4.34it/s]\u001b[A\n",
            "  7%|▋         | 21/318 [00:05<01:08,  4.31it/s]\u001b[A\n",
            "  7%|▋         | 22/318 [00:05<01:05,  4.52it/s]\u001b[A\n",
            "  7%|▋         | 23/318 [00:05<01:11,  4.11it/s]\u001b[A\n",
            "  8%|▊         | 24/318 [00:05<01:11,  4.10it/s]\u001b[A\n",
            "  8%|▊         | 25/318 [00:06<01:13,  4.01it/s]\u001b[A\n",
            "  8%|▊         | 26/318 [00:06<01:17,  3.78it/s]\u001b[A\n",
            "  8%|▊         | 27/318 [00:06<01:14,  3.90it/s]\u001b[A\n",
            "  9%|▉         | 28/318 [00:06<01:12,  4.02it/s]\u001b[A\n",
            "  9%|▉         | 29/318 [00:07<01:09,  4.18it/s]\u001b[A\n",
            "  9%|▉         | 30/318 [00:07<01:09,  4.13it/s]\u001b[A\n",
            " 10%|▉         | 31/318 [00:07<01:07,  4.26it/s]\u001b[A\n",
            " 10%|█         | 32/318 [00:07<01:09,  4.10it/s]\u001b[A\n",
            " 10%|█         | 33/318 [00:08<01:10,  4.02it/s]\u001b[A\n",
            " 11%|█         | 34/318 [00:08<01:07,  4.23it/s]\u001b[A\n",
            " 11%|█         | 35/318 [00:08<01:05,  4.31it/s]\u001b[A\n",
            " 11%|█▏        | 36/318 [00:08<01:07,  4.15it/s]\u001b[A\n",
            " 12%|█▏        | 37/318 [00:09<01:07,  4.18it/s]\u001b[A\n",
            " 12%|█▏        | 38/318 [00:09<01:06,  4.18it/s]\u001b[A\n",
            " 12%|█▏        | 39/318 [00:09<01:12,  3.86it/s]\u001b[A\n",
            " 13%|█▎        | 40/318 [00:09<01:09,  3.98it/s]\u001b[A\n",
            " 13%|█▎        | 41/318 [00:10<01:11,  3.87it/s]\u001b[A\n",
            " 13%|█▎        | 42/318 [00:10<01:07,  4.08it/s]\u001b[A\n",
            " 14%|█▎        | 43/318 [00:10<01:18,  3.50it/s]\u001b[A\n",
            " 14%|█▍        | 44/318 [00:11<01:17,  3.54it/s]\u001b[A\n",
            " 14%|█▍        | 45/318 [00:11<01:14,  3.69it/s]\u001b[A\n",
            " 14%|█▍        | 46/318 [00:11<01:11,  3.82it/s]\u001b[A\n",
            " 15%|█▍        | 47/318 [00:11<01:09,  3.91it/s]\u001b[A\n",
            " 15%|█▌        | 48/318 [00:12<01:10,  3.83it/s]\u001b[A\n",
            " 15%|█▌        | 49/318 [00:12<01:11,  3.76it/s]\u001b[A\n",
            " 16%|█▌        | 50/318 [00:12<01:11,  3.72it/s]\u001b[A\n",
            " 16%|█▌        | 51/318 [00:12<01:10,  3.77it/s]\u001b[A\n",
            " 16%|█▋        | 52/318 [00:13<01:07,  3.94it/s]\u001b[A\n",
            " 17%|█▋        | 53/318 [00:13<01:08,  3.89it/s]\u001b[A\n",
            " 17%|█▋        | 54/318 [00:13<01:05,  4.03it/s]\u001b[A\n",
            " 17%|█▋        | 55/318 [00:13<01:03,  4.13it/s]\u001b[A\n",
            " 18%|█▊        | 56/318 [00:14<01:04,  4.08it/s]\u001b[A\n",
            " 18%|█▊        | 57/318 [00:14<01:01,  4.28it/s]\u001b[A\n",
            " 18%|█▊        | 58/318 [00:14<01:01,  4.24it/s]\u001b[A\n",
            " 19%|█▊        | 59/318 [00:14<01:02,  4.17it/s]\u001b[A\n",
            " 19%|█▉        | 60/318 [00:14<01:02,  4.13it/s]\u001b[A\n",
            " 19%|█▉        | 61/318 [00:15<01:02,  4.12it/s]\u001b[A\n",
            " 19%|█▉        | 62/318 [00:15<01:02,  4.12it/s]\u001b[A\n",
            " 20%|█▉        | 63/318 [00:15<01:00,  4.18it/s]\u001b[A\n",
            " 20%|██        | 64/318 [00:15<01:00,  4.18it/s]\u001b[A\n",
            " 20%|██        | 65/318 [00:16<01:04,  3.94it/s]\u001b[A\n",
            " 21%|██        | 66/318 [00:16<01:05,  3.83it/s]\u001b[A\n",
            " 21%|██        | 67/318 [00:16<01:07,  3.73it/s]\u001b[A\n",
            " 21%|██▏       | 68/318 [00:17<01:06,  3.76it/s]\u001b[A\n",
            " 22%|██▏       | 69/318 [00:17<01:05,  3.77it/s]\u001b[A\n",
            " 22%|██▏       | 70/318 [00:17<01:10,  3.51it/s]\u001b[A\n",
            " 22%|██▏       | 71/318 [00:17<01:08,  3.58it/s]\u001b[A\n",
            " 23%|██▎       | 72/318 [00:18<01:09,  3.54it/s]\u001b[A\n",
            " 23%|██▎       | 73/318 [00:18<01:12,  3.37it/s]\u001b[A\n",
            " 23%|██▎       | 74/318 [00:18<01:09,  3.50it/s]\u001b[A\n",
            " 24%|██▎       | 75/318 [00:18<01:02,  3.90it/s]\u001b[A\n",
            " 24%|██▍       | 76/318 [00:19<01:02,  3.87it/s]\u001b[A\n",
            " 24%|██▍       | 77/318 [00:19<00:57,  4.16it/s]\u001b[A\n",
            " 25%|██▍       | 78/318 [00:19<00:58,  4.13it/s]\u001b[A\n",
            " 25%|██▍       | 79/318 [00:19<00:54,  4.38it/s]\u001b[A\n",
            " 25%|██▌       | 80/318 [00:20<00:56,  4.21it/s]\u001b[A\n",
            " 25%|██▌       | 81/318 [00:20<00:55,  4.24it/s]\u001b[A\n",
            " 26%|██▌       | 82/318 [00:20<00:54,  4.35it/s]\u001b[A\n",
            " 26%|██▌       | 83/318 [00:20<01:01,  3.83it/s]\u001b[A\n",
            " 26%|██▋       | 84/318 [00:21<01:03,  3.68it/s]\u001b[A\n",
            " 27%|██▋       | 85/318 [00:21<01:04,  3.59it/s]\u001b[A\n",
            " 27%|██▋       | 86/318 [00:21<01:01,  3.74it/s]\u001b[A\n",
            " 27%|██▋       | 87/318 [00:22<01:02,  3.72it/s]\u001b[A\n",
            " 28%|██▊       | 88/318 [00:22<01:03,  3.62it/s]\u001b[A\n",
            " 28%|██▊       | 89/318 [00:22<01:01,  3.73it/s]\u001b[A\n",
            " 28%|██▊       | 90/318 [00:22<01:03,  3.59it/s]\u001b[A\n",
            " 29%|██▊       | 91/318 [00:23<01:01,  3.70it/s]\u001b[A\n",
            " 29%|██▉       | 92/318 [00:23<01:01,  3.68it/s]\u001b[A\n",
            " 29%|██▉       | 93/318 [00:23<01:02,  3.59it/s]\u001b[A\n",
            " 30%|██▉       | 94/318 [00:23<01:01,  3.62it/s]\u001b[A\n",
            " 30%|██▉       | 95/318 [00:24<01:02,  3.59it/s]\u001b[A\n",
            " 30%|███       | 96/318 [00:24<01:04,  3.46it/s]\u001b[A\n",
            " 31%|███       | 97/318 [00:24<01:00,  3.65it/s]\u001b[A\n",
            " 31%|███       | 98/318 [00:25<01:01,  3.59it/s]\u001b[A\n",
            " 31%|███       | 99/318 [00:25<01:01,  3.58it/s]\u001b[A\n",
            " 31%|███▏      | 100/318 [00:25<00:58,  3.70it/s]\u001b[A\n",
            " 32%|███▏      | 101/318 [00:25<00:57,  3.79it/s]\u001b[A\n",
            " 32%|███▏      | 102/318 [00:26<00:56,  3.84it/s]\u001b[A\n",
            " 32%|███▏      | 103/318 [00:26<00:56,  3.77it/s]\u001b[A\n",
            " 33%|███▎      | 104/318 [00:26<00:53,  4.04it/s]\u001b[A\n",
            " 33%|███▎      | 105/318 [00:26<00:52,  4.02it/s]\u001b[A\n",
            " 33%|███▎      | 106/318 [00:27<00:48,  4.38it/s]\u001b[A\n",
            " 34%|███▎      | 107/318 [00:27<00:46,  4.50it/s]\u001b[A\n",
            " 34%|███▍      | 108/318 [00:27<00:49,  4.20it/s]\u001b[A\n",
            " 34%|███▍      | 109/318 [00:27<00:50,  4.16it/s]\u001b[A\n",
            " 35%|███▍      | 110/318 [00:27<00:48,  4.30it/s]\u001b[A\n",
            " 35%|███▍      | 111/318 [00:28<00:53,  3.89it/s]\u001b[A\n",
            " 35%|███▌      | 112/318 [00:28<00:52,  3.91it/s]\u001b[A\n",
            " 36%|███▌      | 113/318 [00:28<00:51,  3.96it/s]\u001b[A\n",
            " 36%|███▌      | 114/318 [00:29<00:51,  3.94it/s]\u001b[A\n",
            " 36%|███▌      | 115/318 [00:29<00:50,  3.99it/s]\u001b[A\n",
            " 36%|███▋      | 116/318 [00:29<00:50,  4.03it/s]\u001b[A\n",
            " 37%|███▋      | 117/318 [00:29<00:49,  4.04it/s]\u001b[A\n",
            " 37%|███▋      | 118/318 [00:30<00:50,  3.98it/s]\u001b[A\n",
            " 37%|███▋      | 119/318 [00:30<00:49,  4.06it/s]\u001b[A\n",
            " 38%|███▊      | 120/318 [00:30<00:50,  3.89it/s]\u001b[A\n",
            " 38%|███▊      | 121/318 [00:30<00:50,  3.90it/s]\u001b[A\n",
            " 38%|███▊      | 122/318 [00:31<00:47,  4.14it/s]\u001b[A\n",
            " 39%|███▊      | 123/318 [00:31<00:48,  4.03it/s]\u001b[A\n",
            " 39%|███▉      | 124/318 [00:31<00:48,  3.98it/s]\u001b[A\n",
            " 39%|███▉      | 125/318 [00:31<00:48,  3.98it/s]\u001b[A\n",
            " 40%|███▉      | 126/318 [00:32<00:47,  4.02it/s]\u001b[A\n",
            " 40%|███▉      | 127/318 [00:32<00:46,  4.08it/s]\u001b[A\n",
            " 40%|████      | 128/318 [00:32<00:45,  4.21it/s]\u001b[A\n",
            " 41%|████      | 129/318 [00:32<00:45,  4.19it/s]\u001b[A\n",
            " 41%|████      | 130/318 [00:32<00:45,  4.16it/s]\u001b[A\n",
            " 41%|████      | 131/318 [00:33<00:46,  3.98it/s]\u001b[A\n",
            " 42%|████▏     | 132/318 [00:33<00:44,  4.16it/s]\u001b[A\n",
            " 42%|████▏     | 133/318 [00:33<00:43,  4.28it/s]\u001b[A\n",
            " 42%|████▏     | 134/318 [00:33<00:42,  4.30it/s]\u001b[A\n",
            " 42%|████▏     | 135/318 [00:34<00:44,  4.15it/s]\u001b[A\n",
            " 43%|████▎     | 136/318 [00:34<00:44,  4.07it/s]\u001b[A\n",
            " 43%|████▎     | 137/318 [00:34<00:46,  3.87it/s]\u001b[A\n",
            " 43%|████▎     | 138/318 [00:34<00:46,  3.87it/s]\u001b[A\n",
            " 44%|████▎     | 139/318 [00:35<00:45,  3.91it/s]\u001b[A\n",
            " 44%|████▍     | 140/318 [00:35<00:43,  4.06it/s]\u001b[A\n",
            " 44%|████▍     | 141/318 [00:35<00:45,  3.92it/s]\u001b[A\n",
            " 45%|████▍     | 142/318 [00:35<00:44,  3.94it/s]\u001b[A\n",
            " 45%|████▍     | 143/318 [00:36<00:46,  3.77it/s]\u001b[A\n",
            " 45%|████▌     | 144/318 [00:36<00:44,  3.92it/s]\u001b[A\n",
            " 46%|████▌     | 145/318 [00:36<00:44,  3.85it/s]\u001b[A\n",
            " 46%|████▌     | 146/318 [00:37<00:43,  3.93it/s]\u001b[A\n",
            " 46%|████▌     | 147/318 [00:37<00:46,  3.70it/s]\u001b[A\n",
            " 47%|████▋     | 148/318 [00:37<00:43,  3.93it/s]\u001b[A\n",
            " 47%|████▋     | 149/318 [00:37<00:47,  3.58it/s]\u001b[A\n",
            " 47%|████▋     | 150/318 [00:38<00:47,  3.56it/s]\u001b[A\n",
            " 47%|████▋     | 151/318 [00:38<00:44,  3.77it/s]\u001b[A\n",
            " 48%|████▊     | 152/318 [00:38<00:41,  4.04it/s]\u001b[A\n",
            " 48%|████▊     | 153/318 [00:38<00:39,  4.17it/s]\u001b[A\n",
            " 48%|████▊     | 154/318 [00:39<00:38,  4.26it/s]\u001b[A\n",
            " 49%|████▊     | 155/318 [00:39<00:40,  4.06it/s]\u001b[A\n",
            " 49%|████▉     | 156/318 [00:39<00:39,  4.15it/s]\u001b[A\n",
            " 49%|████▉     | 157/318 [00:39<00:40,  3.98it/s]\u001b[A\n",
            " 50%|████▉     | 158/318 [00:40<00:39,  4.08it/s]\u001b[A\n",
            " 50%|█████     | 159/318 [00:40<00:38,  4.15it/s]\u001b[A\n",
            " 50%|█████     | 160/318 [00:40<00:38,  4.08it/s]\u001b[A\n",
            " 51%|█████     | 161/318 [00:40<00:39,  3.96it/s]\u001b[A\n",
            " 51%|█████     | 162/318 [00:41<00:41,  3.75it/s]\u001b[A\n",
            " 51%|█████▏    | 163/318 [00:41<00:43,  3.60it/s]\u001b[A\n",
            " 52%|█████▏    | 164/318 [00:41<00:38,  3.95it/s]\u001b[A\n",
            " 52%|█████▏    | 165/318 [00:41<00:39,  3.83it/s]\u001b[A\n",
            " 52%|█████▏    | 166/318 [00:42<00:37,  4.08it/s]\u001b[A\n",
            " 53%|█████▎    | 167/318 [00:42<00:38,  3.94it/s]\u001b[A\n",
            " 53%|█████▎    | 168/318 [00:42<00:36,  4.12it/s]\u001b[A\n",
            " 53%|█████▎    | 169/318 [00:42<00:36,  4.12it/s]\u001b[A\n",
            " 53%|█████▎    | 170/318 [00:43<00:34,  4.32it/s]\u001b[A\n",
            " 54%|█████▍    | 171/318 [00:43<00:34,  4.22it/s]\u001b[A\n",
            " 54%|█████▍    | 172/318 [00:43<00:35,  4.13it/s]\u001b[A\n",
            " 54%|█████▍    | 173/318 [00:43<00:34,  4.16it/s]\u001b[A\n",
            " 55%|█████▍    | 174/318 [00:44<00:34,  4.16it/s]\u001b[A\n",
            " 55%|█████▌    | 175/318 [00:44<00:32,  4.43it/s]\u001b[A\n",
            " 55%|█████▌    | 176/318 [00:44<00:31,  4.52it/s]\u001b[A\n",
            " 56%|█████▌    | 177/318 [00:44<00:33,  4.18it/s]\u001b[A\n",
            " 56%|█████▌    | 178/318 [00:44<00:33,  4.18it/s]\u001b[A\n",
            " 56%|█████▋    | 179/318 [00:45<00:34,  4.07it/s]\u001b[A\n",
            " 57%|█████▋    | 180/318 [00:45<00:33,  4.18it/s]\u001b[A\n",
            " 57%|█████▋    | 181/318 [00:45<00:32,  4.25it/s]\u001b[A\n",
            " 57%|█████▋    | 182/318 [00:45<00:33,  4.03it/s]\u001b[A\n",
            " 58%|█████▊    | 183/318 [00:46<00:33,  4.09it/s]\u001b[A\n",
            " 58%|█████▊    | 184/318 [00:46<00:31,  4.23it/s]\u001b[A\n",
            " 58%|█████▊    | 185/318 [00:46<00:30,  4.42it/s]\u001b[A\n",
            " 58%|█████▊    | 186/318 [00:46<00:29,  4.47it/s]\u001b[A\n",
            " 59%|█████▉    | 187/318 [00:47<00:30,  4.24it/s]\u001b[A\n",
            " 59%|█████▉    | 188/318 [00:47<00:30,  4.29it/s]\u001b[A\n",
            " 59%|█████▉    | 189/318 [00:47<00:32,  3.97it/s]\u001b[A\n",
            " 60%|█████▉    | 190/318 [00:47<00:33,  3.79it/s]\u001b[A\n",
            " 60%|██████    | 191/318 [00:48<00:33,  3.75it/s]\u001b[A\n",
            " 60%|██████    | 192/318 [00:48<00:33,  3.74it/s]\u001b[A\n",
            " 61%|██████    | 193/318 [00:48<00:31,  3.95it/s]\u001b[A\n",
            " 61%|██████    | 194/318 [00:48<00:32,  3.86it/s]\u001b[A\n",
            " 61%|██████▏   | 195/318 [00:49<00:34,  3.60it/s]\u001b[A\n",
            " 62%|██████▏   | 196/318 [00:49<00:31,  3.88it/s]\u001b[A\n",
            " 62%|██████▏   | 197/318 [00:49<00:29,  4.16it/s]\u001b[A\n",
            " 62%|██████▏   | 198/318 [00:49<00:30,  3.95it/s]\u001b[A\n",
            " 63%|██████▎   | 199/318 [00:50<00:29,  4.04it/s]\u001b[A\n",
            " 63%|██████▎   | 200/318 [00:50<00:28,  4.13it/s]\u001b[A\n",
            " 63%|██████▎   | 201/318 [00:50<00:29,  3.99it/s]\u001b[A\n",
            " 64%|██████▎   | 202/318 [00:50<00:29,  3.92it/s]\u001b[A\n",
            " 64%|██████▍   | 203/318 [00:51<00:28,  4.11it/s]\u001b[A\n",
            " 64%|██████▍   | 204/318 [00:51<00:29,  3.87it/s]\u001b[A\n",
            " 64%|██████▍   | 205/318 [00:51<00:28,  3.97it/s]\u001b[A\n",
            " 65%|██████▍   | 206/318 [00:51<00:28,  3.98it/s]\u001b[A\n",
            " 65%|██████▌   | 207/318 [00:52<00:27,  4.00it/s]\u001b[A\n",
            " 65%|██████▌   | 208/318 [00:52<00:25,  4.24it/s]\u001b[A\n",
            " 66%|██████▌   | 209/318 [00:52<00:24,  4.40it/s]\u001b[A\n",
            " 66%|██████▌   | 210/318 [00:52<00:23,  4.55it/s]\u001b[A\n",
            " 66%|██████▋   | 211/318 [00:53<00:24,  4.28it/s]\u001b[A\n",
            " 67%|██████▋   | 212/318 [00:53<00:26,  4.06it/s]\u001b[A\n",
            " 67%|██████▋   | 213/318 [00:53<00:26,  4.03it/s]\u001b[A\n",
            " 67%|██████▋   | 214/318 [00:53<00:24,  4.17it/s]\u001b[A\n",
            " 68%|██████▊   | 215/318 [00:53<00:23,  4.41it/s]\u001b[A\n",
            " 68%|██████▊   | 216/318 [00:54<00:23,  4.42it/s]\u001b[A\n",
            " 68%|██████▊   | 217/318 [00:54<00:22,  4.53it/s]\u001b[A\n",
            " 69%|██████▊   | 218/318 [00:54<00:22,  4.49it/s]\u001b[A\n",
            " 69%|██████▉   | 219/318 [00:54<00:22,  4.33it/s]\u001b[A\n",
            " 69%|██████▉   | 220/318 [00:55<00:23,  4.14it/s]\u001b[A\n",
            " 69%|██████▉   | 221/318 [00:55<00:24,  4.03it/s]\u001b[A\n",
            " 70%|██████▉   | 222/318 [00:55<00:24,  3.91it/s]\u001b[A\n",
            " 70%|███████   | 223/318 [00:55<00:25,  3.79it/s]\u001b[A\n",
            " 70%|███████   | 224/318 [00:56<00:24,  3.82it/s]\u001b[A\n",
            " 71%|███████   | 225/318 [00:56<00:24,  3.83it/s]\u001b[A\n",
            " 71%|███████   | 226/318 [00:56<00:23,  3.92it/s]\u001b[A\n",
            " 71%|███████▏  | 227/318 [00:56<00:22,  4.03it/s]\u001b[A\n",
            " 72%|███████▏  | 228/318 [00:57<00:22,  4.08it/s]\u001b[A\n",
            " 72%|███████▏  | 229/318 [00:57<00:22,  3.89it/s]\u001b[A\n",
            " 72%|███████▏  | 230/318 [00:57<00:22,  3.87it/s]\u001b[A\n",
            " 73%|███████▎  | 231/318 [00:58<00:22,  3.93it/s]\u001b[A\n",
            " 73%|███████▎  | 232/318 [00:58<00:21,  4.02it/s]\u001b[A\n",
            " 73%|███████▎  | 233/318 [00:58<00:20,  4.21it/s]\u001b[A\n",
            " 74%|███████▎  | 234/318 [00:58<00:19,  4.33it/s]\u001b[A\n",
            " 74%|███████▍  | 235/318 [00:58<00:20,  3.98it/s]\u001b[A\n",
            " 74%|███████▍  | 236/318 [00:59<00:21,  3.87it/s]\u001b[A\n",
            " 75%|███████▍  | 237/318 [00:59<00:20,  3.93it/s]\u001b[A\n",
            " 75%|███████▍  | 238/318 [00:59<00:19,  4.06it/s]\u001b[A\n",
