flair_cadec_v2_spanbert_large_cased.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "id": "xONJpNLKtGie"
      },
      "outputs": [],
      "source": [
        "import pandas as pd\n",
        "import os\n",
        "import pickle"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nw6RM8uotGih",
        "outputId": "54fb6185-1b90-4ac8-c300-104f053eb237"
      },
      "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",
        "\n",
        "os.chdir('/content/drive/MyDrive/SI 630 project')\n",
        "# os.listdir('.')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "P9LwpBadtGih",
        "outputId": "240b7d75-e069-4909-9e10-cd9a44c9a485"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "['sct',\n",
              " 'text',\n",
              " 'original - unedited version of edited files',\n",
              " 'meddra - unedited version of edited files',\n",
              " 'sct - unedited version of edited files',\n",
              " 'original',\n",
              " 'meddra']"
            ]
          },
          "execution_count": 4,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "os.listdir('datasets/CADEC/CADEC.v2/cadec')\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "VNZtmxiPb5uV"
      },
      "outputs": [],
      "source": [
        "suf = '_v2'\n",
        "\n",
        "data_save_dir = 'data'+suf\n",
        "\n",
        "suf += '_spanbert-large-cased'"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 424
        },
        "id": "BIUl2-W9tGii",
        "outputId": "0fc81059-f659-4872-f513-7f2ba1600c4e"
      },
      "outputs": [
        {
          "data": {
            "text/html": [
              "\n",
              "  <div id=\"df-ced12bca-83e4-4272-bd80-6e67f28ac487\">\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>ner</th>\n",
              "      <th>meddra</th>\n",
              "      <th>counts_exploded</th>\n",
              "      <th>counts</th>\n",
              "      <th>counts_exploded_rank</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Drug</td>\n",
              "      <td>Lipitor</td>\n",
              "      <td>1081</td>\n",
              "      <td>1073</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>ADR</td>\n",
              "      <td>CONCEPT_LESS</td>\n",
              "      <td>1016</td>\n",
              "      <td>266</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Severe_pain</td>\n",
              "      <td>722</td>\n",
              "      <td>247</td>\n",
              "      <td>3</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Myalgia</td>\n",
              "      <td>619</td>\n",
              "      <td>305</td>\n",
              "      <td>4</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Pain</td>\n",
              "      <td>292</td>\n",
              "      <td>260</td>\n",
              "      <td>5</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1294</th>\n",
              "      <td>Finding</td>\n",
              "      <td>Senility</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1233</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1295</th>\n",
              "      <td>Drug</td>\n",
              "      <td>Seroquel</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1233</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1296</th>\n",
              "      <td>Drug</td>\n",
              "      <td>Solaraze</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1233</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1297</th>\n",
              "      <td>Drug</td>\n",
              "      <td>Tagamet</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1233</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1298</th>\n",
              "      <td>Drug</td>\n",
              "      <td>testosterone</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1233</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>1299 rows × 5 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-ced12bca-83e4-4272-bd80-6e67f28ac487')\"\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-ced12bca-83e4-4272-bd80-6e67f28ac487 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-ced12bca-83e4-4272-bd80-6e67f28ac487');\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",