            " 75%|███████▌  | 239/318 [01:00<00:21,  3.68it/s]\u001b[A\n",
            " 75%|███████▌  | 240/318 [01:00<00:20,  3.89it/s]\u001b[A\n",
            " 76%|███████▌  | 241/318 [01:00<00:20,  3.78it/s]\u001b[A\n",
            " 76%|███████▌  | 242/318 [01:00<00:20,  3.75it/s]\u001b[A\n",
            " 76%|███████▋  | 243/318 [01:01<00:20,  3.60it/s]\u001b[A\n",
            " 77%|███████▋  | 244/318 [01:01<00:20,  3.56it/s]\u001b[A\n",
            " 77%|███████▋  | 245/318 [01:01<00:19,  3.82it/s]\u001b[A\n",
            " 77%|███████▋  | 246/318 [01:01<00:18,  3.96it/s]\u001b[A\n",
            " 78%|███████▊  | 247/318 [01:02<00:17,  4.04it/s]\u001b[A\n",
            " 78%|███████▊  | 248/318 [01:02<00:17,  3.90it/s]\u001b[A\n",
            " 78%|███████▊  | 249/318 [01:02<00:16,  4.08it/s]\u001b[A\n",
            " 79%|███████▊  | 250/318 [01:02<00:17,  3.98it/s]\u001b[A\n",
            " 79%|███████▉  | 251/318 [01:03<00:17,  3.83it/s]\u001b[A\n",
            " 79%|███████▉  | 252/318 [01:03<00:16,  4.00it/s]\u001b[A\n",
            " 80%|███████▉  | 253/318 [01:03<00:16,  4.06it/s]\u001b[A\n",
            " 80%|███████▉  | 254/318 [01:03<00:15,  4.24it/s]\u001b[A\n",
            " 80%|████████  | 255/318 [01:04<00:15,  4.06it/s]\u001b[A\n",
            " 81%|████████  | 256/318 [01:04<00:15,  3.96it/s]\u001b[A\n",
            " 81%|████████  | 257/318 [01:04<00:15,  4.03it/s]\u001b[A\n",
            " 81%|████████  | 258/318 [01:04<00:14,  4.00it/s]\u001b[A\n",
            " 81%|████████▏ | 259/318 [01:05<00:13,  4.26it/s]\u001b[A\n",
            " 82%|████████▏ | 260/318 [01:05<00:13,  4.26it/s]\u001b[A\n",
            " 82%|████████▏ | 261/318 [01:05<00:13,  4.19it/s]\u001b[A\n",
            " 82%|████████▏ | 262/318 [01:05<00:14,  3.97it/s]\u001b[A\n",
            " 83%|████████▎ | 263/318 [01:06<00:14,  3.91it/s]\u001b[A\n",
            " 83%|████████▎ | 264/318 [01:06<00:13,  3.86it/s]\u001b[A\n",
            " 83%|████████▎ | 265/318 [01:06<00:13,  4.05it/s]\u001b[A\n",
            " 84%|████████▎ | 266/318 [01:06<00:12,  4.26it/s]\u001b[A\n",
            " 84%|████████▍ | 267/318 [01:07<00:12,  4.05it/s]\u001b[A\n",
            " 84%|████████▍ | 268/318 [01:07<00:12,  3.96it/s]\u001b[A\n",
            " 85%|████████▍ | 269/318 [01:07<00:11,  4.12it/s]\u001b[A\n",
            " 85%|████████▍ | 270/318 [01:07<00:12,  3.90it/s]\u001b[A\n",
            " 85%|████████▌ | 271/318 [01:08<00:12,  3.90it/s]\u001b[A\n",
            " 86%|████████▌ | 272/318 [01:08<00:11,  4.14it/s]\u001b[A\n",
            " 86%|████████▌ | 273/318 [01:08<00:10,  4.41it/s]\u001b[A\n",
            " 86%|████████▌ | 274/318 [01:08<00:09,  4.51it/s]\u001b[A\n",
            " 86%|████████▋ | 275/318 [01:08<00:10,  4.23it/s]\u001b[A\n",
            " 87%|████████▋ | 276/318 [01:09<00:09,  4.30it/s]\u001b[A\n",
            " 87%|████████▋ | 277/318 [01:09<00:09,  4.12it/s]\u001b[A\n",
            " 87%|████████▋ | 278/318 [01:09<00:09,  4.02it/s]\u001b[A\n",
            " 88%|████████▊ | 279/318 [01:09<00:10,  3.87it/s]\u001b[A\n",
            " 88%|████████▊ | 280/318 [01:10<00:09,  3.90it/s]\u001b[A\n",
            " 88%|████████▊ | 281/318 [01:10<00:09,  3.76it/s]\u001b[A\n",
            " 89%|████████▊ | 282/318 [01:10<00:09,  3.70it/s]\u001b[A\n",
            " 89%|████████▉ | 283/318 [01:11<00:09,  3.88it/s]\u001b[A\n",
            " 89%|████████▉ | 284/318 [01:11<00:08,  4.07it/s]\u001b[A\n",
            " 90%|████████▉ | 285/318 [01:11<00:08,  3.78it/s]\u001b[A\n",
            " 90%|████████▉ | 286/318 [01:11<00:08,  3.91it/s]\u001b[A\n",
            " 90%|█████████ | 287/318 [01:12<00:08,  3.80it/s]\u001b[A\n",
            " 91%|█████████ | 288/318 [01:12<00:07,  4.13it/s]\u001b[A\n",
            " 91%|█████████ | 289/318 [01:12<00:07,  4.12it/s]\u001b[A\n",
            " 91%|█████████ | 290/318 [01:12<00:06,  4.09it/s]\u001b[A\n",
            " 92%|█████████▏| 291/318 [01:13<00:06,  3.94it/s]\u001b[A\n",
            " 92%|█████████▏| 292/318 [01:13<00:07,  3.71it/s]\u001b[A\n",
            " 92%|█████████▏| 293/318 [01:13<00:06,  3.68it/s]\u001b[A\n",
            " 92%|█████████▏| 294/318 [01:13<00:06,  3.81it/s]\u001b[A\n",
            " 93%|█████████▎| 295/318 [01:14<00:05,  4.02it/s]\u001b[A\n",
            " 93%|█████████▎| 296/318 [01:14<00:05,  4.17it/s]\u001b[A\n",
            " 93%|█████████▎| 297/318 [01:14<00:05,  4.12it/s]\u001b[A\n",
            " 94%|█████████▎| 298/318 [01:14<00:04,  4.06it/s]\u001b[A\n",
            " 94%|█████████▍| 299/318 [01:15<00:04,  4.07it/s]\u001b[A\n",
            " 94%|█████████▍| 300/318 [01:15<00:04,  3.96it/s]\u001b[A\n",
            " 95%|█████████▍| 301/318 [01:15<00:04,  4.01it/s]\u001b[A\n",
            " 95%|█████████▍| 302/318 [01:15<00:03,  4.25it/s]\u001b[A\n",
            " 95%|█████████▌| 303/318 [01:16<00:03,  4.07it/s]\u001b[A\n",
            " 96%|█████████▌| 304/318 [01:16<00:03,  4.04it/s]\u001b[A\n",
            " 96%|█████████▌| 305/318 [01:16<00:03,  4.13it/s]\u001b[A\n",
            " 96%|█████████▌| 306/318 [01:16<00:02,  4.21it/s]\u001b[A\n",
            " 97%|█████████▋| 307/318 [01:16<00:02,  4.25it/s]\u001b[A\n",
            " 97%|█████████▋| 308/318 [01:17<00:02,  3.95it/s]\u001b[A\n",
            " 97%|█████████▋| 309/318 [01:17<00:02,  4.07it/s]\u001b[A\n",
            " 97%|█████████▋| 310/318 [01:17<00:01,  4.21it/s]\u001b[A\n",
            " 98%|█████████▊| 311/318 [01:17<00:01,  4.18it/s]\u001b[A\n",
            " 98%|█████████▊| 312/318 [01:18<00:01,  4.25it/s]\u001b[A\n",
            " 98%|█████████▊| 313/318 [01:18<00:01,  4.36it/s]\u001b[A\n",
            " 99%|█████████▊| 314/318 [01:18<00:00,  4.30it/s]\u001b[A\n",
            " 99%|█████████▉| 315/318 [01:18<00:00,  4.37it/s]\u001b[A\n",
            " 99%|█████████▉| 316/318 [01:19<00:00,  4.17it/s]\u001b[A\n",
            "100%|█████████▉| 317/318 [01:19<00:00,  3.89it/s]\u001b[A\n",
            "100%|██████████| 318/318 [01:19<00:00,  3.99it/s]\n",
            " 60%|██████    | 6/10 [07:59<05:18, 79.54s/it]\n",
            "  0%|          | 0/318 [00:00<?, ?it/s]\u001b[A\n",
            "  0%|          | 1/318 [00:00<02:10,  2.44it/s]\u001b[A\n",
            "  1%|          | 2/318 [00:00<01:41,  3.11it/s]\u001b[A\n",
            "  1%|          | 3/318 [00:00<01:34,  3.32it/s]\u001b[A\n",
            "  1%|▏         | 4/318 [00:01<01:32,  3.40it/s]\u001b[A\n",
            "  2%|▏         | 5/318 [00:01<01:32,  3.39it/s]\u001b[A\n",
            "  2%|▏         | 6/318 [00:01<01:20,  3.86it/s]\u001b[A\n",
            "  2%|▏         | 7/318 [00:02<01:24,  3.69it/s]\u001b[A\n",
            "  3%|▎         | 8/318 [00:02<01:23,  3.70it/s]\u001b[A\n",
            "  3%|▎         | 9/318 [00:02<01:21,  3.77it/s]\u001b[A\n",
            "  3%|▎         | 10/318 [00:02<01:21,  3.80it/s]\u001b[A\n",
            "  3%|▎         | 11/318 [00:03<01:19,  3.88it/s]\u001b[A\n",
            "  4%|▍         | 12/318 [00:03<01:16,  3.99it/s]\u001b[A\n",
            "  4%|▍         | 13/318 [00:03<01:18,  3.90it/s]\u001b[A\n",
            "  4%|▍         | 14/318 [00:03<01:16,  3.98it/s]\u001b[A\n",
            "  5%|▍         | 15/318 [00:04<01:16,  3.95it/s]\u001b[A\n",
            "  5%|▌         | 16/318 [00:04<01:17,  3.90it/s]\u001b[A\n",
            "  5%|▌         | 17/318 [00:04<01:14,  4.02it/s]\u001b[A\n",
            "  6%|▌         | 18/318 [00:04<01:11,  4.19it/s]\u001b[A\n",
            "  6%|▌         | 19/318 [00:04<01:10,  4.21it/s]\u001b[A\n",
            "  6%|▋         | 20/318 [00:05<01:13,  4.08it/s]\u001b[A\n",
            "  7%|▋         | 21/318 [00:05<01:11,  4.13it/s]\u001b[A\n",
            "  7%|▋         | 22/318 [00:05<01:09,  4.27it/s]\u001b[A\n",
            "  7%|▋         | 23/318 [00:05<01:12,  4.10it/s]\u001b[A\n",
            "  8%|▊         | 24/318 [00:06<01:19,  3.72it/s]\u001b[A\n",
            "  8%|▊         | 25/318 [00:06<01:16,  3.83it/s]\u001b[A\n",
            "  8%|▊         | 26/318 [00:06<01:13,  3.98it/s]\u001b[A\n",
            "  8%|▊         | 27/318 [00:07<01:13,  3.98it/s]\u001b[A\n",
            "  9%|▉         | 28/318 [00:07<01:18,  3.70it/s]\u001b[A\n",
            "  9%|▉         | 29/318 [00:07<01:14,  3.90it/s]\u001b[A\n",
            "  9%|▉         | 30/318 [00:07<01:09,  4.15it/s]\u001b[A\n",
            " 10%|▉         | 31/318 [00:08<01:10,  4.05it/s]\u001b[A\n",
            " 10%|█         | 32/318 [00:08<01:10,  4.04it/s]\u001b[A\n",
            " 10%|█         | 33/318 [00:08<01:09,  4.13it/s]\u001b[A\n",
            " 11%|█         | 34/318 [00:08<01:10,  4.02it/s]\u001b[A\n",
            " 11%|█         | 35/318 [00:08<01:07,  4.17it/s]\u001b[A\n",
            " 11%|█▏        | 36/318 [00:09<01:03,  4.44it/s]\u001b[A\n",
            " 12%|█▏        | 37/318 [00:09<01:01,  4.55it/s]\u001b[A\n",
            " 12%|█▏        | 38/318 [00:09<01:00,  4.63it/s]\u001b[A\n",
            " 12%|█▏        | 39/318 [00:09<01:00,  4.62it/s]\u001b[A\n",
            " 13%|█▎        | 40/318 [00:10<01:06,  4.18it/s]\u001b[A\n",
            " 13%|█▎        | 41/318 [00:10<01:04,  4.27it/s]\u001b[A\n",
            " 13%|█▎        | 42/318 [00:10<01:02,  4.45it/s]\u001b[A\n",
            " 14%|█▎        | 43/318 [00:10<01:01,  4.46it/s]\u001b[A\n",
            " 14%|█▍        | 44/318 [00:10<01:03,  4.32it/s]\u001b[A\n",
            " 14%|█▍        | 45/318 [00:11<01:06,  4.12it/s]\u001b[A\n",
            " 14%|█▍        | 46/318 [00:11<01:06,  4.07it/s]\u001b[A\n",
            " 15%|█▍        | 47/318 [00:11<01:06,  4.10it/s]\u001b[A\n",
            " 15%|█▌        | 48/318 [00:11<01:04,  4.18it/s]\u001b[A\n",
            " 15%|█▌        | 49/318 [00:12<01:04,  4.16it/s]\u001b[A\n",
            " 16%|█▌        | 50/318 [00:12<01:08,  3.92it/s]\u001b[A\n",
            " 16%|█▌        | 51/318 [00:12<01:06,  4.05it/s]\u001b[A\n",
            " 16%|█▋        | 52/318 [00:13<01:12,  3.69it/s]\u001b[A\n",
            " 17%|█▋        | 53/318 [00:13<01:07,  3.93it/s]\u001b[A\n",
            " 17%|█▋        | 54/318 [00:13<01:04,  4.10it/s]\u001b[A\n",
            " 17%|█▋        | 55/318 [00:13<01:06,  3.96it/s]\u001b[A\n",
            " 18%|█▊        | 56/318 [00:14<01:05,  3.99it/s]\u001b[A\n",
            " 18%|█▊        | 57/318 [00:14<00:59,  4.39it/s]\u001b[A\n",
            " 18%|█▊        | 58/318 [00:14<01:00,  4.27it/s]\u001b[A\n",
            " 19%|█▊        | 59/318 [00:14<01:06,  3.89it/s]\u001b[A\n",
            " 19%|█▉        | 60/318 [00:15<01:08,  3.75it/s]\u001b[A\n",
            " 19%|█▉        | 61/318 [00:15<01:08,  3.74it/s]\u001b[A\n",
            " 19%|█▉        | 62/318 [00:15<01:06,  3.84it/s]\u001b[A\n",
            " 20%|█▉        | 63/318 [00:15<01:04,  3.97it/s]\u001b[A\n",
            " 20%|██        | 64/318 [00:16<01:04,  3.92it/s]\u001b[A\n",
            " 20%|██        | 65/318 [00:16<01:03,  3.98it/s]\u001b[A\n",
            " 21%|██        | 66/318 [00:16<01:04,  3.89it/s]\u001b[A\n",
            " 21%|██        | 67/318 [00:16<01:04,  3.88it/s]\u001b[A\n",
            " 21%|██▏       | 68/318 [00:17<01:03,  3.97it/s]\u001b[A\n",
            " 22%|██▏       | 69/318 [00:17<01:01,  4.07it/s]\u001b[A\n",
            " 22%|██▏       | 70/318 [00:17<01:00,  4.09it/s]\u001b[A\n",
            " 22%|██▏       | 71/318 [00:17<01:04,  3.84it/s]\u001b[A\n",
            " 23%|██▎       | 72/318 [00:18<01:04,  3.82it/s]\u001b[A\n",
            " 23%|██▎       | 73/318 [00:18<01:03,  3.87it/s]\u001b[A\n",
            " 23%|██▎       | 74/318 [00:18<01:01,  3.99it/s]\u001b[A\n",
            " 24%|██▎       | 75/318 [00:18<00:56,  4.32it/s]\u001b[A\n",
            " 24%|██▍       | 76/318 [00:18<00:54,  4.45it/s]\u001b[A\n",
            " 24%|██▍       | 77/318 [00:19<00:55,  4.34it/s]\u001b[A\n",
            " 25%|██▍       | 78/318 [00:19<00:56,  4.23it/s]\u001b[A\n",
            " 25%|██▍       | 79/318 [00:19<00:56,  4.20it/s]\u001b[A\n",
            " 25%|██▌       | 80/318 [00:20<01:00,  3.92it/s]\u001b[A\n",
            " 25%|██▌       | 81/318 [00:20<00:58,  4.04it/s]\u001b[A\n",
            " 26%|██▌       | 82/318 [00:20<00:58,  4.05it/s]\u001b[A\n",
            " 26%|██▌       | 83/318 [00:20<00:56,  4.17it/s]\u001b[A\n",
            " 26%|██▋       | 84/318 [00:20<00:58,  4.00it/s]\u001b[A\n",
            " 27%|██▋       | 85/318 [00:21<00:58,  4.00it/s]\u001b[A\n",
            " 27%|██▋       | 86/318 [00:21<00:58,  3.98it/s]\u001b[A\n",
            " 27%|██▋       | 87/318 [00:21<00:58,  3.96it/s]\u001b[A\n",
            " 28%|██▊       | 88/318 [00:21<00:57,  4.02it/s]\u001b[A\n",
            " 28%|██▊       | 89/318 [00:22<00:58,  3.90it/s]\u001b[A\n",
            " 28%|██▊       | 90/318 [00:22<00:58,  3.90it/s]\u001b[A\n",
            " 29%|██▊       | 91/318 [00:22<01:01,  3.68it/s]\u001b[A\n",
            " 29%|██▉       | 92/318 [00:23<00:59,  3.79it/s]\u001b[A\n",
            " 29%|██▉       | 93/318 [00:23<00:59,  3.79it/s]\u001b[A\n",
            " 30%|██▉       | 94/318 [00:23<01:00,  3.68it/s]\u001b[A\n",
            " 30%|██▉       | 95/318 [00:23<00:56,  3.94it/s]\u001b[A\n",
            " 30%|███       | 96/318 [00:24<00:53,  4.12it/s]\u001b[A\n",
            " 31%|███       | 97/318 [00:24<00:54,  4.06it/s]\u001b[A\n",
            " 31%|███       | 98/318 [00:24<00:54,  4.07it/s]\u001b[A\n",
            " 31%|███       | 99/318 [00:24<00:54,  4.05it/s]\u001b[A\n",
            " 31%|███▏      | 100/318 [00:25<00:52,  4.17it/s]\u001b[A\n",
            " 32%|███▏      | 101/318 [00:25<00:51,  4.21it/s]\u001b[A\n",
            " 32%|███▏      | 102/318 [00:25<00:53,  4.07it/s]\u001b[A\n",
            " 32%|███▏      | 103/318 [00:25<00:54,  3.96it/s]\u001b[A\n",
            " 33%|███▎      | 104/318 [00:26<00:51,  4.13it/s]\u001b[A\n",
            " 33%|███▎      | 105/318 [00:26<00:55,  3.87it/s]\u001b[A\n",
            " 33%|███▎      | 106/318 [00:26<00:57,  3.69it/s]\u001b[A\n",
            " 34%|███▎      | 107/318 [00:26<00:58,  3.61it/s]\u001b[A\n",
            " 34%|███▍      | 108/318 [00:27<00:58,  3.59it/s]\u001b[A\n",
            " 34%|███▍      | 109/318 [00:27<00:56,  3.71it/s]\u001b[A\n",
            " 35%|███▍      | 110/318 [00:27<00:53,  3.89it/s]\u001b[A\n",
            " 35%|███▍      | 111/318 [00:27<00:51,  4.00it/s]\u001b[A\n",
            " 35%|███▌      | 112/318 [00:28<00:50,  4.04it/s]\u001b[A\n",
            " 36%|███▌      | 113/318 [00:28<00:50,  4.05it/s]\u001b[A\n",
            " 36%|███▌      | 114/318 [00:28<00:51,  3.97it/s]\u001b[A\n",
            " 36%|███▌      | 115/318 [00:28<00:53,  3.80it/s]\u001b[A\n",
            " 36%|███▋      | 116/318 [00:29<00:51,  3.94it/s]\u001b[A\n",
            " 37%|███▋      | 117/318 [00:29<00:52,  3.82it/s]\u001b[A\n",
            " 37%|███▋      | 118/318 [00:29<00:52,  3.82it/s]\u001b[A\n",
            " 37%|███▋      | 119/318 [00:29<00:52,  3.81it/s]\u001b[A\n",
            " 38%|███▊      | 120/318 [00:30<00:55,  3.56it/s]\u001b[A\n",
            " 38%|███▊      | 121/318 [00:30<00:51,  3.84it/s]\u001b[A\n",
            " 38%|███▊      | 122/318 [00:30<00:51,  3.82it/s]\u001b[A\n",
            " 39%|███▊      | 123/318 [00:31<00:52,  3.72it/s]\u001b[A\n",
            " 39%|███▉      | 124/318 [00:31<00:53,  3.65it/s]\u001b[A\n",
            " 39%|███▉      | 125/318 [00:31<00:51,  3.76it/s]\u001b[A\n",
            " 40%|███▉      | 126/318 [00:31<00:52,  3.64it/s]\u001b[A\n",
            " 40%|███▉      | 127/318 [00:32<00:49,  3.83it/s]\u001b[A\n",
            " 40%|████      | 128/318 [00:32<00:49,  3.82it/s]\u001b[A\n",
            " 41%|████      | 129/318 [00:32<00:48,  3.93it/s]\u001b[A\n",
            " 41%|████      | 130/318 [00:32<00:50,  3.69it/s]\u001b[A\n",
            " 41%|████      | 131/318 [00:33<00:48,  3.86it/s]\u001b[A\n",
            " 42%|████▏     | 132/318 [00:33<00:45,  4.05it/s]\u001b[A\n",
            " 42%|████▏     | 133/318 [00:33<00:46,  3.98it/s]\u001b[A\n",
            " 42%|████▏     | 134/318 [00:33<00:43,  4.26it/s]\u001b[A\n",
            " 42%|████▏     | 135/318 [00:34<00:44,  4.15it/s]\u001b[A\n",
            " 43%|████▎     | 136/318 [00:34<00:41,  4.43it/s]\u001b[A\n",
            " 43%|████▎     | 137/318 [00:34<00:40,  4.46it/s]\u001b[A\n",
            " 43%|████▎     | 138/318 [00:34<00:39,  4.54it/s]\u001b[A\n",
            " 44%|████▎     | 139/318 [00:34<00:40,  4.42it/s]\u001b[A\n",
            " 44%|████▍     | 140/318 [00:35<00:41,  4.29it/s]\u001b[A\n",
            " 44%|████▍     | 141/318 [00:35<00:39,  4.45it/s]\u001b[A\n",
            " 45%|████▍     | 142/318 [00:35<00:41,  4.25it/s]\u001b[A\n",
            " 45%|████▍     | 143/318 [00:35<00:41,  4.27it/s]\u001b[A\n",
            " 45%|████▌     | 144/318 [00:36<00:44,  3.92it/s]\u001b[A\n",
            " 46%|████▌     | 145/318 [00:36<00:45,  3.84it/s]\u001b[A\n",
            " 46%|████▌     | 146/318 [00:36<00:44,  3.84it/s]\u001b[A\n",
            " 46%|████▌     | 147/318 [00:36<00:42,  3.98it/s]\u001b[A\n",
            " 47%|████▋     | 148/318 [00:37<00:41,  4.12it/s]\u001b[A\n",
            " 47%|████▋     | 149/318 [00:37<00:39,  4.29it/s]\u001b[A\n",
            " 47%|████▋     | 150/318 [00:37<00:39,  4.24it/s]\u001b[A\n",
            " 47%|████▋     | 151/318 [00:37<00:39,  4.24it/s]\u001b[A\n",
            " 48%|████▊     | 152/318 [00:38<00:42,  3.94it/s]\u001b[A\n",
            " 48%|████▊     | 153/318 [00:38<00:40,  4.06it/s]\u001b[A\n",
            " 48%|████▊     | 154/318 [00:38<00:37,  4.39it/s]\u001b[A\n",
            " 49%|████▊     | 155/318 [00:38<00:35,  4.65it/s]\u001b[A\n",
            " 49%|████▉     | 156/318 [00:38<00:35,  4.57it/s]\u001b[A\n",
            " 49%|████▉     | 157/318 [00:39<00:34,  4.69it/s]\u001b[A\n",
            " 50%|████▉     | 158/318 [00:39<00:35,  4.45it/s]\u001b[A\n",
            " 50%|█████     | 159/318 [00:39<00:41,  3.80it/s]\u001b[A\n",
            " 50%|█████     | 160/318 [00:40<00:39,  3.98it/s]\u001b[A\n",
            " 51%|█████     | 161/318 [00:40<00:37,  4.19it/s]\u001b[A\n",
            " 51%|█████     | 162/318 [00:40<00:37,  4.15it/s]\u001b[A\n",
            " 51%|█████▏    | 163/318 [00:40<00:41,  3.76it/s]\u001b[A\n",
            " 52%|█████▏    | 164/318 [00:41<00:41,  3.68it/s]\u001b[A\n",
            " 52%|█████▏    | 165/318 [00:41<00:42,  3.56it/s]\u001b[A\n",
            " 52%|█████▏    | 166/318 [00:41<00:42,  3.56it/s]\u001b[A\n",
            " 53%|█████▎    | 167/318 [00:41<00:43,  3.49it/s]\u001b[A\n",
            " 53%|█████▎    | 168/318 [00:42<00:42,  3.54it/s]\u001b[A\n",
            " 53%|█████▎    | 169/318 [00:42<00:39,  3.78it/s]\u001b[A\n",
            " 53%|█████▎    | 170/318 [00:42<00:36,  4.04it/s]\u001b[A\n",
            " 54%|█████▍    | 171/318 [00:42<00:35,  4.10it/s]\u001b[A\n",
            " 54%|█████▍    | 172/318 [00:43<00:36,  3.98it/s]\u001b[A\n",
            " 54%|█████▍    | 173/318 [00:43<00:35,  4.05it/s]\u001b[A\n",
            " 55%|█████▍    | 174/318 [00:43<00:35,  4.02it/s]\u001b[A\n",
            " 55%|█████▌    | 175/318 [00:43<00:35,  3.99it/s]\u001b[A\n",
            " 55%|█████▌    | 176/318 [00:44<00:36,  3.89it/s]\u001b[A\n",
            " 56%|█████▌    | 177/318 [00:44<00:37,  3.80it/s]\u001b[A\n",
            " 56%|█████▌    | 178/318 [00:44<00:33,  4.14it/s]\u001b[A\n",
            " 56%|█████▋    | 179/318 [00:44<00:32,  4.31it/s]\u001b[A\n",
            " 57%|█████▋    | 180/318 [00:45<00:32,  4.20it/s]\u001b[A\n",
            " 57%|█████▋    | 181/318 [00:45<00:34,  4.02it/s]\u001b[A\n",
            " 57%|█████▋    | 182/318 [00:45<00:33,  4.10it/s]\u001b[A\n",
            " 58%|█████▊    | 183/318 [00:45<00:33,  4.08it/s]\u001b[A\n",
            " 58%|█████▊    | 184/318 [00:46<00:32,  4.18it/s]\u001b[A\n",
            " 58%|█████▊    | 185/318 [00:46<00:32,  4.09it/s]\u001b[A\n",
            " 58%|█████▊    | 186/318 [00:46<00:32,  4.11it/s]\u001b[A\n",
            " 59%|█████▉    | 187/318 [00:46<00:32,  4.03it/s]\u001b[A\n",
            " 59%|█████▉    | 188/318 [00:47<00:32,  4.05it/s]\u001b[A\n",
            " 59%|█████▉    | 189/318 [00:47<00:30,  4.27it/s]\u001b[A\n",
            " 60%|█████▉    | 190/318 [00:47<00:29,  4.35it/s]\u001b[A\n",
            " 60%|██████    | 191/318 [00:47<00:30,  4.15it/s]\u001b[A\n",
            " 60%|██████    | 192/318 [00:48<00:31,  4.02it/s]\u001b[A\n",
            " 61%|██████    | 193/318 [00:48<00:28,  4.33it/s]\u001b[A\n",
            " 61%|██████    | 194/318 [00:48<00:29,  4.25it/s]\u001b[A\n",
            " 61%|██████▏   | 195/318 [00:48<00:30,  4.09it/s]\u001b[A\n",
            " 62%|██████▏   | 196/318 [00:49<00:30,  3.98it/s]\u001b[A\n",
            " 62%|██████▏   | 197/318 [00:49<00:29,  4.16it/s]\u001b[A\n",
            " 62%|██████▏   | 198/318 [00:49<00:30,  3.99it/s]\u001b[A\n",
            " 63%|██████▎   | 199/318 [00:49<00:30,  3.93it/s]\u001b[A\n",
            " 63%|██████▎   | 200/318 [00:49<00:28,  4.12it/s]\u001b[A\n",
            " 63%|██████▎   | 201/318 [00:50<00:30,  3.82it/s]\u001b[A\n",
            " 64%|██████▎   | 202/318 [00:50<00:28,  4.08it/s]\u001b[A\n",
            " 64%|██████▍   | 203/318 [00:50<00:29,  3.85it/s]\u001b[A\n",
            " 64%|██████▍   | 204/318 [00:51<00:29,  3.89it/s]\u001b[A\n",
            " 64%|██████▍   | 205/318 [00:51<00:27,  4.07it/s]\u001b[A\n",
            " 65%|██████▍   | 206/318 [00:51<00:27,  4.08it/s]\u001b[A\n",
            " 65%|██████▌   | 207/318 [00:51<00:29,  3.71it/s]\u001b[A\n",
            " 65%|██████▌   | 208/318 [00:52<00:30,  3.58it/s]\u001b[A\n",
            " 66%|██████▌   | 209/318 [00:52<00:28,  3.81it/s]\u001b[A\n",
            " 66%|██████▌   | 210/318 [00:52<00:28,  3.72it/s]\u001b[A\n",
            " 66%|██████▋   | 211/318 [00:52<00:28,  3.79it/s]\u001b[A\n",
            " 67%|██████▋   | 212/318 [00:53<00:26,  4.02it/s]\u001b[A\n",
            " 67%|██████▋   | 213/318 [00:53<00:26,  3.99it/s]\u001b[A\n",
            " 67%|██████▋   | 214/318 [00:53<00:28,  3.68it/s]\u001b[A\n",
            " 68%|██████▊   | 215/318 [00:53<00:26,  3.83it/s]\u001b[A\n",
            " 68%|██████▊   | 216/318 [00:54<00:26,  3.78it/s]\u001b[A\n",
            " 68%|██████▊   | 217/318 [00:54<00:26,  3.87it/s]\u001b[A\n",
            " 69%|██████▊   | 218/318 [00:54<00:27,  3.66it/s]\u001b[A\n",
            " 69%|██████▉   | 219/318 [00:55<00:26,  3.75it/s]\u001b[A\n",
            " 69%|██████▉   | 220/318 [00:55<00:26,  3.74it/s]\u001b[A\n",
            " 69%|██████▉   | 221/318 [00:55<00:24,  3.98it/s]\u001b[A\n",
            " 70%|██████▉   | 222/318 [00:55<00:24,  3.97it/s]\u001b[A\n",
            " 70%|███████   | 223/318 [00:55<00:23,  4.11it/s]\u001b[A\n",
            " 70%|███████   | 224/318 [00:56<00:23,  4.02it/s]\u001b[A\n",
            " 71%|███████   | 225/318 [00:56<00:23,  3.98it/s]\u001b[A\n",
            " 71%|███████   | 226/318 [00:56<00:22,  4.15it/s]\u001b[A\n",
            " 71%|███████▏  | 227/318 [00:56<00:23,  3.94it/s]\u001b[A\n",
            " 72%|███████▏  | 228/318 [00:57<00:26,  3.42it/s]\u001b[A\n",
            " 72%|███████▏  | 229/318 [00:57<00:24,  3.59it/s]\u001b[A\n",
            " 72%|███████▏  | 230/318 [00:57<00:24,  3.63it/s]\u001b[A\n",
            " 73%|███████▎  | 231/318 [00:58<00:23,  3.77it/s]\u001b[A\n",
            " 73%|███████▎  | 232/318 [00:58<00:22,  3.89it/s]\u001b[A\n",
            " 73%|███████▎  | 233/318 [00:58<00:21,  4.03it/s]\u001b[A\n",
            " 74%|███████▎  | 234/318 [00:58<00:19,  4.33it/s]\u001b[A\n",
            " 74%|███████▍  | 235/318 [00:59<00:19,  4.16it/s]\u001b[A\n",
            " 74%|███████▍  | 236/318 [00:59<00:19,  4.13it/s]\u001b[A\n",
            " 75%|███████▍  | 237/318 [00:59<00:19,  4.17it/s]\u001b[A\n",
            " 75%|███████▍  | 238/318 [00:59<00:19,  4.12it/s]\u001b[A\n",
            " 75%|███████▌  | 239/318 [01:00<00:19,  4.12it/s]\u001b[A\n",
            " 75%|███████▌  | 240/318 [01:00<00:18,  4.20it/s]\u001b[A\n",
            " 76%|███████▌  | 241/318 [01:00<00:18,  4.07it/s]\u001b[A\n",
            " 76%|███████▌  | 242/318 [01:00<00:19,  3.98it/s]\u001b[A\n",
            " 76%|███████▋  | 243/318 [01:01<00:18,  4.00it/s]\u001b[A\n",
            " 77%|███████▋  | 244/318 [01:01<00:19,  3.82it/s]\u001b[A\n",
            " 77%|███████▋  | 245/318 [01:01<00:19,  3.82it/s]\u001b[A\n",
            " 77%|███████▋  | 246/318 [01:01<00:19,  3.76it/s]\u001b[A\n",
            " 78%|███████▊  | 247/318 [01:02<00:18,  3.85it/s]\u001b[A\n",
            " 78%|███████▊  | 248/318 [01:02<00:19,  3.61it/s]\u001b[A\n",
            " 78%|███████▊  | 249/318 [01:02<00:17,  3.90it/s]\u001b[A\n",
            " 79%|███████▊  | 250/318 [01:02<00:17,  3.88it/s]\u001b[A\n",
            " 79%|███████▉  | 251/318 [01:03<00:16,  3.99it/s]\u001b[A\n",
            " 79%|███████▉  | 252/318 [01:03<00:16,  4.11it/s]\u001b[A\n",
            " 80%|███████▉  | 253/318 [01:03<00:15,  4.19it/s]\u001b[A\n",
            " 80%|███████▉  | 254/318 [01:03<00:14,  4.33it/s]\u001b[A\n",
            " 80%|████████  | 255/318 [01:04<00:14,  4.28it/s]\u001b[A\n",
            " 81%|████████  | 256/318 [01:04<00:14,  4.19it/s]\u001b[A\n",
            " 81%|████████  | 257/318 [01:04<00:14,  4.16it/s]\u001b[A\n",
            " 81%|████████  | 258/318 [01:04<00:15,  3.93it/s]\u001b[A\n",
            " 81%|████████▏ | 259/318 [01:05<00:15,  3.90it/s]\u001b[A\n",
            " 82%|████████▏ | 260/318 [01:05<00:15,  3.70it/s]\u001b[A\n",
            " 82%|████████▏ | 261/318 [01:05<00:15,  3.77it/s]\u001b[A\n",
            " 82%|████████▏ | 262/318 [01:05<00:14,  3.86it/s]\u001b[A\n",
            " 83%|████████▎ | 263/318 [01:06<00:15,  3.51it/s]\u001b[A\n",
            " 83%|████████▎ | 264/318 [01:06<00:15,  3.58it/s]\u001b[A\n",
            " 83%|████████▎ | 265/318 [01:06<00:15,  3.37it/s]\u001b[A\n",
            " 84%|████████▎ | 266/318 [01:07<00:16,  3.11it/s]\u001b[A\n",
            " 84%|████████▍ | 267/318 [01:07<00:15,  3.35it/s]\u001b[A\n",
            " 84%|████████▍ | 268/318 [01:07<00:13,  3.64it/s]\u001b[A\n",
            " 85%|████████▍ | 269/318 [01:07<00:12,  3.86it/s]\u001b[A\n",
            " 85%|████████▍ | 270/318 [01:08<00:12,  3.99it/s]\u001b[A\n",
            " 85%|████████▌ | 271/318 [01:08<00:11,  4.14it/s]\u001b[A\n",
            " 86%|████████▌ | 272/318 [01:08<00:11,  3.84it/s]\u001b[A\n",
            " 86%|████████▌ | 273/318 [01:08<00:12,  3.71it/s]\u001b[A\n",
            " 86%|████████▌ | 274/318 [01:09<00:11,  3.69it/s]\u001b[A\n",
            " 86%|████████▋ | 275/318 [01:09<00:10,  3.93it/s]\u001b[A\n",
            " 87%|████████▋ | 276/318 [01:09<00:10,  4.08it/s]\u001b[A\n",
            " 87%|████████▋ | 277/318 [01:09<00:10,  4.06it/s]\u001b[A\n",
            " 87%|████████▋ | 278/318 [01:10<00:09,  4.21it/s]\u001b[A\n",
            " 88%|████████▊ | 279/318 [01:10<00:09,  4.21it/s]\u001b[A\n",
            " 88%|████████▊ | 280/318 [01:10<00:08,  4.41it/s]\u001b[A\n",
            " 88%|████████▊ | 281/318 [01:10<00:08,  4.28it/s]\u001b[A\n",
            " 89%|████████▊ | 282/318 [01:11<00:08,  4.06it/s]\u001b[A\n",
            " 89%|████████▉ | 283/318 [01:11<00:08,  4.12it/s]\u001b[A\n",
            " 89%|████████▉ | 284/318 [01:11<00:08,  4.25it/s]\u001b[A\n",
            " 90%|████████▉ | 285/318 [01:11<00:08,  4.09it/s]\u001b[A\n",
            " 90%|████████▉ | 286/318 [01:12<00:07,  4.04it/s]\u001b[A\n",
            " 90%|█████████ | 287/318 [01:12<00:07,  3.90it/s]\u001b[A\n",
            " 91%|█████████ | 288/318 [01:12<00:07,  4.10it/s]\u001b[A\n",
            " 91%|█████████ | 289/318 [01:12<00:06,  4.55it/s]\u001b[A\n",
            " 91%|█████████ | 290/318 [01:12<00:05,  4.73it/s]\u001b[A\n",
            " 92%|█████████▏| 291/318 [01:13<00:06,  4.27it/s]\u001b[A\n",
            " 92%|█████████▏| 292/318 [01:13<00:05,  4.40it/s]\u001b[A\n",
            " 92%|█████████▏| 293/318 [01:13<00:05,  4.56it/s]\u001b[A\n",
            " 92%|█████████▏| 294/318 [01:13<00:05,  4.57it/s]\u001b[A\n",
            " 93%|█████████▎| 295/318 [01:14<00:05,  4.40it/s]\u001b[A\n",
            " 93%|█████████▎| 296/318 [01:14<00:05,  4.26it/s]\u001b[A\n",
            " 93%|█████████▎| 297/318 [01:14<00:04,  4.69it/s]\u001b[A\n",
            " 94%|█████████▎| 298/318 [01:14<00:04,  4.61it/s]\u001b[A\n",
            " 94%|█████████▍| 299/318 [01:15<00:04,  4.01it/s]\u001b[A\n",
            " 94%|█████████▍| 300/318 [01:15<00:04,  3.99it/s]\u001b[A\n",
            " 95%|█████████▍| 301/318 [01:15<00:04,  4.07it/s]\u001b[A\n",
            " 95%|█████████▍| 302/318 [01:15<00:04,  3.97it/s]\u001b[A\n",
            " 95%|█████████▌| 303/318 [01:16<00:03,  3.97it/s]\u001b[A\n",
            " 96%|█████████▌| 304/318 [01:16<00:03,  3.93it/s]\u001b[A\n",
            " 96%|█████████▌| 305/318 [01:16<00:03,  4.08it/s]\u001b[A\n",
            " 96%|█████████▌| 306/318 [01:16<00:03,  3.92it/s]\u001b[A\n",
            " 97%|█████████▋| 307/318 [01:17<00:02,  3.92it/s]\u001b[A\n",
            " 97%|█████████▋| 308/318 [01:17<00:02,  3.78it/s]\u001b[A\n",
            " 97%|█████████▋| 309/318 [01:17<00:02,  3.77it/s]\u001b[A\n",
            " 97%|█████████▋| 310/318 [01:17<00:02,  3.43it/s]\u001b[A\n",
            " 98%|█████████▊| 311/318 [01:18<00:01,  3.70it/s]\u001b[A\n",
            " 98%|█████████▊| 312/318 [01:18<00:01,  3.91it/s]\u001b[A\n",
            " 98%|█████████▊| 313/318 [01:18<00:01,  3.99it/s]\u001b[A\n",
            " 99%|█████████▊| 314/318 [01:18<00:00,  4.02it/s]\u001b[A\n",
            " 99%|█████████▉| 315/318 [01:19<00:00,  4.01it/s]\u001b[A\n",
            " 99%|█████████▉| 316/318 [01:19<00:00,  4.07it/s]\u001b[A\n",
            "100%|█████████▉| 317/318 [01:19<00:00,  4.23it/s]\u001b[A\n",
            "100%|██████████| 318/318 [01:19<00:00,  3.98it/s]\n",
            " 70%|███████   | 7/10 [09:19<03:59, 79.67s/it]\n",
            "  0%|          | 0/318 [00:00<?, ?it/s]\u001b[A\n",
            "  0%|          | 1/318 [00:00<01:58,  2.68it/s]\u001b[A\n",
            "  1%|          | 2/318 [00:00<01:34,  3.35it/s]\u001b[A\n",
            "  1%|          | 3/318 [00:00<01:28,  3.56it/s]\u001b[A\n",
            "  1%|▏         | 4/318 [00:01<01:29,  3.50it/s]\u001b[A\n",
            "  2%|▏         | 5/318 [00:01<01:24,  3.72it/s]\u001b[A\n",
            "  2%|▏         | 6/318 [00:01<01:21,  3.82it/s]\u001b[A\n",
            "  2%|▏         | 7/318 [00:01<01:21,  3.81it/s]\u001b[A\n",
            "  3%|▎         | 8/318 [00:02<01:15,  4.10it/s]\u001b[A\n",
            "  3%|▎         | 9/318 [00:02<01:20,  3.84it/s]\u001b[A\n",
            "  3%|▎         | 10/318 [00:02<01:16,  4.00it/s]\u001b[A\n",
            "  3%|▎         | 11/318 [00:02<01:13,  4.15it/s]\u001b[A\n",
            "  4%|▍         | 12/318 [00:03<01:18,  3.89it/s]\u001b[A\n",
            "  4%|▍         | 13/318 [00:03<01:14,  4.11it/s]\u001b[A\n",
            "  4%|▍         | 14/318 [00:03<01:11,  4.28it/s]\u001b[A\n",
            "  5%|▍         | 15/318 [00:03<01:12,  4.19it/s]\u001b[A\n",
            "  5%|▌         | 16/318 [00:04<01:10,  4.26it/s]\u001b[A\n",
            "  5%|▌         | 17/318 [00:04<01:13,  4.07it/s]\u001b[A\n",
            "  6%|▌         | 18/318 [00:04<01:15,  3.98it/s]\u001b[A\n",
            "  6%|▌         | 19/318 [00:04<01:13,  4.04it/s]\u001b[A\n",
            "  6%|▋         | 20/318 [00:05<01:12,  4.09it/s]\u001b[A\n",
            "  7%|▋         | 21/318 [00:05<01:08,  4.33it/s]\u001b[A\n",
            "  7%|▋         | 22/318 [00:05<01:08,  4.30it/s]\u001b[A\n",
            "  7%|▋         | 23/318 [00:05<01:07,  4.34it/s]\u001b[A\n",
            "  8%|▊         | 24/318 [00:05<01:05,  4.49it/s]\u001b[A\n",
            "  8%|▊         | 25/318 [00:06<01:05,  4.45it/s]\u001b[A\n",
            "  8%|▊         | 26/318 [00:06<01:03,  4.60it/s]\u001b[A\n",
            "  8%|▊         | 27/318 [00:06<01:07,  4.33it/s]\u001b[A\n",
            "  9%|▉         | 28/318 [00:06<01:06,  4.35it/s]\u001b[A\n",
            "  9%|▉         | 29/318 [00:07<01:11,  4.06it/s]\u001b[A\n",
            "  9%|▉         | 30/318 [00:07<01:10,  4.08it/s]\u001b[A\n",
            " 10%|▉         | 31/318 [00:07<01:11,  4.00it/s]\u001b[A\n",
            " 10%|█         | 32/318 [00:07<01:12,  3.93it/s]\u001b[A\n",
            " 10%|█         | 33/318 [00:08<01:11,  3.99it/s]\u001b[A\n",
            " 11%|█         | 34/318 [00:08<01:09,  4.07it/s]\u001b[A\n",