              "  "
            ],
            "text/plain": [
              "          ner        meddra  counts_exploded  counts  counts_exploded_rank\n",
              "0        Drug       Lipitor             1081    1073                     1\n",
              "1         ADR  CONCEPT_LESS             1016     266                     2\n",
              "2         ADR   Severe_pain              722     247                     3\n",
              "3         ADR       Myalgia              619     305                     4\n",
              "4         ADR          Pain              292     260                     5\n",
              "...       ...           ...              ...     ...                   ...\n",
              "1294  Finding      Senility                1       1                  1233\n",
              "1295     Drug      Seroquel                1       1                  1233\n",
              "1296     Drug      Solaraze                1       1                  1233\n",
              "1297     Drug       Tagamet                1       1                  1233\n",
              "1298     Drug  testosterone                1       1                  1233\n",
              "\n",
              "[1299 rows x 5 columns]"
            ]
          },
          "execution_count": 6,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "meddra_counts = pd.read_csv(data_save_dir+'/meddra_counts.csv')\n",
        "meddra_counts"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 424
        },
        "id": "xzc4Zo0PtGii",
        "outputId": "366f27ac-ae00-4d3c-e789-42e135fde8f9"
      },
      "outputs": [
        {
          "data": {
            "text/html": [
              "\n",
              "  <div id=\"df-963691c3-160b-4d31-9b6d-97254851f133\">\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>ner</th>\n",
              "      <th>meddra</th>\n",
              "      <th>counts_exploded</th>\n",
              "      <th>counts</th>\n",
              "      <th>counts_exploded_rank</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>ADR</td>\n",
              "      <td>CONCEPT_LESS</td>\n",
              "      <td>1016</td>\n",
              "      <td>266</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Severe_pain</td>\n",
              "      <td>722</td>\n",
              "      <td>247</td>\n",
              "      <td>3</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Myalgia</td>\n",
              "      <td>619</td>\n",
              "      <td>305</td>\n",
              "      <td>4</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Pain</td>\n",
              "      <td>292</td>\n",
              "      <td>260</td>\n",
              "      <td>5</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Arthralgia</td>\n",
              "      <td>280</td>\n",
              "      <td>163</td>\n",
              "      <td>6</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1268</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Labyrinthitis</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1233</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1269</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Dysphagia</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1233</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1275</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Lupus_erythematosus</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1233</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1278</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Cataract</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1233</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1283</th>\n",
              "      <td>ADR</td>\n",
              "      <td>Crying_associated_with_mood</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1233</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>800 rows × 5 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-963691c3-160b-4d31-9b6d-97254851f133')\"\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-963691c3-160b-4d31-9b6d-97254851f133 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-963691c3-160b-4d31-9b6d-97254851f133');\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",