            " 11%|█         | 35/318 [00:08<01:10,  3.99it/s]\u001b[A\n",
            " 11%|█▏        | 36/318 [00:08<01:18,  3.59it/s]\u001b[A\n",
            " 12%|█▏        | 37/318 [00:09<01:16,  3.69it/s]\u001b[A\n",
            " 12%|█▏        | 38/318 [00:09<01:12,  3.86it/s]\u001b[A\n",
            " 12%|█▏        | 39/318 [00:09<01:17,  3.59it/s]\u001b[A\n",
            " 13%|█▎        | 40/318 [00:10<01:12,  3.84it/s]\u001b[A\n",
            " 13%|█▎        | 41/318 [00:10<01:15,  3.68it/s]\u001b[A\n",
            " 13%|█▎        | 42/318 [00:10<01:15,  3.67it/s]\u001b[A\n",
            " 14%|█▎        | 43/318 [00:10<01:13,  3.76it/s]\u001b[A\n",
            " 14%|█▍        | 44/318 [00:11<01:13,  3.72it/s]\u001b[A\n",
            " 14%|█▍        | 45/318 [00:11<01:12,  3.76it/s]\u001b[A\n",
            " 14%|█▍        | 46/318 [00:11<01:09,  3.91it/s]\u001b[A\n",
            " 15%|█▍        | 47/318 [00:11<01:05,  4.14it/s]\u001b[A\n",
            " 15%|█▌        | 48/318 [00:12<01:04,  4.20it/s]\u001b[A\n",
            " 15%|█▌        | 49/318 [00:12<01:06,  4.02it/s]\u001b[A\n",
            " 16%|█▌        | 50/318 [00:12<01:04,  4.16it/s]\u001b[A\n",
            " 16%|█▌        | 51/318 [00:12<01:06,  4.00it/s]\u001b[A\n",
            " 16%|█▋        | 52/318 [00:13<01:06,  3.98it/s]\u001b[A\n",
            " 17%|█▋        | 53/318 [00:13<01:06,  3.98it/s]\u001b[A\n",
            " 17%|█▋        | 54/318 [00:13<01:05,  4.03it/s]\u001b[A\n",
            " 17%|█▋        | 55/318 [00:13<01:06,  3.96it/s]\u001b[A\n",
            " 18%|█▊        | 56/318 [00:14<01:02,  4.17it/s]\u001b[A\n",
            " 18%|█▊        | 57/318 [00:14<01:00,  4.29it/s]\u001b[A\n",
            " 18%|█▊        | 58/318 [00:14<01:03,  4.11it/s]\u001b[A\n",
            " 19%|█▊        | 59/318 [00:14<01:04,  4.02it/s]\u001b[A\n",
            " 19%|█▉        | 60/318 [00:15<01:06,  3.85it/s]\u001b[A\n",
            " 19%|█▉        | 61/318 [00:15<01:04,  4.01it/s]\u001b[A\n",
            " 19%|█▉        | 62/318 [00:15<01:02,  4.06it/s]\u001b[A\n",
            " 20%|█▉        | 63/318 [00:15<01:01,  4.17it/s]\u001b[A\n",
            " 20%|██        | 64/318 [00:16<01:01,  4.12it/s]\u001b[A\n",
            " 20%|██        | 65/318 [00:16<00:58,  4.29it/s]\u001b[A\n",
            " 21%|██        | 66/318 [00:16<00:58,  4.33it/s]\u001b[A\n",
            " 21%|██        | 67/318 [00:16<00:58,  4.28it/s]\u001b[A\n",
            " 21%|██▏       | 68/318 [00:16<01:02,  3.98it/s]\u001b[A\n",
            " 22%|██▏       | 69/318 [00:17<01:04,  3.84it/s]\u001b[A\n",
            " 22%|██▏       | 70/318 [00:17<01:00,  4.12it/s]\u001b[A\n",
            " 22%|██▏       | 71/318 [00:17<00:58,  4.25it/s]\u001b[A\n",
            " 23%|██▎       | 72/318 [00:17<00:59,  4.11it/s]\u001b[A\n",
            " 23%|██▎       | 73/318 [00:18<00:58,  4.22it/s]\u001b[A\n",
            " 23%|██▎       | 74/318 [00:18<00:57,  4.25it/s]\u001b[A\n",
            " 24%|██▎       | 75/318 [00:18<01:05,  3.73it/s]\u001b[A\n",
            " 24%|██▍       | 76/318 [00:18<01:02,  3.90it/s]\u001b[A\n",
            " 24%|██▍       | 77/318 [00:19<01:01,  3.93it/s]\u001b[A\n",
            " 25%|██▍       | 78/318 [00:19<01:00,  3.97it/s]\u001b[A\n",
            " 25%|██▍       | 79/318 [00:19<00:59,  4.04it/s]\u001b[A\n",
            " 25%|██▌       | 80/318 [00:19<00:56,  4.22it/s]\u001b[A\n",
            " 25%|██▌       | 81/318 [00:20<00:56,  4.21it/s]\u001b[A\n",
            " 26%|██▌       | 82/318 [00:20<00:55,  4.27it/s]\u001b[A\n",
            " 26%|██▌       | 83/318 [00:20<00:57,  4.06it/s]\u001b[A\n",
            " 26%|██▋       | 84/318 [00:20<00:55,  4.22it/s]\u001b[A\n",
            " 27%|██▋       | 85/318 [00:21<00:56,  4.09it/s]\u001b[A\n",
            " 27%|██▋       | 86/318 [00:21<00:55,  4.16it/s]\u001b[A\n",
            " 27%|██▋       | 87/318 [00:21<00:57,  4.01it/s]\u001b[A\n",
            " 28%|██▊       | 88/318 [00:21<00:58,  3.91it/s]\u001b[A\n",
            " 28%|██▊       | 89/318 [00:22<01:03,  3.59it/s]\u001b[A\n",
            " 28%|██▊       | 90/318 [00:22<00:59,  3.82it/s]\u001b[A\n",
            " 29%|██▊       | 91/318 [00:22<00:58,  3.91it/s]\u001b[A\n",
            " 29%|██▉       | 92/318 [00:22<00:57,  3.91it/s]\u001b[A\n",
            " 29%|██▉       | 93/318 [00:23<00:58,  3.84it/s]\u001b[A\n",
            " 30%|██▉       | 94/318 [00:23<00:56,  3.96it/s]\u001b[A\n",
            " 30%|██▉       | 95/318 [00:23<00:56,  3.94it/s]\u001b[A\n",
            " 30%|███       | 96/318 [00:23<00:56,  3.90it/s]\u001b[A\n",
            " 31%|███       | 97/318 [00:24<00:59,  3.69it/s]\u001b[A\n",
            " 31%|███       | 98/318 [00:24<00:53,  4.10it/s]\u001b[A\n",
            " 31%|███       | 99/318 [00:24<00:52,  4.18it/s]\u001b[A\n",
            " 31%|███▏      | 100/318 [00:24<00:53,  4.09it/s]\u001b[A\n",
            " 32%|███▏      | 101/318 [00:25<00:54,  3.98it/s]\u001b[A\n",
            " 32%|███▏      | 102/318 [00:25<00:54,  4.00it/s]\u001b[A\n",
            " 32%|███▏      | 103/318 [00:25<00:54,  3.94it/s]\u001b[A\n",
            " 33%|███▎      | 104/318 [00:26<00:55,  3.85it/s]\u001b[A\n",
            " 33%|███▎      | 105/318 [00:26<00:53,  3.96it/s]\u001b[A\n",
            " 33%|███▎      | 106/318 [00:26<00:52,  4.03it/s]\u001b[A\n",
            " 34%|███▎      | 107/318 [00:26<00:50,  4.18it/s]\u001b[A\n",
            " 34%|███▍      | 108/318 [00:27<00:56,  3.70it/s]\u001b[A\n",
            " 34%|███▍      | 109/318 [00:27<00:56,  3.70it/s]\u001b[A\n",
            " 35%|███▍      | 110/318 [00:27<00:54,  3.79it/s]\u001b[A\n",
            " 35%|███▍      | 111/318 [00:27<00:55,  3.76it/s]\u001b[A\n",
            " 35%|███▌      | 112/318 [00:28<00:51,  3.96it/s]\u001b[A\n",
            " 36%|███▌      | 113/318 [00:28<00:50,  4.03it/s]\u001b[A\n",
            " 36%|███▌      | 114/318 [00:28<00:51,  3.95it/s]\u001b[A\n",
            " 36%|███▌      | 115/318 [00:28<00:51,  3.94it/s]\u001b[A\n",
            " 36%|███▋      | 116/318 [00:29<00:49,  4.11it/s]\u001b[A\n",
            " 37%|███▋      | 117/318 [00:29<00:48,  4.16it/s]\u001b[A\n",
            " 37%|███▋      | 118/318 [00:29<00:47,  4.22it/s]\u001b[A\n",
            " 37%|███▋      | 119/318 [00:29<00:46,  4.32it/s]\u001b[A\n",
            " 38%|███▊      | 120/318 [00:29<00:45,  4.35it/s]\u001b[A\n",
            " 38%|███▊      | 121/318 [00:30<00:43,  4.53it/s]\u001b[A\n",
            " 38%|███▊      | 122/318 [00:30<00:44,  4.38it/s]\u001b[A\n",
            " 39%|███▊      | 123/318 [00:30<00:44,  4.42it/s]\u001b[A\n",
            " 39%|███▉      | 124/318 [00:30<00:43,  4.48it/s]\u001b[A\n",
            " 39%|███▉      | 125/318 [00:31<00:46,  4.12it/s]\u001b[A\n",
            " 40%|███▉      | 126/318 [00:31<00:45,  4.18it/s]\u001b[A\n",
            " 40%|███▉      | 127/318 [00:31<00:44,  4.30it/s]\u001b[A\n",
            " 40%|████      | 128/318 [00:31<00:46,  4.06it/s]\u001b[A\n",
            " 41%|████      | 129/318 [00:32<00:46,  4.05it/s]\u001b[A\n",
            " 41%|████      | 130/318 [00:32<00:45,  4.13it/s]\u001b[A\n",
            " 41%|████      | 131/318 [00:32<00:44,  4.18it/s]\u001b[A\n",
            " 42%|████▏     | 132/318 [00:32<00:46,  3.99it/s]\u001b[A\n",
            " 42%|████▏     | 133/318 [00:33<00:46,  4.01it/s]\u001b[A\n",
            " 42%|████▏     | 134/318 [00:33<00:50,  3.67it/s]\u001b[A\n",
            " 42%|████▏     | 135/318 [00:33<00:48,  3.80it/s]\u001b[A\n",
            " 43%|████▎     | 136/318 [00:33<00:44,  4.05it/s]\u001b[A\n",
            " 43%|████▎     | 137/318 [00:34<00:45,  4.00it/s]\u001b[A\n",
            " 43%|████▎     | 138/318 [00:34<00:44,  4.01it/s]\u001b[A\n",
            " 44%|████▎     | 139/318 [00:34<00:44,  4.03it/s]\u001b[A\n",
            " 44%|████▍     | 140/318 [00:34<00:44,  3.96it/s]\u001b[A\n",
            " 44%|████▍     | 141/318 [00:35<00:46,  3.83it/s]\u001b[A\n",
            " 45%|████▍     | 142/318 [00:35<00:49,  3.53it/s]\u001b[A\n",
            " 45%|████▍     | 143/318 [00:35<00:45,  3.84it/s]\u001b[A\n",
            " 45%|████▌     | 144/318 [00:35<00:43,  3.96it/s]\u001b[A\n",
            " 46%|████▌     | 145/318 [00:36<00:46,  3.73it/s]\u001b[A\n",
            " 46%|████▌     | 146/318 [00:36<00:45,  3.76it/s]\u001b[A\n",
            " 46%|████▌     | 147/318 [00:36<00:46,  3.71it/s]\u001b[A\n",
            " 47%|████▋     | 148/318 [00:36<00:42,  4.04it/s]\u001b[A\n",
            " 47%|████▋     | 149/318 [00:37<00:41,  4.05it/s]\u001b[A\n",
            " 47%|████▋     | 150/318 [00:37<00:40,  4.14it/s]\u001b[A\n",
            " 47%|████▋     | 151/318 [00:37<00:39,  4.20it/s]\u001b[A\n",
            " 48%|████▊     | 152/318 [00:37<00:39,  4.18it/s]\u001b[A\n",
            " 48%|████▊     | 153/318 [00:38<00:40,  4.08it/s]\u001b[A\n",
            " 48%|████▊     | 154/318 [00:38<00:39,  4.17it/s]\u001b[A\n",
            " 49%|████▊     | 155/318 [00:38<00:38,  4.22it/s]\u001b[A\n",
            " 49%|████▉     | 156/318 [00:38<00:39,  4.13it/s]\u001b[A\n",
            " 49%|████▉     | 157/318 [00:39<00:38,  4.14it/s]\u001b[A\n",
            " 50%|████▉     | 158/318 [00:39<00:37,  4.21it/s]\u001b[A\n",
            " 50%|█████     | 159/318 [00:39<00:37,  4.23it/s]\u001b[A\n",
            " 50%|█████     | 160/318 [00:39<00:38,  4.06it/s]\u001b[A\n",
            " 51%|█████     | 161/318 [00:40<00:40,  3.89it/s]\u001b[A\n",
            " 51%|█████     | 162/318 [00:40<00:38,  4.07it/s]\u001b[A\n",
            " 51%|█████▏    | 163/318 [00:40<00:38,  4.08it/s]\u001b[A\n",
            " 52%|█████▏    | 164/318 [00:40<00:40,  3.83it/s]\u001b[A\n",
            " 52%|█████▏    | 165/318 [00:41<00:41,  3.65it/s]\u001b[A\n",
            " 52%|█████▏    | 166/318 [00:41<00:40,  3.74it/s]\u001b[A\n",
            " 53%|█████▎    | 167/318 [00:41<00:40,  3.77it/s]\u001b[A\n",
            " 53%|█████▎    | 168/318 [00:41<00:41,  3.64it/s]\u001b[A\n",
            " 53%|█████▎    | 169/318 [00:42<00:38,  3.92it/s]\u001b[A\n",
            " 53%|█████▎    | 170/318 [00:42<00:38,  3.88it/s]\u001b[A\n",
            " 54%|█████▍    | 171/318 [00:42<00:37,  3.94it/s]\u001b[A\n",
            " 54%|█████▍    | 172/318 [00:42<00:34,  4.24it/s]\u001b[A\n",
            " 54%|█████▍    | 173/318 [00:43<00:35,  4.04it/s]\u001b[A\n",
            " 55%|█████▍    | 174/318 [00:43<00:36,  3.91it/s]\u001b[A\n",
            " 55%|█████▌    | 175/318 [00:43<00:39,  3.65it/s]\u001b[A\n",
            " 55%|█████▌    | 176/318 [00:44<00:37,  3.81it/s]\u001b[A\n",
            " 56%|█████▌    | 177/318 [00:44<00:34,  4.11it/s]\u001b[A\n",
            " 56%|█████▌    | 178/318 [00:44<00:32,  4.30it/s]\u001b[A\n",
            " 56%|█████▋    | 179/318 [00:44<00:33,  4.11it/s]\u001b[A\n",
            " 57%|█████▋    | 180/318 [00:44<00:34,  3.98it/s]\u001b[A\n",
            " 57%|█████▋    | 181/318 [00:45<00:36,  3.72it/s]\u001b[A\n",
            " 57%|█████▋    | 182/318 [00:45<00:34,  3.90it/s]\u001b[A\n",
            " 58%|█████▊    | 183/318 [00:45<00:34,  3.86it/s]\u001b[A\n",
            " 58%|█████▊    | 184/318 [00:45<00:32,  4.10it/s]\u001b[A\n",
            " 58%|█████▊    | 185/318 [00:46<00:33,  3.95it/s]\u001b[A\n",
            " 58%|█████▊    | 186/318 [00:46<00:33,  4.00it/s]\u001b[A\n",
            " 59%|█████▉    | 187/318 [00:46<00:33,  3.86it/s]\u001b[A\n",
            " 59%|█████▉    | 188/318 [00:47<00:35,  3.68it/s]\u001b[A\n",
            " 59%|█████▉    | 189/318 [00:47<00:35,  3.62it/s]\u001b[A\n",
            " 60%|█████▉    | 190/318 [00:47<00:36,  3.52it/s]\u001b[A\n",
            " 60%|██████    | 191/318 [00:47<00:32,  3.87it/s]\u001b[A\n",
            " 60%|██████    | 192/318 [00:48<00:32,  3.91it/s]\u001b[A\n",
            " 61%|██████    | 193/318 [00:48<00:31,  3.92it/s]\u001b[A\n",
            " 61%|██████    | 194/318 [00:48<00:30,  4.02it/s]\u001b[A\n",
            " 61%|██████▏   | 195/318 [00:48<00:31,  3.97it/s]\u001b[A\n",
            " 62%|██████▏   | 196/318 [00:49<00:29,  4.18it/s]\u001b[A\n",
            " 62%|██████▏   | 197/318 [00:49<00:29,  4.08it/s]\u001b[A\n",
            " 62%|██████▏   | 198/318 [00:49<00:30,  3.89it/s]\u001b[A\n",
            " 63%|██████▎   | 199/318 [00:49<00:30,  3.91it/s]\u001b[A\n",
            " 63%|██████▎   | 200/318 [00:50<00:29,  3.94it/s]\u001b[A\n",
            " 63%|██████▎   | 201/318 [00:50<00:30,  3.86it/s]\u001b[A\n",
            " 64%|██████▎   | 202/318 [00:50<00:30,  3.80it/s]\u001b[A\n",
            " 64%|██████▍   | 203/318 [00:50<00:29,  3.95it/s]\u001b[A\n",
            " 64%|██████▍   | 204/318 [00:51<00:28,  4.06it/s]\u001b[A\n",
            " 64%|██████▍   | 205/318 [00:51<00:28,  3.95it/s]\u001b[A\n",
            " 65%|██████▍   | 206/318 [00:51<00:29,  3.84it/s]\u001b[A\n",
            " 65%|██████▌   | 207/318 [00:51<00:28,  3.90it/s]\u001b[A\n",
            " 65%|██████▌   | 208/318 [00:52<00:28,  3.88it/s]\u001b[A\n",
            " 66%|██████▌   | 209/318 [00:52<00:27,  3.97it/s]\u001b[A\n",
            " 66%|██████▌   | 210/318 [00:52<00:25,  4.19it/s]\u001b[A\n",
            " 66%|██████▋   | 211/318 [00:52<00:25,  4.15it/s]\u001b[A\n",
            " 67%|██████▋   | 212/318 [00:53<00:25,  4.19it/s]\u001b[A\n",
            " 67%|██████▋   | 213/318 [00:53<00:24,  4.26it/s]\u001b[A\n",
            " 67%|██████▋   | 214/318 [00:53<00:26,  3.95it/s]\u001b[A\n",
            " 68%|██████▊   | 215/318 [00:53<00:26,  3.96it/s]\u001b[A\n",
            " 68%|██████▊   | 216/318 [00:54<00:25,  4.04it/s]\u001b[A\n",
            " 68%|██████▊   | 217/318 [00:54<00:24,  4.11it/s]\u001b[A\n",
            " 69%|██████▊   | 218/318 [00:54<00:23,  4.21it/s]\u001b[A\n",
            " 69%|██████▉   | 219/318 [00:54<00:25,  3.88it/s]\u001b[A\n",
            " 69%|██████▉   | 220/318 [00:55<00:24,  4.06it/s]\u001b[A\n",
            " 69%|██████▉   | 221/318 [00:55<00:23,  4.16it/s]\u001b[A\n",
            " 70%|██████▉   | 222/318 [00:55<00:22,  4.21it/s]\u001b[A\n",
            " 70%|███████   | 223/318 [00:55<00:22,  4.25it/s]\u001b[A\n",
            " 70%|███████   | 224/318 [00:56<00:23,  4.01it/s]\u001b[A\n",
            " 71%|███████   | 225/318 [00:56<00:22,  4.12it/s]\u001b[A\n",
            " 71%|███████   | 226/318 [00:56<00:21,  4.20it/s]\u001b[A\n",
            " 71%|███████▏  | 227/318 [00:56<00:21,  4.20it/s]\u001b[A\n",
            " 72%|███████▏  | 228/318 [00:57<00:22,  3.95it/s]\u001b[A\n",
            " 72%|███████▏  | 229/318 [00:57<00:22,  3.92it/s]\u001b[A\n",
            " 72%|███████▏  | 230/318 [00:57<00:22,  3.87it/s]\u001b[A\n",
            " 73%|███████▎  | 231/318 [00:57<00:22,  3.90it/s]\u001b[A\n",
            " 73%|███████▎  | 232/318 [00:58<00:21,  4.06it/s]\u001b[A\n",
            " 73%|███████▎  | 233/318 [00:58<00:22,  3.85it/s]\u001b[A\n",
            " 74%|███████▎  | 234/318 [00:58<00:20,  4.19it/s]\u001b[A\n",
            " 74%|███████▍  | 235/318 [00:58<00:18,  4.40it/s]\u001b[A\n",
            " 74%|███████▍  | 236/318 [00:58<00:18,  4.37it/s]\u001b[A\n",
            " 75%|███████▍  | 237/318 [00:59<00:19,  4.22it/s]\u001b[A\n",
            " 75%|███████▍  | 238/318 [00:59<00:19,  4.17it/s]\u001b[A\n",
            " 75%|███████▌  | 239/318 [00:59<00:19,  4.08it/s]\u001b[A\n",
            " 75%|███████▌  | 240/318 [00:59<00:18,  4.26it/s]\u001b[A\n",
            " 76%|███████▌  | 241/318 [01:00<00:17,  4.41it/s]\u001b[A\n",
            " 76%|███████▌  | 242/318 [01:00<00:16,  4.50it/s]\u001b[A\n",
            " 76%|███████▋  | 243/318 [01:00<00:17,  4.32it/s]\u001b[A\n",
            " 77%|███████▋  | 244/318 [01:00<00:19,  3.82it/s]\u001b[A\n",
            " 77%|███████▋  | 245/318 [01:01<00:17,  4.14it/s]\u001b[A\n",
            " 77%|███████▋  | 246/318 [01:01<00:17,  4.05it/s]\u001b[A\n",
            " 78%|███████▊  | 247/318 [01:01<00:17,  4.09it/s]\u001b[A\n",
            " 78%|███████▊  | 248/318 [01:01<00:16,  4.18it/s]\u001b[A\n",
            " 78%|███████▊  | 249/318 [01:02<00:16,  4.15it/s]\u001b[A\n",
            " 79%|███████▊  | 250/318 [01:02<00:17,  4.00it/s]\u001b[A\n",
            " 79%|███████▉  | 251/318 [01:02<00:16,  3.98it/s]\u001b[A\n",
            " 79%|███████▉  | 252/318 [01:02<00:16,  3.94it/s]\u001b[A\n",
            " 80%|███████▉  | 253/318 [01:03<00:16,  3.97it/s]\u001b[A\n",
            " 80%|███████▉  | 254/318 [01:03<00:16,  3.95it/s]\u001b[A\n",
            " 80%|████████  | 255/318 [01:03<00:16,  3.86it/s]\u001b[A\n",
            " 81%|████████  | 256/318 [01:03<00:16,  3.79it/s]\u001b[A\n",
            " 81%|████████  | 257/318 [01:04<00:15,  3.82it/s]\u001b[A\n",
            " 81%|████████  | 258/318 [01:04<00:15,  3.89it/s]\u001b[A\n",
            " 81%|████████▏ | 259/318 [01:04<00:15,  3.88it/s]\u001b[A\n",
            " 82%|████████▏ | 260/318 [01:04<00:14,  3.98it/s]\u001b[A\n",
            " 82%|████████▏ | 261/318 [01:05<00:13,  4.21it/s]\u001b[A\n",
            " 82%|████████▏ | 262/318 [01:05<00:13,  4.11it/s]\u001b[A\n",
            " 83%|████████▎ | 263/318 [01:05<00:12,  4.32it/s]\u001b[A\n",
            " 83%|████████▎ | 264/318 [01:05<00:12,  4.17it/s]\u001b[A\n",
            " 83%|████████▎ | 265/318 [01:06<00:13,  3.93it/s]\u001b[A\n",
            " 84%|████████▎ | 266/318 [01:06<00:14,  3.69it/s]\u001b[A\n",
            " 84%|████████▍ | 267/318 [01:06<00:13,  3.90it/s]\u001b[A\n",
            " 84%|████████▍ | 268/318 [01:06<00:12,  4.06it/s]\u001b[A\n",
            " 85%|████████▍ | 269/318 [01:07<00:11,  4.19it/s]\u001b[A\n",
            " 85%|████████▍ | 270/318 [01:07<00:11,  4.11it/s]\u001b[A\n",
            " 85%|████████▌ | 271/318 [01:07<00:11,  4.19it/s]\u001b[A\n",
            " 86%|████████▌ | 272/318 [01:07<00:11,  4.01it/s]\u001b[A\n",
            " 86%|████████▌ | 273/318 [01:08<00:10,  4.12it/s]\u001b[A\n",
            " 86%|████████▌ | 274/318 [01:08<00:10,  4.22it/s]\u001b[A\n",
            " 86%|████████▋ | 275/318 [01:08<00:10,  3.93it/s]\u001b[A\n",
            " 87%|████████▋ | 276/318 [01:08<00:10,  3.83it/s]\u001b[A\n",
            " 87%|████████▋ | 277/318 [01:09<00:10,  3.99it/s]\u001b[A\n",
            " 87%|████████▋ | 278/318 [01:09<00:09,  4.27it/s]\u001b[A\n",
            " 88%|████████▊ | 279/318 [01:09<00:09,  4.32it/s]\u001b[A\n",
            " 88%|████████▊ | 280/318 [01:09<00:09,  4.04it/s]\u001b[A\n",
            " 88%|████████▊ | 281/318 [01:10<00:09,  4.05it/s]\u001b[A\n",
            " 89%|████████▊ | 282/318 [01:10<00:08,  4.11it/s]\u001b[A\n",
            " 89%|████████▉ | 283/318 [01:10<00:08,  4.11it/s]\u001b[A\n",
            " 89%|████████▉ | 284/318 [01:10<00:08,  4.22it/s]\u001b[A\n",
            " 90%|████████▉ | 285/318 [01:11<00:07,  4.16it/s]\u001b[A\n",
            " 90%|████████▉ | 286/318 [01:11<00:07,  4.10it/s]\u001b[A\n",
            " 90%|█████████ | 287/318 [01:11<00:07,  3.97it/s]\u001b[A\n",
            " 91%|█████████ | 288/318 [01:11<00:07,  4.06it/s]\u001b[A\n",
            " 91%|█████████ | 289/318 [01:12<00:07,  3.88it/s]\u001b[A\n",
            " 91%|█████████ | 290/318 [01:12<00:08,  3.48it/s]\u001b[A\n",
            " 92%|█████████▏| 291/318 [01:12<00:07,  3.68it/s]\u001b[A\n",
            " 92%|█████████▏| 292/318 [01:12<00:06,  3.82it/s]\u001b[A\n",
            " 92%|█████████▏| 293/318 [01:13<00:06,  3.99it/s]\u001b[A\n",
            " 92%|█████████▏| 294/318 [01:13<00:05,  4.12it/s]\u001b[A\n",
            " 93%|█████████▎| 295/318 [01:13<00:05,  4.00it/s]\u001b[A\n",
            " 93%|█████████▎| 296/318 [01:13<00:05,  4.02it/s]\u001b[A\n",
            " 93%|█████████▎| 297/318 [01:14<00:05,  4.14it/s]\u001b[A\n",
            " 94%|█████████▎| 298/318 [01:14<00:04,  4.31it/s]\u001b[A\n",
            " 94%|█████████▍| 299/318 [01:14<00:04,  3.99it/s]\u001b[A\n",
            " 94%|█████████▍| 300/318 [01:14<00:04,  4.20it/s]\u001b[A\n",
            " 95%|█████████▍| 301/318 [01:15<00:04,  4.19it/s]\u001b[A\n",
            " 95%|█████████▍| 302/318 [01:15<00:04,  3.92it/s]\u001b[A\n",
            " 95%|█████████▌| 303/318 [01:15<00:03,  4.05it/s]\u001b[A\n",
            " 96%|█████████▌| 304/318 [01:15<00:03,  3.91it/s]\u001b[A\n",
            " 96%|█████████▌| 305/318 [01:16<00:03,  3.83it/s]\u001b[A\n",
            " 96%|█████████▌| 306/318 [01:16<00:03,  3.78it/s]\u001b[A\n",
            " 97%|█████████▋| 307/318 [01:16<00:02,  3.83it/s]\u001b[A\n",
            " 97%|█████████▋| 308/318 [01:16<00:02,  3.89it/s]\u001b[A\n",
            " 97%|█████████▋| 309/318 [01:17<00:02,  4.06it/s]\u001b[A\n",
            " 97%|█████████▋| 310/318 [01:17<00:01,  4.05it/s]\u001b[A\n",
            " 98%|█████████▊| 311/318 [01:17<00:01,  4.15it/s]\u001b[A\n",
            " 98%|█████████▊| 312/318 [01:17<00:01,  4.08it/s]\u001b[A\n",
            " 98%|█████████▊| 313/318 [01:18<00:01,  3.87it/s]\u001b[A\n",
            " 99%|█████████▊| 314/318 [01:18<00:01,  3.99it/s]\u001b[A\n",
            " 99%|█████████▉| 315/318 [01:18<00:00,  4.08it/s]\u001b[A\n",
            " 99%|█████████▉| 316/318 [01:18<00:00,  4.10it/s]\u001b[A\n",
            "100%|█████████▉| 317/318 [01:19<00:00,  4.37it/s]\u001b[A\n",
            "100%|██████████| 318/318 [01:19<00:00,  4.01it/s]\n",
            " 80%|████████  | 8/10 [10:38<02:39, 79.57s/it]\n",
            "  0%|          | 0/318 [00:00<?, ?it/s]\u001b[A\n",
            "  0%|          | 1/318 [00:00<02:16,  2.32it/s]\u001b[A\n",
            "  1%|          | 2/318 [00:00<01:37,  3.23it/s]\u001b[A\n",
            "  1%|          | 3/318 [00:00<01:22,  3.80it/s]\u001b[A\n",
            "  1%|▏         | 4/318 [00:01<01:20,  3.89it/s]\u001b[A\n",
            "  2%|▏         | 5/318 [00:01<01:21,  3.84it/s]\u001b[A\n",
            "  2%|▏         | 6/318 [00:01<01:22,  3.76it/s]\u001b[A\n",
            "  2%|▏         | 7/318 [00:01<01:19,  3.92it/s]\u001b[A\n",
            "  3%|▎         | 8/318 [00:02<01:16,  4.04it/s]\u001b[A\n",
            "  3%|▎         | 9/318 [00:02<01:14,  4.14it/s]\u001b[A\n",
            "  3%|▎         | 10/318 [00:02<01:12,  4.25it/s]\u001b[A\n",
            "  3%|▎         | 11/318 [00:02<01:16,  4.01it/s]\u001b[A\n",
            "  4%|▍         | 12/318 [00:03<01:16,  4.02it/s]\u001b[A\n",
            "  4%|▍         | 13/318 [00:03<01:17,  3.93it/s]\u001b[A\n",
            "  4%|▍         | 14/318 [00:03<01:20,  3.80it/s]\u001b[A\n",
            "  5%|▍         | 15/318 [00:03<01:18,  3.86it/s]\u001b[A\n",
            "  5%|▌         | 16/318 [00:04<01:18,  3.87it/s]\u001b[A\n",
            "  5%|▌         | 17/318 [00:04<01:15,  4.01it/s]\u001b[A\n",
            "  6%|▌         | 18/318 [00:04<01:16,  3.93it/s]\u001b[A\n",
            "  6%|▌         | 19/318 [00:04<01:14,  4.01it/s]\u001b[A\n",
            "  6%|▋         | 20/318 [00:05<01:16,  3.90it/s]\u001b[A\n",
            "  7%|▋         | 21/318 [00:05<01:13,  4.04it/s]\u001b[A\n",
            "  7%|▋         | 22/318 [00:05<01:13,  4.02it/s]\u001b[A\n",
            "  7%|▋         | 23/318 [00:05<01:11,  4.10it/s]\u001b[A\n",
            "  8%|▊         | 24/318 [00:06<01:12,  4.04it/s]\u001b[A\n",
            "  8%|▊         | 25/318 [00:06<01:14,  3.92it/s]\u001b[A\n",
            "  8%|▊         | 26/318 [00:06<01:13,  3.99it/s]\u001b[A\n",
            "  8%|▊         | 27/318 [00:06<01:12,  4.01it/s]\u001b[A\n",
            "  9%|▉         | 28/318 [00:07<01:09,  4.17it/s]\u001b[A\n",
            "  9%|▉         | 29/318 [00:07<01:08,  4.22it/s]\u001b[A\n",
            "  9%|▉         | 30/318 [00:07<01:08,  4.18it/s]\u001b[A\n",
            " 10%|▉         | 31/318 [00:07<01:06,  4.31it/s]\u001b[A\n",
            " 10%|█         | 32/318 [00:08<01:06,  4.32it/s]\u001b[A\n",
            " 10%|█         | 33/318 [00:08<01:08,  4.17it/s]\u001b[A\n",
            " 11%|█         | 34/318 [00:08<01:05,  4.36it/s]\u001b[A\n",
            " 11%|█         | 35/318 [00:08<01:04,  4.39it/s]\u001b[A\n",
            " 11%|█▏        | 36/318 [00:08<01:05,  4.29it/s]\u001b[A\n",
            " 12%|█▏        | 37/318 [00:09<01:04,  4.36it/s]\u001b[A\n",
            " 12%|█▏        | 38/318 [00:09<01:06,  4.22it/s]\u001b[A\n",
            " 12%|█▏        | 39/318 [00:09<01:04,  4.32it/s]\u001b[A\n",
            " 13%|█▎        | 40/318 [00:09<01:01,  4.51it/s]\u001b[A\n",
            " 13%|█▎        | 41/318 [00:10<01:01,  4.50it/s]\u001b[A\n",
            " 13%|█▎        | 42/318 [00:10<01:00,  4.59it/s]\u001b[A\n",
            " 14%|█▎        | 43/318 [00:10<01:05,  4.18it/s]\u001b[A\n",
            " 14%|█▍        | 44/318 [00:10<01:14,  3.67it/s]\u001b[A\n",
            " 14%|█▍        | 45/318 [00:11<01:10,  3.87it/s]\u001b[A\n",
            " 14%|█▍        | 46/318 [00:11<01:08,  3.96it/s]\u001b[A\n",
            " 15%|█▍        | 47/318 [00:11<01:11,  3.77it/s]\u001b[A\n",
            " 15%|█▌        | 48/318 [00:11<01:11,  3.75it/s]\u001b[A\n",
            " 15%|█▌        | 49/318 [00:12<01:10,  3.84it/s]\u001b[A\n",
            " 16%|█▌        | 50/318 [00:12<01:08,  3.93it/s]\u001b[A\n",
            " 16%|█▌        | 51/318 [00:12<01:05,  4.06it/s]\u001b[A\n",
            " 16%|█▋        | 52/318 [00:12<01:06,  4.00it/s]\u001b[A\n",
            " 17%|█▋        | 53/318 [00:13<01:05,  4.06it/s]\u001b[A\n",
            " 17%|█▋        | 54/318 [00:13<01:03,  4.14it/s]\u001b[A\n",
            " 17%|█▋        | 55/318 [00:13<01:07,  3.92it/s]\u001b[A\n",
            " 18%|█▊        | 56/318 [00:13<01:07,  3.90it/s]\u001b[A\n",
            " 18%|█▊        | 57/318 [00:14<01:06,  3.92it/s]\u001b[A\n",
            " 18%|█▊        | 58/318 [00:14<01:06,  3.93it/s]\u001b[A\n",
            " 19%|█▊        | 59/318 [00:14<01:12,  3.58it/s]\u001b[A\n",
            " 19%|█▉        | 60/318 [00:15<01:09,  3.73it/s]\u001b[A\n",
            " 19%|█▉        | 61/318 [00:15<01:08,  3.77it/s]\u001b[A\n",
            " 19%|█▉        | 62/318 [00:15<01:07,  3.81it/s]\u001b[A\n",
            " 20%|█▉        | 63/318 [00:15<01:06,  3.82it/s]\u001b[A\n",
            " 20%|██        | 64/318 [00:16<01:05,  3.86it/s]\u001b[A\n",
            " 20%|██        | 65/318 [00:16<01:07,  3.77it/s]\u001b[A\n",
            " 21%|██        | 66/318 [00:16<01:10,  3.56it/s]\u001b[A\n",
            " 21%|██        | 67/318 [00:16<01:11,  3.52it/s]\u001b[A\n",
            " 21%|██▏       | 68/318 [00:17<01:11,  3.50it/s]\u001b[A\n",
            " 22%|██▏       | 69/318 [00:17<01:07,  3.69it/s]\u001b[A\n",
            " 22%|██▏       | 70/318 [00:17<01:06,  3.73it/s]\u001b[A\n",
            " 22%|██▏       | 71/318 [00:17<01:05,  3.76it/s]\u001b[A\n",
            " 23%|██▎       | 72/318 [00:18<01:00,  4.07it/s]\u001b[A\n",
            " 23%|██▎       | 73/318 [00:18<01:04,  3.82it/s]\u001b[A\n",
            " 23%|██▎       | 74/318 [00:18<01:04,  3.77it/s]\u001b[A\n",
            " 24%|██▎       | 75/318 [00:18<01:01,  3.92it/s]\u001b[A\n",
            " 24%|██▍       | 76/318 [00:19<00:59,  4.08it/s]\u001b[A\n",
            " 24%|██▍       | 77/318 [00:19<00:56,  4.24it/s]\u001b[A\n",
            " 25%|██▍       | 78/318 [00:19<00:56,  4.24it/s]\u001b[A\n",
            " 25%|██▍       | 79/318 [00:19<00:57,  4.17it/s]\u001b[A\n",
            " 25%|██▌       | 80/318 [00:20<00:58,  4.10it/s]\u001b[A\n",
            " 25%|██▌       | 81/318 [00:20<00:59,  4.02it/s]\u001b[A\n",
            " 26%|██▌       | 82/318 [00:20<01:01,  3.82it/s]\u001b[A\n",
            " 26%|██▌       | 83/318 [00:20<00:57,  4.09it/s]\u001b[A\n",
            " 26%|██▋       | 84/318 [00:21<00:56,  4.15it/s]\u001b[A\n",
            " 27%|██▋       | 85/318 [00:21<01:00,  3.84it/s]\u001b[A\n",
            " 27%|██▋       | 86/318 [00:21<00:57,  4.04it/s]\u001b[A\n",
            " 27%|██▋       | 87/318 [00:21<00:55,  4.18it/s]\u001b[A\n",
            " 28%|██▊       | 88/318 [00:22<00:55,  4.17it/s]\u001b[A\n",
            " 28%|██▊       | 89/318 [00:22<00:54,  4.17it/s]\u001b[A\n",
            " 28%|██▊       | 90/318 [00:22<00:55,  4.09it/s]\u001b[A\n",
            " 29%|██▊       | 91/318 [00:22<00:54,  4.16it/s]\u001b[A\n",
            " 29%|██▉       | 92/318 [00:23<00:56,  4.02it/s]\u001b[A\n",
            " 29%|██▉       | 93/318 [00:23<00:57,  3.93it/s]\u001b[A\n",
            " 30%|██▉       | 94/318 [00:23<00:57,  3.87it/s]\u001b[A\n",
            " 30%|██▉       | 95/318 [00:23<00:55,  4.01it/s]\u001b[A\n",
            " 30%|███       | 96/318 [00:24<00:52,  4.26it/s]\u001b[A\n",
            " 31%|███       | 97/318 [00:24<00:53,  4.14it/s]\u001b[A\n",
            " 31%|███       | 98/318 [00:24<00:55,  3.99it/s]\u001b[A\n",
            " 31%|███       | 99/318 [00:24<00:53,  4.11it/s]\u001b[A\n",
            " 31%|███▏      | 100/318 [00:25<00:53,  4.07it/s]\u001b[A\n",
            " 32%|███▏      | 101/318 [00:25<00:55,  3.94it/s]\u001b[A\n",
            " 32%|███▏      | 102/318 [00:25<00:52,  4.13it/s]\u001b[A\n",
            " 32%|███▏      | 103/318 [00:25<00:50,  4.22it/s]\u001b[A\n",
            " 33%|███▎      | 104/318 [00:26<00:53,  4.02it/s]\u001b[A\n",
            " 33%|███▎      | 105/318 [00:26<00:52,  4.09it/s]\u001b[A\n",
            " 33%|███▎      | 106/318 [00:26<00:51,  4.10it/s]\u001b[A\n",
            " 34%|███▎      | 107/318 [00:26<00:50,  4.19it/s]\u001b[A\n",
            " 34%|███▍      | 108/318 [00:27<00:49,  4.28it/s]\u001b[A\n",
            " 34%|███▍      | 109/318 [00:27<00:51,  4.03it/s]\u001b[A\n",
            " 35%|███▍      | 110/318 [00:27<00:51,  4.05it/s]\u001b[A\n",
            " 35%|███▍      | 111/318 [00:27<00:50,  4.09it/s]\u001b[A\n",
            " 35%|███▌      | 112/318 [00:28<00:53,  3.82it/s]\u001b[A\n",
            " 36%|███▌      | 113/318 [00:28<00:54,  3.76it/s]\u001b[A\n",
            " 36%|███▌      | 114/318 [00:28<00:50,  4.00it/s]\u001b[A\n",
            " 36%|███▌      | 115/318 [00:28<00:54,  3.74it/s]\u001b[A\n",
            " 36%|███▋      | 116/318 [00:29<00:52,  3.82it/s]\u001b[A\n",
            " 37%|███▋      | 117/318 [00:29<00:50,  4.00it/s]\u001b[A\n",
            " 37%|███▋      | 118/318 [00:29<00:51,  3.88it/s]\u001b[A\n",
            " 37%|███▋      | 119/318 [00:29<00:55,  3.58it/s]\u001b[A\n",
            " 38%|███▊      | 120/318 [00:30<00:52,  3.75it/s]\u001b[A\n",
            " 38%|███▊      | 121/318 [00:30<00:51,  3.86it/s]\u001b[A\n",
            " 38%|███▊      | 122/318 [00:30<00:48,  4.04it/s]\u001b[A\n",
            " 39%|███▊      | 123/318 [00:30<00:46,  4.17it/s]\u001b[A\n",
            " 39%|███▉      | 124/318 [00:31<00:46,  4.17it/s]\u001b[A\n",
            " 39%|███▉      | 125/318 [00:31<00:45,  4.20it/s]\u001b[A\n",
            " 40%|███▉      | 126/318 [00:31<00:52,  3.66it/s]\u001b[A\n",
            " 40%|███▉      | 127/318 [00:31<00:48,  3.92it/s]\u001b[A\n",
            " 40%|████      | 128/318 [00:32<00:44,  4.23it/s]\u001b[A\n",
            " 41%|████      | 129/318 [00:32<00:44,  4.27it/s]\u001b[A\n",
            " 41%|████      | 130/318 [00:32<00:47,  3.94it/s]\u001b[A\n",
            " 41%|████      | 131/318 [00:32<00:47,  3.98it/s]\u001b[A\n",
            " 42%|████▏     | 132/318 [00:33<00:46,  3.97it/s]\u001b[A\n",
            " 42%|████▏     | 133/318 [00:33<00:44,  4.16it/s]\u001b[A\n",
            " 42%|████▏     | 134/318 [00:33<00:46,  3.93it/s]\u001b[A\n",
            " 42%|████▏     | 135/318 [00:33<00:45,  4.01it/s]\u001b[A\n",
            " 43%|████▎     | 136/318 [00:34<00:43,  4.20it/s]\u001b[A\n",
            " 43%|████▎     | 137/318 [00:34<00:43,  4.19it/s]\u001b[A\n",
            " 43%|████▎     | 138/318 [00:34<00:43,  4.11it/s]\u001b[A\n",
            " 44%|████▎     | 139/318 [00:34<00:45,  3.97it/s]\u001b[A\n",
            " 44%|████▍     | 140/318 [00:35<00:43,  4.08it/s]\u001b[A\n",
            " 44%|████▍     | 141/318 [00:35<00:41,  4.24it/s]\u001b[A\n",
            " 45%|████▍     | 142/318 [00:35<00:38,  4.52it/s]\u001b[A\n",
            " 45%|████▍     | 143/318 [00:35<00:38,  4.60it/s]\u001b[A\n",
            " 45%|████▌     | 144/318 [00:35<00:41,  4.24it/s]\u001b[A\n",
            " 46%|████▌     | 145/318 [00:36<00:41,  4.17it/s]\u001b[A\n",
            " 46%|████▌     | 146/318 [00:36<00:42,  4.09it/s]\u001b[A\n",
            " 46%|████▌     | 147/318 [00:36<00:41,  4.16it/s]\u001b[A\n",
            " 47%|████▋     | 148/318 [00:36<00:39,  4.26it/s]\u001b[A\n",
            " 47%|████▋     | 149/318 [00:37<00:39,  4.23it/s]\u001b[A\n",
            " 47%|████▋     | 150/318 [00:37<00:40,  4.14it/s]\u001b[A\n",
            " 47%|████▋     | 151/318 [00:37<00:42,  3.92it/s]\u001b[A\n",