              "  "
            ],
            "text/plain": [
              "      ner                       meddra  counts_exploded  counts  \\\n",
              "1     ADR                 CONCEPT_LESS             1016     266   \n",
              "2     ADR                  Severe_pain              722     247   \n",
              "3     ADR                      Myalgia              619     305   \n",
              "4     ADR                         Pain              292     260   \n",
              "5     ADR                   Arthralgia              280     163   \n",
              "...   ...                          ...              ...     ...   \n",
              "1268  ADR                Labyrinthitis                1       1   \n",
              "1269  ADR                    Dysphagia                1       1   \n",
              "1275  ADR          Lupus_erythematosus                1       1   \n",
              "1278  ADR                     Cataract                1       1   \n",
              "1283  ADR  Crying_associated_with_mood                1       1   \n",
              "\n",
              "      counts_exploded_rank  \n",
              "1                        2  \n",
              "2                        3  \n",
              "3                        4  \n",
              "4                        5  \n",
              "5                        6  \n",
              "...                    ...  \n",
              "1268                  1233  \n",
              "1269                  1233  \n",
              "1275                  1233  \n",
              "1278                  1233  \n",
              "1283                  1233  \n",
              "\n",
              "[800 rows x 5 columns]"
            ]
          },
          "execution_count": 7,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "meddra_counts.loc[meddra_counts.ner == 'ADR',:]"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hAGPini7tGii",
        "outputId": "ba3ad5e9-abab-4ff4-8b44-cea6f9255708"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "[Sentence[117]: \"i feel a bit drowsy & have a little blurred vision , so far no gastric problems . ive been on arthrotec 50 for over 10 years on and off , only taking it when i needed it . due to my arthritis getting progressively worse , to the point where i am in tears with the agony , gp ' s started me on 75 twice a day and i have to take it . every day for the next month to see how i get on , here goes . so far its been very good , pains almost gone , but i feel a bit weird , didnt have that when on 50 .\" → [\"i\"/O/O, \"feel\"/O/O, \"a\"/O/O, \"bit\"/B-ADR/B-Drowsy, \"drowsy\"/E-ADR/E-Drowsy, \"&\"/O/O, \"have\"/O/O, \"a\"/O/O, \"little\"/B-ADR/B-Blurred_vision_-_hazy, \"blurred\"/I-ADR/I-Blurred_vision_-_hazy, \"vision\"/E-ADR/E-Blurred_vision_-_hazy, \",\"/O/O, \"so\"/O/O, \"far\"/O/O, \"no\"/O/O, \"gastric\"/B-ADR/B-Excessive_upper_gastrointestinal_gas, \"problems\"/E-ADR/E-Excessive_upper_gastrointestinal_gas, \".\"/O/O, \"ive\"/O/O, \"been\"/O/O, \"on\"/O/O, \"arthrotec\"/S-Drug/S-Arthrotec, \"50\"/O/O, \"for\"/O/O, \"over\"/O/O, \"10\"/O/O, \"years\"/O/O, \"on\"/O/O, \"and\"/O/O, \"off\"/O/O, \",\"/O/O, \"only\"/O/O, \"taking\"/O/O, \"it\"/O/O, \"when\"/O/O, \"i\"/O/O, \"needed\"/O/O, \"it\"/O/O, \".\"/O/O, \"due\"/O/O, \"to\"/O/O, \"my\"/O/O, \"arthritis\"/S-Disease/S-Arthritis, \"getting\"/O/O, \"progressively\"/O/O, \"worse\"/O/O, \",\"/O/O, \"to\"/O/O, \"the\"/O/O, \"point\"/O/O, \"where\"/O/O, \"i\"/O/O, \"am\"/O/O, \"in\"/O/O, \"tears\"/O/O, \"with\"/O/O, \"the\"/O/O, \"agony\"/S-Symptom/S-Agony, \",\"/O/O, \"gp\"/O/O, \"'\"/O/O, \"s\"/O/O, \"started\"/O/O, \"me\"/O/O, \"on\"/O/O, \"75\"/O/O, \"twice\"/O/O, \"a\"/O/O, \"day\"/O/O, \"and\"/O/O, \"i\"/O/O, \"have\"/O/O, \"to\"/O/O, \"take\"/O/O, \"it\"/O/O, \".\"/O/O, \"every\"/O/O, \"day\"/O/O, \"for\"/O/O, \"the\"/O/O, \"next\"/O/O, \"month\"/O/O, \"to\"/O/O, \"see\"/O/O, \"how\"/O/O, \"i\"/O/O, \"get\"/O/O, \"on\"/O/O, \",\"/O/O, \"here\"/O/O, \"goes\"/O/O, \".\"/O/O, \"so\"/O/O, \"far\"/O/O, \"its\"/O/O, \"been\"/O/O, \"very\"/O/O, \"good\"/O/O, \",\"/O/O, \"pains\"/S-Symptom/S-Pain, \"almost\"/O/O, \"gone\"/O/O, \",\"/O/O, \"but\"/O/O, \"i\"/O/O, \"feel\"/B-ADR/B-Malaise, \"a\"/I-ADR/I-Malaise, \"bit\"/I-ADR/I-Malaise, \"weird\"/E-ADR/E-Malaise, \",\"/O/O, \"didnt\"/O/O, \"have\"/O/O, \"that\"/O/O, \"when\"/O/O, \"on\"/O/O, \"50\"/O/O, \".\"/O/O],\n",