            " 48%|████▊     | 152/318 [00:37<00:43,  3.78it/s]\u001b[A\n",
            " 48%|████▊     | 153/318 [00:38<00:40,  4.08it/s]\u001b[A\n",
            " 48%|████▊     | 154/318 [00:38<00:40,  4.03it/s]\u001b[A\n",
            " 49%|████▊     | 155/318 [00:38<00:40,  4.05it/s]\u001b[A\n",
            " 49%|████▉     | 156/318 [00:38<00:40,  3.98it/s]\u001b[A\n",
            " 49%|████▉     | 157/318 [00:39<00:41,  3.86it/s]\u001b[A\n",
            " 50%|████▉     | 158/318 [00:39<00:39,  4.04it/s]\u001b[A\n",
            " 50%|█████     | 159/318 [00:39<00:39,  3.99it/s]\u001b[A\n",
            " 50%|█████     | 160/318 [00:39<00:38,  4.08it/s]\u001b[A\n",
            " 51%|█████     | 161/318 [00:40<00:39,  3.95it/s]\u001b[A\n",
            " 51%|█████     | 162/318 [00:40<00:39,  3.96it/s]\u001b[A\n",
            " 51%|█████▏    | 163/318 [00:40<00:40,  3.82it/s]\u001b[A\n",
            " 52%|█████▏    | 164/318 [00:40<00:38,  4.00it/s]\u001b[A\n",
            " 52%|█████▏    | 165/318 [00:41<00:39,  3.92it/s]\u001b[A\n",
            " 52%|█████▏    | 166/318 [00:41<00:38,  3.97it/s]\u001b[A\n",
            " 53%|█████▎    | 167/318 [00:41<00:38,  3.94it/s]\u001b[A\n",
            " 53%|█████▎    | 168/318 [00:41<00:36,  4.13it/s]\u001b[A\n",
            " 53%|█████▎    | 169/318 [00:42<00:36,  4.10it/s]\u001b[A\n",
            " 53%|█████▎    | 170/318 [00:42<00:34,  4.23it/s]\u001b[A\n",
            " 54%|█████▍    | 171/318 [00:42<00:34,  4.27it/s]\u001b[A\n",
            " 54%|█████▍    | 172/318 [00:42<00:34,  4.27it/s]\u001b[A\n",
            " 54%|█████▍    | 173/318 [00:43<00:37,  3.87it/s]\u001b[A\n",
            " 55%|█████▍    | 174/318 [00:43<00:35,  4.03it/s]\u001b[A\n",
            " 55%|█████▌    | 175/318 [00:43<00:38,  3.74it/s]\u001b[A\n",
            " 55%|█████▌    | 176/318 [00:43<00:37,  3.80it/s]\u001b[A\n",
            " 56%|█████▌    | 177/318 [00:44<00:39,  3.58it/s]\u001b[A\n",
            " 56%|█████▌    | 178/318 [00:44<00:35,  3.89it/s]\u001b[A\n",
            " 56%|█████▋    | 179/318 [00:44<00:37,  3.73it/s]\u001b[A\n",
            " 57%|█████▋    | 180/318 [00:45<00:37,  3.73it/s]\u001b[A\n",
            " 57%|█████▋    | 181/318 [00:45<00:35,  3.90it/s]\u001b[A\n",
            " 57%|█████▋    | 182/318 [00:45<00:33,  4.04it/s]\u001b[A\n",
            " 58%|█████▊    | 183/318 [00:45<00:34,  3.88it/s]\u001b[A\n",
            " 58%|█████▊    | 184/318 [00:46<00:34,  3.90it/s]\u001b[A\n",
            " 58%|█████▊    | 185/318 [00:46<00:33,  3.96it/s]\u001b[A\n",
            " 58%|█████▊    | 186/318 [00:46<00:31,  4.15it/s]\u001b[A\n",
            " 59%|█████▉    | 187/318 [00:46<00:32,  4.03it/s]\u001b[A\n",
            " 59%|█████▉    | 188/318 [00:47<00:31,  4.09it/s]\u001b[A\n",
            " 59%|█████▉    | 189/318 [00:47<00:32,  4.00it/s]\u001b[A\n",
            " 60%|█████▉    | 190/318 [00:47<00:30,  4.16it/s]\u001b[A\n",
            " 60%|██████    | 191/318 [00:47<00:30,  4.21it/s]\u001b[A\n",
            " 60%|██████    | 192/318 [00:47<00:30,  4.15it/s]\u001b[A\n",
            " 61%|██████    | 193/318 [00:48<00:30,  4.17it/s]\u001b[A\n",
            " 61%|██████    | 194/318 [00:48<00:31,  3.99it/s]\u001b[A\n",
            " 61%|██████▏   | 195/318 [00:48<00:30,  4.07it/s]\u001b[A\n",
            " 62%|██████▏   | 196/318 [00:48<00:30,  4.01it/s]\u001b[A\n",
            " 62%|██████▏   | 197/318 [00:49<00:30,  3.99it/s]\u001b[A\n",
            " 62%|██████▏   | 198/318 [00:49<00:33,  3.57it/s]\u001b[A\n",
            " 63%|██████▎   | 199/318 [00:49<00:31,  3.83it/s]\u001b[A\n",
            " 63%|██████▎   | 200/318 [00:50<00:32,  3.64it/s]\u001b[A\n",
            " 63%|██████▎   | 201/318 [00:50<00:32,  3.57it/s]\u001b[A\n",
            " 64%|██████▎   | 202/318 [00:50<00:30,  3.79it/s]\u001b[A\n",
            " 64%|██████▍   | 203/318 [00:50<00:30,  3.74it/s]\u001b[A\n",
            " 64%|██████▍   | 204/318 [00:51<00:28,  3.95it/s]\u001b[A\n",
            " 64%|██████▍   | 205/318 [00:51<00:27,  4.08it/s]\u001b[A\n",
            " 65%|██████▍   | 206/318 [00:51<00:30,  3.71it/s]\u001b[A\n",
            " 65%|██████▌   | 207/318 [00:51<00:29,  3.82it/s]\u001b[A\n",
            " 65%|██████▌   | 208/318 [00:52<00:29,  3.72it/s]\u001b[A\n",
            " 66%|██████▌   | 209/318 [00:52<00:29,  3.68it/s]\u001b[A\n",
            " 66%|██████▌   | 210/318 [00:52<00:28,  3.83it/s]\u001b[A\n",
            " 66%|██████▋   | 211/318 [00:53<00:28,  3.78it/s]\u001b[A\n",
            " 67%|██████▋   | 212/318 [00:53<00:28,  3.78it/s]\u001b[A\n",
            " 67%|██████▋   | 213/318 [00:53<00:27,  3.84it/s]\u001b[A\n",
            " 67%|██████▋   | 214/318 [00:53<00:26,  3.99it/s]\u001b[A\n",
            " 68%|██████▊   | 215/318 [00:54<00:26,  3.93it/s]\u001b[A\n",
            " 68%|██████▊   | 216/318 [00:54<00:25,  4.04it/s]\u001b[A\n",
            " 68%|██████▊   | 217/318 [00:54<00:25,  3.89it/s]\u001b[A\n",
            " 69%|██████▊   | 218/318 [00:54<00:25,  3.98it/s]\u001b[A\n",
            " 69%|██████▉   | 219/318 [00:55<00:24,  3.96it/s]\u001b[A\n",
            " 69%|██████▉   | 220/318 [00:55<00:24,  4.07it/s]\u001b[A\n",
            " 69%|██████▉   | 221/318 [00:55<00:25,  3.79it/s]\u001b[A\n",
            " 70%|██████▉   | 222/318 [00:55<00:25,  3.83it/s]\u001b[A\n",
            " 70%|███████   | 223/318 [00:56<00:23,  4.01it/s]\u001b[A\n",
            " 70%|███████   | 224/318 [00:56<00:24,  3.83it/s]\u001b[A\n",
            " 71%|███████   | 225/318 [00:56<00:23,  3.93it/s]\u001b[A\n",
            " 71%|███████   | 226/318 [00:56<00:22,  4.12it/s]\u001b[A\n",
            " 71%|███████▏  | 227/318 [00:57<00:23,  3.95it/s]\u001b[A\n",
            " 72%|███████▏  | 228/318 [00:57<00:24,  3.75it/s]\u001b[A\n",
            " 72%|███████▏  | 229/318 [00:57<00:23,  3.73it/s]\u001b[A\n",
            " 72%|███████▏  | 230/318 [00:57<00:23,  3.74it/s]\u001b[A\n",
            " 73%|███████▎  | 231/318 [00:58<00:24,  3.58it/s]\u001b[A\n",
            " 73%|███████▎  | 232/318 [00:58<00:24,  3.58it/s]\u001b[A\n",
            " 73%|███████▎  | 233/318 [00:58<00:23,  3.66it/s]\u001b[A\n",
            " 74%|███████▎  | 234/318 [00:58<00:21,  3.82it/s]\u001b[A\n",
            " 74%|███████▍  | 235/318 [00:59<00:21,  3.79it/s]\u001b[A\n",
            " 74%|███████▍  | 236/318 [00:59<00:20,  3.92it/s]\u001b[A\n",
            " 75%|███████▍  | 237/318 [00:59<00:20,  3.96it/s]\u001b[A\n",
            " 75%|███████▍  | 238/318 [00:59<00:20,  3.83it/s]\u001b[A\n",
            " 75%|███████▌  | 239/318 [01:00<00:20,  3.86it/s]\u001b[A\n",
            " 75%|███████▌  | 240/318 [01:00<00:18,  4.13it/s]\u001b[A\n",
            " 76%|███████▌  | 241/318 [01:00<00:19,  3.94it/s]\u001b[A\n",
            " 76%|███████▌  | 242/318 [01:01<00:20,  3.77it/s]\u001b[A\n",
            " 76%|███████▋  | 243/318 [01:01<00:19,  3.94it/s]\u001b[A\n",
            " 77%|███████▋  | 244/318 [01:01<00:19,  3.77it/s]\u001b[A\n",
            " 77%|███████▋  | 245/318 [01:01<00:19,  3.81it/s]\u001b[A\n",
            " 77%|███████▋  | 246/318 [01:02<00:18,  3.79it/s]\u001b[A\n",
            " 78%|███████▊  | 247/318 [01:02<00:19,  3.73it/s]\u001b[A\n",
            " 78%|███████▊  | 248/318 [01:02<00:19,  3.66it/s]\u001b[A\n",
            " 78%|███████▊  | 249/318 [01:02<00:18,  3.72it/s]\u001b[A\n",
            " 79%|███████▊  | 250/318 [01:03<00:18,  3.63it/s]\u001b[A\n",
            " 79%|███████▉  | 251/318 [01:03<00:17,  3.86it/s]\u001b[A\n",
            " 79%|███████▉  | 252/318 [01:03<00:17,  3.87it/s]\u001b[A\n",
            " 80%|███████▉  | 253/318 [01:03<00:16,  3.86it/s]\u001b[A\n",
            " 80%|███████▉  | 254/318 [01:04<00:15,  4.03it/s]\u001b[A\n",
            " 80%|████████  | 255/318 [01:04<00:15,  4.10it/s]\u001b[A\n",
            " 81%|████████  | 256/318 [01:04<00:14,  4.20it/s]\u001b[A\n",
            " 81%|████████  | 257/318 [01:04<00:14,  4.26it/s]\u001b[A\n",
            " 81%|████████  | 258/318 [01:05<00:14,  4.08it/s]\u001b[A\n",
            " 81%|████████▏ | 259/318 [01:05<00:14,  4.01it/s]\u001b[A\n",
            " 82%|████████▏ | 260/318 [01:05<00:15,  3.73it/s]\u001b[A\n",
            " 82%|████████▏ | 261/318 [01:05<00:15,  3.64it/s]\u001b[A\n",
            " 82%|████████▏ | 262/318 [01:06<00:15,  3.70it/s]\u001b[A\n",
            " 83%|████████▎ | 263/318 [01:06<00:14,  3.83it/s]\u001b[A\n",
            " 83%|████████▎ | 264/318 [01:06<00:13,  4.02it/s]\u001b[A\n",
            " 83%|████████▎ | 265/318 [01:06<00:12,  4.08it/s]\u001b[A\n",
            " 84%|████████▎ | 266/318 [01:07<00:12,  4.28it/s]\u001b[A\n",
            " 84%|████████▍ | 267/318 [01:07<00:11,  4.40it/s]\u001b[A\n",
            " 84%|████████▍ | 268/318 [01:07<00:11,  4.34it/s]\u001b[A\n",
            " 85%|████████▍ | 269/318 [01:07<00:11,  4.29it/s]\u001b[A\n",
            " 85%|████████▍ | 270/318 [01:08<00:11,  4.17it/s]\u001b[A\n",
            " 85%|████████▌ | 271/318 [01:08<00:11,  4.12it/s]\u001b[A\n",
            " 86%|████████▌ | 272/318 [01:08<00:10,  4.25it/s]\u001b[A\n",
            " 86%|████████▌ | 273/318 [01:08<00:11,  4.04it/s]\u001b[A\n",
            " 86%|████████▌ | 274/318 [01:09<00:10,  4.01it/s]\u001b[A\n",
            " 86%|████████▋ | 275/318 [01:09<00:09,  4.34it/s]\u001b[A\n",
            " 87%|████████▋ | 276/318 [01:09<00:10,  4.05it/s]\u001b[A\n",
            " 87%|████████▋ | 277/318 [01:09<00:09,  4.25it/s]\u001b[A\n",
            " 87%|████████▋ | 278/318 [01:09<00:09,  4.27it/s]\u001b[A\n",
            " 88%|████████▊ | 279/318 [01:10<00:09,  4.30it/s]\u001b[A\n",
            " 88%|████████▊ | 280/318 [01:10<00:09,  4.13it/s]\u001b[A\n",
            " 88%|████████▊ | 281/318 [01:10<00:09,  3.97it/s]\u001b[A\n",
            " 89%|████████▊ | 282/318 [01:10<00:08,  4.15it/s]\u001b[A\n",
            " 89%|████████▉ | 283/318 [01:11<00:08,  3.98it/s]\u001b[A\n",
            " 89%|████████▉ | 284/318 [01:11<00:08,  3.90it/s]\u001b[A\n",
            " 90%|████████▉ | 285/318 [01:11<00:08,  4.05it/s]\u001b[A\n",
            " 90%|████████▉ | 286/318 [01:11<00:07,  4.04it/s]\u001b[A\n",
            " 90%|█████████ | 287/318 [01:12<00:07,  4.00it/s]\u001b[A\n",
            " 91%|█████████ | 288/318 [01:12<00:07,  3.89it/s]\u001b[A\n",
            " 91%|█████████ | 289/318 [01:12<00:06,  4.18it/s]\u001b[A\n",
            " 91%|█████████ | 290/318 [01:12<00:06,  4.46it/s]\u001b[A\n",
            " 92%|█████████▏| 291/318 [01:13<00:06,  4.41it/s]\u001b[A\n",
            " 92%|█████████▏| 292/318 [01:13<00:06,  4.07it/s]\u001b[A\n",
            " 92%|█████████▏| 293/318 [01:13<00:06,  3.96it/s]\u001b[A\n",
            " 92%|█████████▏| 294/318 [01:13<00:05,  4.06it/s]\u001b[A\n",
            " 93%|█████████▎| 295/318 [01:14<00:05,  4.00it/s]\u001b[A\n",
            " 93%|█████████▎| 296/318 [01:14<00:05,  4.02it/s]\u001b[A\n",
            " 93%|█████████▎| 297/318 [01:14<00:05,  3.89it/s]\u001b[A\n",
            " 94%|█████████▎| 298/318 [01:14<00:05,  3.89it/s]\u001b[A\n",
            " 94%|█████████▍| 299/318 [01:15<00:05,  3.80it/s]\u001b[A\n",
            " 94%|█████████▍| 300/318 [01:15<00:04,  3.80it/s]\u001b[A\n",
            " 95%|█████████▍| 301/318 [01:15<00:04,  3.66it/s]\u001b[A\n",
            " 95%|█████████▍| 302/318 [01:16<00:04,  3.66it/s]\u001b[A\n",
            " 95%|█████████▌| 303/318 [01:16<00:03,  3.81it/s]\u001b[A\n",
            " 96%|█████████▌| 304/318 [01:16<00:03,  3.70it/s]\u001b[A\n",
            " 96%|█████████▌| 305/318 [01:16<00:03,  3.80it/s]\u001b[A\n",
            " 96%|█████████▌| 306/318 [01:17<00:03,  3.62it/s]\u001b[A\n",
            " 97%|█████████▋| 307/318 [01:17<00:02,  3.81it/s]\u001b[A\n",
            " 97%|█████████▋| 308/318 [01:17<00:02,  3.79it/s]\u001b[A\n",
            " 97%|█████████▋| 309/318 [01:17<00:02,  3.89it/s]\u001b[A\n",
            " 97%|█████████▋| 310/318 [01:18<00:02,  3.93it/s]\u001b[A\n",
            " 98%|█████████▊| 311/318 [01:18<00:01,  4.03it/s]\u001b[A\n",
            " 98%|█████████▊| 312/318 [01:18<00:01,  4.31it/s]\u001b[A\n",
            " 98%|█████████▊| 313/318 [01:18<00:01,  4.32it/s]\u001b[A\n",
            " 99%|█████████▊| 314/318 [01:19<00:01,  3.99it/s]\u001b[A\n",
            " 99%|█████████▉| 315/318 [01:19<00:00,  4.03it/s]\u001b[A\n",
            " 99%|█████████▉| 316/318 [01:19<00:00,  3.86it/s]\u001b[A\n",
            "100%|█████████▉| 317/318 [01:19<00:00,  4.08it/s]\u001b[A\n",
            "100%|██████████| 318/318 [01:20<00:00,  3.97it/s]\n",
            " 90%|█████████ | 9/10 [11:58<01:19, 79.75s/it]\n",
            "  0%|          | 0/318 [00:00<?, ?it/s]\u001b[A\n",
            "  0%|          | 1/318 [00:00<02:05,  2.53it/s]\u001b[A\n",
            "  1%|          | 2/318 [00:00<01:30,  3.51it/s]\u001b[A\n",
            "  1%|          | 3/318 [00:00<01:22,  3.84it/s]\u001b[A\n",
            "  1%|▏         | 4/318 [00:01<01:21,  3.86it/s]\u001b[A\n",
            "  2%|▏         | 5/318 [00:01<01:24,  3.69it/s]\u001b[A\n",
            "  2%|▏         | 6/318 [00:01<01:29,  3.50it/s]\u001b[A\n",
            "  2%|▏         | 7/318 [00:01<01:23,  3.71it/s]\u001b[A\n",
            "  3%|▎         | 8/318 [00:02<01:19,  3.88it/s]\u001b[A\n",
            "  3%|▎         | 9/318 [00:02<01:22,  3.75it/s]\u001b[A\n",
            "  3%|▎         | 10/318 [00:02<01:22,  3.71it/s]\u001b[A\n",
            "  3%|▎         | 11/318 [00:02<01:19,  3.87it/s]\u001b[A\n",
            "  4%|▍         | 12/318 [00:03<01:16,  3.98it/s]\u001b[A\n",
            "  4%|▍         | 13/318 [00:03<01:19,  3.84it/s]\u001b[A\n",
            "  4%|▍         | 14/318 [00:03<01:20,  3.78it/s]\u001b[A\n",
            "  5%|▍         | 15/318 [00:04<01:19,  3.82it/s]\u001b[A\n",
            "  5%|▌         | 16/318 [00:04<01:17,  3.90it/s]\u001b[A\n",
            "  5%|▌         | 17/318 [00:04<01:16,  3.95it/s]\u001b[A\n",
            "  6%|▌         | 18/318 [00:04<01:14,  4.02it/s]\u001b[A\n",
            "  6%|▌         | 19/318 [00:04<01:15,  3.95it/s]\u001b[A\n",
            "  6%|▋         | 20/318 [00:05<01:16,  3.88it/s]\u001b[A\n",
            "  7%|▋         | 21/318 [00:05<01:15,  3.91it/s]\u001b[A\n",
            "  7%|▋         | 22/318 [00:05<01:15,  3.93it/s]\u001b[A\n",
            "  7%|▋         | 23/318 [00:05<01:12,  4.07it/s]\u001b[A\n",
            "  8%|▊         | 24/318 [00:06<01:15,  3.90it/s]\u001b[A\n",
            "  8%|▊         | 25/318 [00:06<01:19,  3.70it/s]\u001b[A\n",
            "  8%|▊         | 26/318 [00:06<01:13,  3.96it/s]\u001b[A\n",
            "  8%|▊         | 27/318 [00:07<01:13,  3.99it/s]\u001b[A\n",
            "  9%|▉         | 28/318 [00:07<01:14,  3.88it/s]\u001b[A\n",
            "  9%|▉         | 29/318 [00:07<01:12,  4.00it/s]\u001b[A\n",
            "  9%|▉         | 30/318 [00:07<01:09,  4.13it/s]\u001b[A\n",
            " 10%|▉         | 31/318 [00:08<01:10,  4.07it/s]\u001b[A\n",
            " 10%|█         | 32/318 [00:08<01:10,  4.03it/s]\u001b[A\n",
            " 10%|█         | 33/318 [00:08<01:13,  3.88it/s]\u001b[A\n",
            " 11%|█         | 34/318 [00:08<01:11,  3.97it/s]\u001b[A\n",
            " 11%|█         | 35/318 [00:09<01:14,  3.79it/s]\u001b[A\n",
            " 11%|█▏        | 36/318 [00:09<01:13,  3.85it/s]\u001b[A\n",
            " 12%|█▏        | 37/318 [00:09<01:09,  4.05it/s]\u001b[A\n",
            " 12%|█▏        | 38/318 [00:09<01:10,  3.98it/s]\u001b[A\n",
            " 12%|█▏        | 39/318 [00:10<01:08,  4.05it/s]\u001b[A\n",
            " 13%|█▎        | 40/318 [00:10<01:14,  3.72it/s]\u001b[A\n",
            " 13%|█▎        | 41/318 [00:10<01:13,  3.78it/s]\u001b[A\n",
            " 13%|█▎        | 42/318 [00:10<01:13,  3.73it/s]\u001b[A\n",
            " 14%|█▎        | 43/318 [00:11<01:13,  3.72it/s]\u001b[A\n",
            " 14%|█▍        | 44/318 [00:11<01:08,  4.02it/s]\u001b[A\n",
            " 14%|█▍        | 45/318 [00:11<01:07,  4.04it/s]\u001b[A\n",
            " 14%|█▍        | 46/318 [00:11<01:09,  3.93it/s]\u001b[A\n",
            " 15%|█▍        | 47/318 [00:12<01:06,  4.05it/s]\u001b[A\n",
            " 15%|█▌        | 48/318 [00:12<01:02,  4.34it/s]\u001b[A\n",
            " 15%|█▌        | 49/318 [00:12<01:03,  4.23it/s]\u001b[A\n",
            " 16%|█▌        | 50/318 [00:12<01:00,  4.39it/s]\u001b[A\n",
            " 16%|█▌        | 51/318 [00:12<01:00,  4.43it/s]\u001b[A\n",
            " 16%|█▋        | 52/318 [00:13<01:01,  4.35it/s]\u001b[A\n",
            " 17%|█▋        | 53/318 [00:13<01:04,  4.13it/s]\u001b[A\n",
            " 17%|█▋        | 54/318 [00:13<01:05,  4.01it/s]\u001b[A\n",
            " 17%|█▋        | 55/318 [00:14<01:06,  3.96it/s]\u001b[A\n",
            " 18%|█▊        | 56/318 [00:14<01:04,  4.05it/s]\u001b[A\n",
            " 18%|█▊        | 57/318 [00:14<01:07,  3.86it/s]\u001b[A\n",
            " 18%|█▊        | 58/318 [00:14<01:05,  3.98it/s]\u001b[A\n",
            " 19%|█▊        | 59/318 [00:15<01:04,  4.03it/s]\u001b[A\n",
            " 19%|█▉        | 60/318 [00:15<01:01,  4.19it/s]\u001b[A\n",
            " 19%|█▉        | 61/318 [00:15<01:01,  4.20it/s]\u001b[A\n",
            " 19%|█▉        | 62/318 [00:15<01:00,  4.22it/s]\u001b[A\n",
            " 20%|█▉        | 63/318 [00:15<00:59,  4.30it/s]\u001b[A\n",
            " 20%|██        | 64/318 [00:16<00:59,  4.26it/s]\u001b[A\n",
            " 20%|██        | 65/318 [00:16<01:07,  3.76it/s]\u001b[A\n",
            " 21%|██        | 66/318 [00:16<01:05,  3.87it/s]\u001b[A\n",
            " 21%|██        | 67/318 [00:16<01:03,  3.93it/s]\u001b[A\n",
            " 21%|██▏       | 68/318 [00:17<01:07,  3.69it/s]\u001b[A\n",
            " 22%|██▏       | 69/318 [00:17<01:07,  3.67it/s]\u001b[A\n",
            " 22%|██▏       | 70/318 [00:17<01:11,  3.47it/s]\u001b[A\n",
            " 22%|██▏       | 71/318 [00:18<01:07,  3.64it/s]\u001b[A\n",
            " 23%|██▎       | 72/318 [00:18<01:08,  3.59it/s]\u001b[A\n",
            " 23%|██▎       | 73/318 [00:18<01:02,  3.93it/s]\u001b[A\n",
            " 23%|██▎       | 74/318 [00:18<01:04,  3.81it/s]\u001b[A\n",
            " 24%|██▎       | 75/318 [00:19<00:59,  4.08it/s]\u001b[A\n",
            " 24%|██▍       | 76/318 [00:19<00:58,  4.11it/s]\u001b[A\n",
            " 24%|██▍       | 77/318 [00:19<00:59,  4.04it/s]\u001b[A\n",
            " 25%|██▍       | 78/318 [00:19<00:56,  4.23it/s]\u001b[A\n",
            " 25%|██▍       | 79/318 [00:20<01:03,  3.78it/s]\u001b[A\n",
            " 25%|██▌       | 80/318 [00:20<01:00,  3.94it/s]\u001b[A\n",
            " 25%|██▌       | 81/318 [00:20<00:59,  3.99it/s]\u001b[A\n",
            " 26%|██▌       | 82/318 [00:20<01:01,  3.83it/s]\u001b[A\n",
            " 26%|██▌       | 83/318 [00:21<01:03,  3.70it/s]\u001b[A\n",
            " 26%|██▋       | 84/318 [00:21<01:01,  3.77it/s]\u001b[A\n",
            " 27%|██▋       | 85/318 [00:21<01:03,  3.69it/s]\u001b[A\n",
            " 27%|██▋       | 86/318 [00:22<01:03,  3.65it/s]\u001b[A\n",
            " 27%|██▋       | 87/318 [00:22<01:02,  3.69it/s]\u001b[A\n",
            " 28%|██▊       | 88/318 [00:22<01:01,  3.75it/s]\u001b[A\n",
            " 28%|██▊       | 89/318 [00:22<00:56,  4.07it/s]\u001b[A\n",
            " 28%|██▊       | 90/318 [00:22<00:53,  4.28it/s]\u001b[A\n",
            " 29%|██▊       | 91/318 [00:23<00:52,  4.33it/s]\u001b[A\n",
            " 29%|██▉       | 92/318 [00:23<00:53,  4.19it/s]\u001b[A\n",
            " 29%|██▉       | 93/318 [00:23<00:50,  4.48it/s]\u001b[A\n",
            " 30%|██▉       | 94/318 [00:23<00:52,  4.24it/s]\u001b[A\n",
            " 30%|██▉       | 95/318 [00:24<00:52,  4.25it/s]\u001b[A\n",
            " 30%|███       | 96/318 [00:24<00:53,  4.16it/s]\u001b[A\n",
            " 31%|███       | 97/318 [00:24<00:50,  4.34it/s]\u001b[A\n",
            " 31%|███       | 98/318 [00:24<00:53,  4.08it/s]\u001b[A\n",
            " 31%|███       | 99/318 [00:25<00:55,  3.97it/s]\u001b[A\n",
            " 31%|███▏      | 100/318 [00:25<00:54,  4.00it/s]\u001b[A\n",
            " 32%|███▏      | 101/318 [00:25<00:50,  4.26it/s]\u001b[A\n",
            " 32%|███▏      | 102/318 [00:25<00:54,  4.00it/s]\u001b[A\n",
            " 32%|███▏      | 103/318 [00:26<00:49,  4.32it/s]\u001b[A\n",
            " 33%|███▎      | 104/318 [00:26<00:49,  4.35it/s]\u001b[A\n",
            " 33%|███▎      | 105/318 [00:26<00:49,  4.28it/s]\u001b[A\n",
            " 33%|███▎      | 106/318 [00:26<00:53,  4.00it/s]\u001b[A\n",
            " 34%|███▎      | 107/318 [00:27<00:51,  4.09it/s]\u001b[A\n",
            " 34%|███▍      | 108/318 [00:27<00:50,  4.17it/s]\u001b[A\n",
            " 34%|███▍      | 109/318 [00:27<00:49,  4.26it/s]\u001b[A\n",
            " 35%|███▍      | 110/318 [00:27<00:47,  4.35it/s]\u001b[A\n",
            " 35%|███▍      | 111/318 [00:27<00:48,  4.26it/s]\u001b[A\n",
            " 35%|███▌      | 112/318 [00:28<00:48,  4.27it/s]\u001b[A\n",
            " 36%|███▌      | 113/318 [00:28<00:46,  4.39it/s]\u001b[A\n",
            " 36%|███▌      | 114/318 [00:28<00:47,  4.34it/s]\u001b[A\n",
            " 36%|███▌      | 115/318 [00:28<00:47,  4.26it/s]\u001b[A\n",
            " 36%|███▋      | 116/318 [00:29<00:48,  4.21it/s]\u001b[A\n",
            " 37%|███▋      | 117/318 [00:29<00:49,  4.09it/s]\u001b[A\n",
            " 37%|███▋      | 118/318 [00:29<00:54,  3.69it/s]\u001b[A\n",
            " 37%|███▋      | 119/318 [00:29<00:51,  3.85it/s]\u001b[A\n",
            " 38%|███▊      | 120/318 [00:30<00:50,  3.93it/s]\u001b[A\n",
            " 38%|███▊      | 121/318 [00:30<00:47,  4.13it/s]\u001b[A\n",
            " 38%|███▊      | 122/318 [00:30<00:49,  3.99it/s]\u001b[A\n",
            " 39%|███▊      | 123/318 [00:30<00:49,  3.96it/s]\u001b[A\n",
            " 39%|███▉      | 124/318 [00:31<00:46,  4.21it/s]\u001b[A\n",
            " 39%|███▉      | 125/318 [00:31<00:47,  4.09it/s]\u001b[A\n",
            " 40%|███▉      | 126/318 [00:31<00:48,  3.99it/s]\u001b[A\n",
            " 40%|███▉      | 127/318 [00:31<00:47,  4.02it/s]\u001b[A\n",
            " 40%|████      | 128/318 [00:32<00:45,  4.16it/s]\u001b[A\n",
            " 41%|████      | 129/318 [00:32<00:46,  4.08it/s]\u001b[A\n",
            " 41%|████      | 130/318 [00:32<00:43,  4.28it/s]\u001b[A\n",
            " 41%|████      | 131/318 [00:32<00:47,  3.97it/s]\u001b[A\n",
            " 42%|████▏     | 132/318 [00:33<00:46,  4.00it/s]\u001b[A\n",
            " 42%|████▏     | 133/318 [00:33<00:47,  3.88it/s]\u001b[A\n",
            " 42%|████▏     | 134/318 [00:33<00:44,  4.13it/s]\u001b[A\n",
            " 42%|████▏     | 135/318 [00:33<00:43,  4.23it/s]\u001b[A\n",
            " 43%|████▎     | 136/318 [00:34<00:43,  4.16it/s]\u001b[A\n",
            " 43%|████▎     | 137/318 [00:34<00:45,  3.94it/s]\u001b[A\n",
            " 43%|████▎     | 138/318 [00:34<00:46,  3.83it/s]\u001b[A\n",
            " 44%|████▎     | 139/318 [00:34<00:48,  3.68it/s]\u001b[A\n",
            " 44%|████▍     | 140/318 [00:35<00:44,  3.99it/s]\u001b[A\n",
            " 44%|████▍     | 141/318 [00:35<00:42,  4.21it/s]\u001b[A\n",
            " 45%|████▍     | 142/318 [00:35<00:42,  4.13it/s]\u001b[A\n",
            " 45%|████▍     | 143/318 [00:35<00:42,  4.12it/s]\u001b[A\n",
            " 45%|████▌     | 144/318 [00:36<00:40,  4.28it/s]\u001b[A\n",
            " 46%|████▌     | 145/318 [00:36<00:44,  3.92it/s]\u001b[A\n",
            " 46%|████▌     | 146/318 [00:36<00:43,  3.94it/s]\u001b[A\n",
            " 46%|████▌     | 147/318 [00:36<00:43,  3.94it/s]\u001b[A\n",
            " 47%|████▋     | 148/318 [00:37<00:44,  3.85it/s]\u001b[A\n",
            " 47%|████▋     | 149/318 [00:37<00:40,  4.18it/s]\u001b[A\n",
            " 47%|████▋     | 150/318 [00:37<00:42,  3.92it/s]\u001b[A\n",
            " 47%|████▋     | 151/318 [00:37<00:41,  4.00it/s]\u001b[A\n",
            " 48%|████▊     | 152/318 [00:38<00:41,  3.96it/s]\u001b[A\n",
            " 48%|████▊     | 153/318 [00:38<00:42,  3.88it/s]\u001b[A\n",
            " 48%|████▊     | 154/318 [00:38<00:41,  3.95it/s]\u001b[A\n",
            " 49%|████▊     | 155/318 [00:38<00:43,  3.77it/s]\u001b[A\n",
            " 49%|████▉     | 156/318 [00:39<00:41,  3.86it/s]\u001b[A\n",
            " 49%|████▉     | 157/318 [00:39<00:41,  3.89it/s]\u001b[A\n",
            " 50%|████▉     | 158/318 [00:39<00:43,  3.70it/s]\u001b[A\n",
            " 50%|█████     | 159/318 [00:39<00:41,  3.83it/s]\u001b[A\n",
            " 50%|█████     | 160/318 [00:40<00:41,  3.78it/s]\u001b[A\n",
            " 51%|█████     | 161/318 [00:40<00:41,  3.82it/s]\u001b[A\n",
            " 51%|█████     | 162/318 [00:40<00:40,  3.85it/s]\u001b[A\n",
            " 51%|█████▏    | 163/318 [00:41<00:41,  3.76it/s]\u001b[A\n",
            " 52%|█████▏    | 164/318 [00:41<00:39,  3.91it/s]\u001b[A\n",
            " 52%|█████▏    | 165/318 [00:41<00:41,  3.65it/s]\u001b[A\n",
            " 52%|█████▏    | 166/318 [00:41<00:39,  3.89it/s]\u001b[A\n",
            " 53%|█████▎    | 167/318 [00:42<00:36,  4.10it/s]\u001b[A\n",
            " 53%|█████▎    | 168/318 [00:42<00:35,  4.23it/s]\u001b[A\n",
            " 53%|█████▎    | 169/318 [00:42<00:36,  4.12it/s]\u001b[A\n",
            " 53%|█████▎    | 170/318 [00:42<00:38,  3.86it/s]\u001b[A\n",
            " 54%|█████▍    | 171/318 [00:43<00:36,  4.03it/s]\u001b[A\n",
            " 54%|█████▍    | 172/318 [00:43<00:34,  4.17it/s]\u001b[A\n",
            " 54%|█████▍    | 173/318 [00:43<00:34,  4.26it/s]\u001b[A\n",
            " 55%|█████▍    | 174/318 [00:43<00:34,  4.18it/s]\u001b[A\n",
            " 55%|█████▌    | 175/318 [00:43<00:36,  3.96it/s]\u001b[A\n",
            " 55%|█████▌    | 176/318 [00:44<00:38,  3.71it/s]\u001b[A\n",
            " 56%|█████▌    | 177/318 [00:44<00:38,  3.70it/s]\u001b[A\n",
            " 56%|█████▌    | 178/318 [00:44<00:38,  3.65it/s]\u001b[A\n",
            " 56%|█████▋    | 179/318 [00:45<00:38,  3.61it/s]\u001b[A\n",
            " 57%|█████▋    | 180/318 [00:45<00:37,  3.64it/s]\u001b[A\n",
            " 57%|█████▋    | 181/318 [00:45<00:38,  3.52it/s]\u001b[A\n",
            " 57%|█████▋    | 182/318 [00:45<00:37,  3.68it/s]\u001b[A\n",
            " 58%|█████▊    | 183/318 [00:46<00:35,  3.79it/s]\u001b[A\n",
            " 58%|█████▊    | 184/318 [00:46<00:35,  3.73it/s]\u001b[A\n",
            " 58%|█████▊    | 185/318 [00:46<00:35,  3.78it/s]\u001b[A\n",
            " 58%|█████▊    | 186/318 [00:47<00:35,  3.73it/s]\u001b[A\n",
            " 59%|█████▉    | 187/318 [00:47<00:34,  3.76it/s]\u001b[A\n",
            " 59%|█████▉    | 188/318 [00:47<00:33,  3.88it/s]\u001b[A\n",
            " 59%|█████▉    | 189/318 [00:47<00:32,  3.92it/s]\u001b[A\n",
            " 60%|█████▉    | 190/318 [00:48<00:32,  3.95it/s]\u001b[A\n",
            " 60%|██████    | 191/318 [00:48<00:33,  3.80it/s]\u001b[A\n",
            " 60%|██████    | 192/318 [00:48<00:32,  3.93it/s]\u001b[A\n",
            " 61%|██████    | 193/318 [00:48<00:33,  3.78it/s]\u001b[A\n",
            " 61%|██████    | 194/318 [00:49<00:31,  3.95it/s]\u001b[A\n",
            " 61%|██████▏   | 195/318 [00:49<00:31,  3.96it/s]\u001b[A\n",
            " 62%|██████▏   | 196/318 [00:49<00:29,  4.14it/s]\u001b[A\n",
            " 62%|██████▏   | 197/318 [00:49<00:32,  3.70it/s]\u001b[A\n",
            " 62%|██████▏   | 198/318 [00:50<00:30,  3.91it/s]\u001b[A\n",
            " 63%|██████▎   | 199/318 [00:50<00:30,  3.92it/s]\u001b[A\n",
            " 63%|██████▎   | 200/318 [00:50<00:30,  3.91it/s]\u001b[A\n",
            " 63%|██████▎   | 201/318 [00:50<00:30,  3.78it/s]\u001b[A\n",
            " 64%|██████▎   | 202/318 [00:51<00:32,  3.60it/s]\u001b[A\n",
            " 64%|██████▍   | 203/318 [00:51<00:30,  3.75it/s]\u001b[A\n",
            " 64%|██████▍   | 204/318 [00:51<00:29,  3.86it/s]\u001b[A\n",
            " 64%|██████▍   | 205/318 [00:51<00:29,  3.87it/s]\u001b[A\n",
            " 65%|██████▍   | 206/318 [00:52<00:26,  4.18it/s]\u001b[A\n",
            " 65%|██████▌   | 207/318 [00:52<00:27,  4.06it/s]\u001b[A\n",
            " 65%|██████▌   | 208/318 [00:52<00:25,  4.31it/s]\u001b[A\n",
            " 66%|██████▌   | 209/318 [00:52<00:26,  4.11it/s]\u001b[A\n",
            " 66%|██████▌   | 210/318 [00:53<00:24,  4.49it/s]\u001b[A\n",
            " 66%|██████▋   | 211/318 [00:53<00:23,  4.48it/s]\u001b[A\n",
            " 67%|██████▋   | 212/318 [00:53<00:23,  4.43it/s]\u001b[A\n",
            " 67%|██████▋   | 213/318 [00:53<00:23,  4.43it/s]\u001b[A\n",
            " 67%|██████▋   | 214/318 [00:53<00:23,  4.40it/s]\u001b[A\n",
            " 68%|██████▊   | 215/318 [00:54<00:26,  3.83it/s]\u001b[A\n",
            " 68%|██████▊   | 216/318 [00:54<00:26,  3.83it/s]\u001b[A\n",
            " 68%|██████▊   | 217/318 [00:54<00:25,  4.00it/s]\u001b[A\n",
            " 69%|██████▊   | 218/318 [00:55<00:25,  3.95it/s]\u001b[A\n",
            " 69%|██████▉   | 219/318 [00:55<00:24,  4.02it/s]\u001b[A\n",
            " 69%|██████▉   | 220/318 [00:55<00:23,  4.12it/s]\u001b[A\n",
            " 69%|██████▉   | 221/318 [00:55<00:22,  4.24it/s]\u001b[A\n",
            " 70%|██████▉   | 222/318 [00:55<00:22,  4.23it/s]\u001b[A\n",
            " 70%|███████   | 223/318 [00:56<00:24,  3.94it/s]\u001b[A\n",
            " 70%|███████   | 224/318 [00:56<00:23,  4.01it/s]\u001b[A\n",
            " 71%|███████   | 225/318 [00:56<00:23,  3.95it/s]\u001b[A\n",
            " 71%|███████   | 226/318 [00:56<00:22,  4.16it/s]\u001b[A\n",
            " 71%|███████▏  | 227/318 [00:57<00:22,  3.98it/s]\u001b[A\n",
            " 72%|███████▏  | 228/318 [00:57<00:21,  4.11it/s]\u001b[A\n",
            " 72%|███████▏  | 229/318 [00:57<00:20,  4.38it/s]\u001b[A\n",
            " 72%|███████▏  | 230/318 [00:57<00:19,  4.42it/s]\u001b[A\n",
            " 73%|███████▎  | 231/318 [00:58<00:20,  4.23it/s]\u001b[A\n",
            " 73%|███████▎  | 232/318 [00:58<00:21,  4.09it/s]\u001b[A\n",
            " 73%|███████▎  | 233/318 [00:58<00:20,  4.08it/s]\u001b[A\n",
            " 74%|███████▎  | 234/318 [00:58<00:19,  4.30it/s]\u001b[A\n",
            " 74%|███████▍  | 235/318 [00:59<00:19,  4.25it/s]\u001b[A\n",
            " 74%|███████▍  | 236/318 [00:59<00:19,  4.27it/s]\u001b[A\n",
            " 75%|███████▍  | 237/318 [00:59<00:18,  4.39it/s]\u001b[A\n",
            " 75%|███████▍  | 238/318 [00:59<00:18,  4.43it/s]\u001b[A\n",
            " 75%|███████▌  | 239/318 [00:59<00:17,  4.58it/s]\u001b[A\n",
            " 75%|███████▌  | 240/318 [01:00<00:17,  4.50it/s]\u001b[A\n",
            " 76%|███████▌  | 241/318 [01:00<00:17,  4.45it/s]\u001b[A\n",
            " 76%|███████▌  | 242/318 [01:00<00:17,  4.39it/s]\u001b[A\n",
            " 76%|███████▋  | 243/318 [01:00<00:17,  4.34it/s]\u001b[A\n",
            " 77%|███████▋  | 244/318 [01:01<00:17,  4.30it/s]\u001b[A\n",
            " 77%|███████▋  | 245/318 [01:01<00:17,  4.11it/s]\u001b[A\n",
            " 77%|███████▋  | 246/318 [01:01<00:16,  4.26it/s]\u001b[A\n",
            " 78%|███████▊  | 247/318 [01:01<00:16,  4.32it/s]\u001b[A\n",
            " 78%|███████▊  | 248/318 [01:02<00:16,  4.12it/s]\u001b[A\n",
            " 78%|███████▊  | 249/318 [01:02<00:17,  3.96it/s]\u001b[A\n",
            " 79%|███████▊  | 250/318 [01:02<00:17,  3.84it/s]\u001b[A\n",
            " 79%|███████▉  | 251/318 [01:02<00:16,  4.08it/s]\u001b[A\n",
            " 79%|███████▉  | 252/318 [01:03<00:15,  4.18it/s]\u001b[A\n",
            " 80%|███████▉  | 253/318 [01:03<00:15,  4.14it/s]\u001b[A\n",
            " 80%|███████▉  | 254/318 [01:03<00:16,  3.94it/s]\u001b[A\n",
            " 80%|████████  | 255/318 [01:03<00:17,  3.69it/s]\u001b[A\n",
            " 81%|████████  | 256/318 [01:04<00:15,  3.99it/s]\u001b[A\n",
            " 81%|████████  | 257/318 [01:04<00:14,  4.11it/s]\u001b[A\n",
            " 81%|████████  | 258/318 [01:04<00:14,  4.11it/s]\u001b[A\n",
            " 81%|████████▏ | 259/318 [01:04<00:14,  4.03it/s]\u001b[A\n",
            " 82%|████████▏ | 260/318 [01:05<00:14,  4.10it/s]\u001b[A\n",
            " 82%|████████▏ | 261/318 [01:05<00:14,  3.99it/s]\u001b[A\n",
            " 82%|████████▏ | 262/318 [01:05<00:13,  4.07it/s]\u001b[A\n",
            " 83%|████████▎ | 263/318 [01:05<00:13,  4.23it/s]\u001b[A\n",
            " 83%|████████▎ | 264/318 [01:06<00:12,  4.33it/s]\u001b[A\n",
            " 83%|████████▎ | 265/318 [01:06<00:12,  4.26it/s]\u001b[A\n",
            " 84%|████████▎ | 266/318 [01:06<00:12,  4.20it/s]\u001b[A\n",
            " 84%|████████▍ | 267/318 [01:06<00:12,  4.10it/s]\u001b[A\n",
            " 84%|████████▍ | 268/318 [01:06<00:11,  4.28it/s]\u001b[A\n",
            " 85%|████████▍ | 269/318 [01:07<00:11,  4.31it/s]\u001b[A\n",
            " 85%|████████▍ | 270/318 [01:07<00:11,  4.11it/s]\u001b[A\n",
            " 85%|████████▌ | 271/318 [01:07<00:11,  3.96it/s]\u001b[A\n",
            " 86%|████████▌ | 272/318 [01:07<00:11,  4.07it/s]\u001b[A\n",
            " 86%|████████▌ | 273/318 [01:08<00:11,  3.90it/s]\u001b[A\n",
            " 86%|████████▌ | 274/318 [01:08<00:10,  4.10it/s]\u001b[A\n",
            " 86%|████████▋ | 275/318 [01:08<00:10,  4.09it/s]\u001b[A\n",
            " 87%|████████▋ | 276/318 [01:08<00:10,  4.18it/s]\u001b[A\n",
            " 87%|████████▋ | 277/318 [01:09<00:09,  4.17it/s]\u001b[A\n",
            " 87%|████████▋ | 278/318 [01:09<00:10,  3.95it/s]\u001b[A\n",
            " 88%|████████▊ | 279/318 [01:09<00:09,  3.91it/s]\u001b[A\n",
            " 88%|████████▊ | 280/318 [01:09<00:09,  3.82it/s]\u001b[A\n",
            " 88%|████████▊ | 281/318 [01:10<00:09,  3.93it/s]\u001b[A\n",
            " 89%|████████▊ | 282/318 [01:10<00:09,  3.94it/s]\u001b[A\n",
            " 89%|████████▉ | 283/318 [01:10<00:09,  3.80it/s]\u001b[A\n",
            " 89%|████████▉ | 284/318 [01:10<00:08,  4.09it/s]\u001b[A\n",
            " 90%|████████▉ | 285/318 [01:11<00:07,  4.21it/s]\u001b[A\n",
            " 90%|████████▉ | 286/318 [01:11<00:07,  4.05it/s]\u001b[A\n",
            " 90%|█████████ | 287/318 [01:11<00:07,  4.14it/s]\u001b[A\n",
            " 91%|█████████ | 288/318 [01:11<00:07,  3.96it/s]\u001b[A\n",
            " 91%|█████████ | 289/318 [01:12<00:07,  3.76it/s]\u001b[A\n",
            " 91%|█████████ | 290/318 [01:12<00:07,  3.92it/s]\u001b[A\n",
            " 92%|█████████▏| 291/318 [01:12<00:06,  3.89it/s]\u001b[A\n",
            " 92%|█████████▏| 292/318 [01:13<00:06,  3.91it/s]\u001b[A\n",
            " 92%|█████████▏| 293/318 [01:13<00:06,  3.89it/s]\u001b[A\n",
            " 92%|█████████▏| 294/318 [01:13<00:06,  3.73it/s]\u001b[A\n",
            " 93%|█████████▎| 295/318 [01:13<00:05,  3.96it/s]\u001b[A\n",
            " 93%|█████████▎| 296/318 [01:14<00:06,  3.58it/s]\u001b[A\n",
            " 93%|█████████▎| 297/318 [01:14<00:05,  3.69it/s]\u001b[A\n",
            " 94%|█████████▎| 298/318 [01:14<00:05,  3.83it/s]\u001b[A\n",
            " 94%|█████████▍| 299/318 [01:14<00:04,  3.96it/s]\u001b[A\n",
            " 94%|█████████▍| 300/318 [01:15<00:04,  3.97it/s]\u001b[A\n",
            " 95%|█████████▍| 301/318 [01:15<00:04,  4.02it/s]\u001b[A\n",
            " 95%|█████████▍| 302/318 [01:15<00:03,  4.14it/s]\u001b[A\n",
            " 95%|█████████▌| 303/318 [01:15<00:03,  4.06it/s]\u001b[A\n",
            " 96%|█████████▌| 304/318 [01:16<00:03,  3.99it/s]\u001b[A\n",
            " 96%|█████████▌| 305/318 [01:16<00:03,  4.18it/s]\u001b[A\n",
            " 96%|█████████▌| 306/318 [01:16<00:02,  4.36it/s]\u001b[A\n",
            " 97%|█████████▋| 307/318 [01:16<00:02,  4.32it/s]\u001b[A\n",
            " 97%|█████████▋| 308/318 [01:16<00:02,  4.33it/s]\u001b[A\n",
            " 97%|█████████▋| 309/318 [01:17<00:02,  4.00it/s]\u001b[A\n",
            " 97%|█████████▋| 310/318 [01:17<00:02,  3.88it/s]\u001b[A\n",
            " 98%|█████████▊| 311/318 [01:17<00:01,  4.17it/s]\u001b[A\n",
            " 98%|█████████▊| 312/318 [01:17<00:01,  4.29it/s]\u001b[A\n",
            " 98%|█████████▊| 313/318 [01:18<00:01,  4.28it/s]\u001b[A\n",
            " 99%|█████████▊| 314/318 [01:18<00:00,  4.07it/s]\u001b[A\n",
            " 99%|█████████▉| 315/318 [01:18<00:00,  3.54it/s]\u001b[A\n",
            " 99%|█████████▉| 316/318 [01:19<00:00,  3.91it/s]\u001b[A\n",
            "100%|█████████▉| 317/318 [01:19<00:00,  4.05it/s]\u001b[A\n",
            "100%|██████████| 318/318 [01:19<00:00,  4.00it/s]\n",
            "100%|██████████| 10/10 [13:18<00:00, 79.82s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Saving...\n",
            "Done\n"
          ]
        }
      ],
      "source": [
        "if __name__ == '__main__':\n",
        "  # Load data\n",
        "  print('Load data...')\n",
        "  data = np.load(data_dir + 'data_arrays.npz', allow_pickle=True)\n",
        "  \n",
        "  print(data_dir)\n",
        "  # Training and validation data\n",
        "  if all_train:\n",
        "    trainloader, num_batches, pos_weight = get_trainloader(data, 'ALL')\n",
        "  else:\n",
        "    trainloader, num_batches, pos_weight = get_trainloader(data, 'TRAIN')\n",
        "  \n",
        "  # Vocabulary sizes\n",
        "  #num_static = num_static(data)\n",
        "  num_static = data['static_vars'].shape[0]\n",
        "  num_dp_codes, num_cp_codes = vocab_sizes(data)\n",
        "  \n",
        "  print('-----------------------------------------')\n",
        "  print('Train...')\n",
        "\n",
        "  # CUDA for PyTorch\n",
        "  use_cuda = torch.cuda.is_available()\n",
        "  #device = torch.device('cuda:0' if use_cuda else 'GPU:0') #changed to gpu\n",
        "  device = torch.device('cuda:0' if use_cuda else 'cpu') #changed to gpu\n",
        "  torch.backends.cudnn.benchmark = True\n",
        "\n",
        "  # Network\n",
        "  net = Net(num_static, num_dp_codes, num_cp_codes).to(device)\n",
        "\n",
        "  # Loss function and optimizer\n",
        "  criterion = nn.BCEWithLogitsLoss(pos_weight=pos_weight).to(device)\n",
        "  optimizer = optim.Adam(net.parameters(), lr = 0.001)  \n",
        "\n",
        "  # Create log dir\n",
        "  logdir_temp = logdir + net_variant + '/'\n",
        "\n",
        "  if not os.path.exists(logdir_temp):\n",
        "    os.makedirs(logdir_temp)\n",
        "  \n",
        "  # Store times\n",
        "  epoch_times = []\n",
        "\n",
        "  # Train\n",
        "  for epoch in tqdm(range(num_epochs)): \n",
        "    # print('-----------------------------------------')\n",
        "    # print('Epoch: {}'.format(epoch))\n",
        "    net.train()\n",
        "    time_start = time()\n",
        "    for i, (stat, dp, cp, dp_t, cp_t, label) in enumerate(tqdm(trainloader), 0):\n",
        "      # move to GPU if available\n",
        "      stat  = stat.to(device)\n",
        "      dp    = dp.to(device)\n",
        "      cp    = cp.to(device)\n",
        "      dp_t  = dp_t.to(device)\n",
        "      cp_t  = cp_t.to(device)\n",
        "      label = label.to(device)\n",
        "    \n",
        "      # zero the parameter gradients\n",
        "      optimizer.zero_grad()\n",
        "\n",
        "      # forward + backward + optimize\n",
        "      label_pred, _ = net(stat, dp, cp, dp_t, cp_t)\n",
        "      loss = criterion(label_pred, label)\n",
        "      loss.backward()\n",
        "      optimizer.step()\n",
        "    \n",
        "    # timing\n",
        "    time_end = time()\n",
        "    epoch_times.append(time_end-time_start)\n",
        "\n",
        "  # Save\n",
        "  print('Saving...')\n",
        "  torch.save(net.state_dict(), logdir_temp + 'final_model.pt')\n",
        "  np.savez(logdir_temp + 'epoch_times', epoch_times=epoch_times)\n",
        "  print('Done')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LAtDGNoLP-hK"
      },
      "source": [
        "Test"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "a-ZcwkxHP_D1"
      },
      "outputs": [],
      "source": [
        "'''\n",
        "Test.py\n",
        "Mar 2019 by Sebastiano Barbieri\n",
        "s.barbieri@unsw.edu.au\n",
        "'''\n",
        "\n",
        "def round(num):\n",
        "  return np.round(num*1000)/1000\n",
        "\n",
        "if __name__ == '__main__':\n",
        "  # Load data\n",
        "  print('Load data...')\n",
        "  data = np.load(data_dir + 'data_arrays.npz')\n",
        "  test_ids_patients = pd.read_pickle(data_dir + 'test_ids_patients.pkl')\n",
        "  \n",
        "  # Patients in test data\n",
        "  patients = test_ids_patients.drop_duplicates()\n",
        "  num_patients = patients.shape[0]\n",
        "  row_ids = pd.DataFrame({'ROW_IDX': test_ids_patients.index}, index=test_ids_patients)\n",
        "  \n",
        "  # Vocabulary sizes\n",
        "  #num_static = num_static(data)\n",
        "  num_static = data['static_vars'].shape[0]\n",
        "  num_dp_codes, num_cp_codes = vocab_sizes(data)\n",
        "\n",
        "  # CUDA for PyTorch\n",
        "  use_cuda = torch.cuda.is_available()\n",
        "  device = torch.device('cuda:0' if use_cuda else 'cpu')\n",
        "  torch.backends.cudnn.benchmark = True\n",
        "\n",
        "  # Network\n",
        "  net = Net(num_static, num_dp_codes, num_cp_codes).to(device)\n",
        "  \n",
        "  print('Evaluate...')\n",
        "  # Set log dir to read trained model from\n",
        "  logdir_temp = logdir + net_variant + '/'\n",
        "  #logdir = logdir + net_variant + '/'\n",
        "\n",
        "  # Restore variables from disk\n",
        "  net.load_state_dict(torch.load(logdir_temp + 'final_model.pt', map_location=device))\n",
        "\n",
        "  # Bootstrapping\n",
        "  np.random.seed(np_seed)\n",
        "  avpre_vec = np.zeros(bootstrap_samples)\n",
        "  auroc_vec = np.zeros(bootstrap_samples)\n",
        "  f1_vec    = np.zeros(bootstrap_samples)\n",
        "  sensitivity_vec = np.zeros(bootstrap_samples)\n",
        "  specificity_vec = np.zeros(bootstrap_samples)\n",
        "  ppv_vec = np.zeros(bootstrap_samples)\n",
        "  npv_vec = np.zeros(bootstrap_samples)  \n",
        "  \n",
        "  for sample in range(bootstrap_samples):\n",
        "    print('Bootstrap sample {}'.format(sample))\n",
        "  \n",
        "    # Test data\n",
        "    sample_patients = patients.sample(n=num_patients, replace=True)\n",
        "    idx = np.squeeze(row_ids.loc[sample_patients].values)\n",
        "    testloader, _, _ = get_trainloader(data, 'TEST', shuffle=False, idx=idx)\n",
        "\n",
        "    # evaluate on test data\n",
        "    net.eval()\n",
        "    label_pred = torch.Tensor([])\n",
        "    label_test = torch.Tensor([])\n",
        "    with torch.no_grad():\n",
        "      for i, (stat, dp, cp, dp_t, cp_t, label_batch) in enumerate(tqdm(testloader), 0):\n",
        "        # move to GPU if available\n",
        "        stat  = stat.to(device)\n",
        "        dp    = dp.to(device)\n",
        "        cp    = cp.to(device)\n",
        "        dp_t  = dp_t.to(device)\n",
        "        cp_t  = cp_t.to(device)\n",
        "      \n",
        "        label_pred_batch, _ = net(stat, dp, cp, dp_t, cp_t)\n",
        "        label_pred = torch.cat((label_pred, label_pred_batch.cpu()))\n",
        "        label_test = torch.cat((label_test, label_batch))\n",
        "        \n",
        "    label_sigmoids = torch.sigmoid(label_pred).cpu().numpy()\n",
        "\n",
        "    # Average precision\n",
        "    avpre = average_precision_score(label_test, label_sigmoids)\n",
        "    \n",
        "    # Determine AUROC score\n",
        "    auroc = roc_auc_score(label_test, label_sigmoids)\n",
        "\n",
        "    # Sensitivity, specificity\n",
        "    fpr, tpr, thresholds = roc_curve(label_test, label_sigmoids)\n",
        "    youden_idx = np.argmax(tpr - fpr)\n",
        "    sensitivity = tpr[youden_idx]\n",
        "    specificity = 1-fpr[youden_idx]\n",
        "    \n",
        "    # F1, PPV, NPV score\n",
        "    f1 = 0\n",
        "    ppv = 0\n",
        "    npv = 0\n",
        "    for t in thresholds:\n",
        "      label_pred = (np.array(label_sigmoids) >= t).astype(int)\n",
        "      f1_temp = f1_score(label_test, label_pred)\n",
        "      ppv_temp = precision_score(label_test, label_pred, pos_label=1)\n",
        "      npv_temp = precision_score(label_test, label_pred, pos_label=0)\n",
        "      if f1_temp > f1:\n",
        "        f1 = f1_temp\n",
        "      if (ppv_temp+npv_temp) > (ppv+npv):\n",
        "        ppv = ppv_temp\n",
        "        npv = npv_temp\n",
        "\n",
        "    # Store in vectors\n",
        "    avpre_vec[sample] = avpre\n",
        "    auroc_vec[sample] = auroc\n",
        "    f1_vec[sample]    = f1\n",
        "    sensitivity_vec[sample]  = sensitivity\n",
        "    specificity_vec[sample]  = specificity\n",
        "    ppv_vec[sample]  = ppv\n",
        "    npv_vec[sample]  = npv    \n",
        "\n",
        "  avpre_mean = np.mean(avpre_vec)\n",
        "  avpre_lci, avpre_uci = st.t.interval(0.95, bootstrap_samples-1, loc=avpre_mean, scale=st.sem(avpre_vec))\n",
        "  auroc_mean = np.mean(auroc_vec)\n",
        "  auroc_lci, auroc_uci = st.t.interval(0.95, bootstrap_samples-1, loc=auroc_mean, scale=st.sem(auroc_vec))\n",
        "  f1_mean = np.mean(f1_vec)\n",
        "  f1_lci, f1_uci = st.t.interval(0.95, bootstrap_samples-1, loc=f1_mean, scale=st.sem(f1_vec))\n",
        "  ppv_mean = np.mean(ppv_vec)\n",
        "  ppv_lci, ppv_uci = st.t.interval(0.95, bootstrap_samples-1, loc=ppv_mean, scale=st.sem(ppv_vec))\n",
        "  npv_mean = np.mean(npv_vec)\n",
        "  npv_lci, npv_uci = st.t.interval(0.95, bootstrap_samples-1, loc=npv_mean, scale=st.sem(npv_vec))    \n",
        "  sensitivity_mean = np.mean(sensitivity_vec)\n",
        "  sensitivity_lci, sensitivity_uci = st.t.interval(0.95, bootstrap_samples-1, loc=sensitivity_mean, scale=st.sem(sensitivity_vec))\n",
        "  specificity_mean = np.mean(specificity_vec)\n",
        "  specificity_lci, specificity_uci = st.t.interval(0.95, bootstrap_samples-1, loc=specificity_mean, scale=st.sem(specificity_vec))\n",
        "\n",
        "  epoch_times = np.load(logdir + net_variant + '/epoch_times.npz')['epoch_times']\n",
        "  times_mean = np.mean(epoch_times)\n",
        "  times_lci, times_uci = st.t.interval(0.95, len(epoch_times)-1, loc=np.mean(epoch_times), scale=st.sem(epoch_times))\n",
        "  times_std = np.std(epoch_times)\n",
        "  \n",
        "  print('------------------------------------------------')\n",
        "  print('Net variant: {}'.format(net_variant))\n",
        "  print('Average Precision: {} [{},{}]'.format(round(avpre_mean), round(avpre_lci), round(avpre_uci)))\n",
        "  print('AUROC: {} [{},{}]'.format(round(auroc_mean), round(auroc_lci), round(auroc_uci)))\n",
        "  print('F1: {} [{},{}]'.format(round(f1_mean), round(f1_lci), round(f1_uci)))  \n",
        "  print('PPV: {} [{},{}]'.format(round(ppv_mean), round(ppv_lci), round(ppv_uci)))\n",
        "  print('NPV: {} [{},{}]'.format(round(npv_mean), round(npv_lci), round(npv_uci)))  \n",
        "  print('Sensitivity: {} [{},{}]'.format(round(sensitivity_mean), round(sensitivity_lci), round(sensitivity_uci)))\n",
        "  print('Specificity: {} [{},{}]'.format(round(specificity_mean), round(specificity_lci), round(specificity_uci)))\n",
        "  print('Time: {} [{},{}] std: {}'.format(round(times_mean), round(times_lci), round(times_uci), round(times_std)))\n",
        "  print('Done')"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#References\n",
        "\n",
        "\n",
        "1.   Sebastiano Barbieri, James Kemp, Oscar Perez-Concha, Sradha Kotwal, Martin Gallagher, Angus Ritchie, and Louisa Jorm. 2020. Benchmarking deep learning architectures for predicting readmission to the ICU and describing patients-at-risk. Volume 10. \n",
        "\n",
        "2.  Alistair Johnson, Tom Pollard, Lu Shen, Li-wei Lehman, Mengling Feng, Mohammad Ghassemi, Benjamin Moody, Peter Szolovits, Leo Celi, and Roger Mark. 2016. Mimic-iii, a freely accessible critical care database. Volume 3, page 160035.\n",
        "\n",
        "3. A.L. Rosenberg and C. Watts. 2000. Patients readmitted to ICUs* : a systematic review of risk factors and outcomes. Chest.118.2.492\n",
        "\n"
      ],
      "metadata": {
        "id": "rvccdpZ8Cs5s"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "sXDDPLzuDqZ9"
      },
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "machine_shape": "hm",
      "provenance": []
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}