              " Sentence[39]: \"hunger pangs . brilliant , i have a new lease of life , i walk up & down steps properly , no longer sideways like a toddler , hip pain as gone other than if i jar it .\" → [\"hunger\"/B-ADR/B-Hunger_pain, \"pangs\"/E-ADR/E-Hunger_pain, \".\"/O/O, \"brilliant\"/O/O, \",\"/O/O, \"i\"/O/O, \"have\"/O/O, \"a\"/O/O, \"new\"/O/O, \"lease\"/O/O, \"of\"/O/O, \"life\"/O/O, \",\"/O/O, \"i\"/O/O, \"walk\"/B-Symptom/B-CONCEPT_LESS, \"up\"/I-Symptom/I-CONCEPT_LESS, \"&\"/I-Symptom/I-CONCEPT_LESS, \"down\"/I-Symptom/I-CONCEPT_LESS, \"steps\"/I-Symptom/I-CONCEPT_LESS, \"properly\"/O/O, \",\"/O/O, \"no\"/O/O, \"longer\"/O/O, \"sideways\"/E-Symptom/E-CONCEPT_LESS, \"like\"/O/O, \"a\"/O/O, \"toddler\"/O/O, \",\"/O/O, \"hip\"/B-Symptom/B-Hip_pain, \"pain\"/E-Symptom/E-Hip_pain, \"as\"/O/O, \"gone\"/O/O, \"other\"/O/O, \"than\"/O/O, \"if\"/O/O, \"i\"/O/O, \"jar\"/O/O, \"it\"/O/O, \".\"/O/O],\n",
              " Sentence[51]: \"no side effects for the first two months . then vaginal bleeding 2 wks after menstral cycle . stomach pain . canker sores in my mouth . headache . been off for 1 week still have bleeding . helped my pain alot . too scared to take this drug again .\" → [\"no\"/O/O, \"side\"/O/O, \"effects\"/O/O, \"for\"/O/O, \"the\"/O/O, \"first\"/O/O, \"two\"/O/O, \"months\"/O/O, \".\"/O/O, \"then\"/O/O, \"vaginal\"/B-ADR/B-Vaginal_bleeding, \"bleeding\"/E-ADR/E-Vaginal_bleeding, \"2\"/O/O, \"wks\"/O/O, \"after\"/O/O, \"menstral\"/O/O, \"cycle\"/O/O, \".\"/O/O, \"stomach\"/B-ADR/B-Stomach_ache, \"pain\"/E-ADR/E-Stomach_ache, \".\"/O/O, \"canker\"/B-ADR/B-Aphthous_ulcer, \"sores\"/E-ADR/E-Aphthous_ulcer, \"in\"/O/O, \"my\"/O/O, \"mouth\"/O/O, \".\"/O/O, \"headache\"/S-ADR/S-Headache, \".\"/O/O, \"been\"/O/O, \"off\"/O/O, \"for\"/O/O, \"1\"/O/O, \"week\"/O/O, \"still\"/O/O, \"have\"/O/O, \"bleeding\"/O/O, \".\"/O/O, \"helped\"/O/O, \"my\"/O/O, \"pain\"/S-Symptom/S-Pain, \"alot\"/O/O, \".\"/O/O, \"too\"/O/O, \"scared\"/O/O, \"to\"/O/O, \"take\"/O/O, \"this\"/O/O, \"drug\"/O/O, \"again\"/O/O, \".\"/O/O],\n",
              " Sentence[68]: \"1st pill taken with food , a few hours after i experienced shortness of breath , a sense of depression , cramping , upset stomache will stop taking immediately . honestly can not recommend this drug , i am still experiencing side effects just after 1 pill and im supposed to take twice / day 75mg . i just cant keep going have to try something else .\" → [\"1st\"/O/O, \"pill\"/O/O, \"taken\"/O/O, \"with\"/O/O, \"food\"/O/O, \",\"/O/O, \"a\"/O/O, \"few\"/O/O, \"hours\"/O/O, \"after\"/O/O, \"i\"/O/O, \"experienced\"/O/O, \"shortness\"/B-ADR/B-Dyspnoea, \"of\"/I-ADR/I-Dyspnoea, \"breath\"/E-ADR/E-Dyspnoea, \",\"/O/O, \"a\"/O/O, \"sense\"/O/O, \"of\"/O/O, \"depression\"/S-ADR/S-Depression, \",\"/O/O, \"cramping\"/S-ADR/S-Muscle_cramp, \",\"/O/O, \"upset\"/B-ADR/B-Upset_stomach, \"stomache\"/E-ADR/E-Upset_stomach, \"will\"/O/O, \"stop\"/O/O, \"taking\"/O/O, \"immediately\"/O/O, \".\"/O/O, \"honestly\"/O/O, \"can\"/O/O, \"not\"/O/O, \"recommend\"/O/O, \"this\"/O/O, \"drug\"/O/O, \",\"/O/O, \"i\"/O/O, \"am\"/O/O, \"still\"/O/O, \"experiencing\"/O/O, \"side\"/O/O, \"effects\"/O/O, \"just\"/O/O, \"after\"/O/O, \"1\"/O/O, \"pill\"/O/O, \"and\"/O/O, \"im\"/O/O, \"supposed\"/O/O, \"to\"/O/O, \"take\"/O/O, \"twice\"/O/O, \"/\"/O/O, \"day\"/O/O, \"75mg\"/O/O, \".\"/O/O, \"i\"/O/O, \"just\"/O/O, \"cant\"/O/O, \"keep\"/O/O, \"going\"/O/O, \"have\"/O/O, \"to\"/O/O, \"try\"/O/O, \"something\"/O/O, \"else\"/O/O, \".\"/O/O],\n",
              " Sentence[50]: \"i have had no side effects been taking arthrotec a little over a year , have not noticed any side effects . it does help alot i noticed that when there are times when i forget to take it i cant stand or walk for any lengths of time .\" → [\"i\"/O/O, \"have\"/O/O, \"had\"/O/O, \"no\"/O/O, \"side\"/O/O, \"effects\"/O/O, \"been\"/O/O, \"taking\"/O/O, \"arthrotec\"/S-Drug/S-Arthrotec, \"a\"/O/O, \"little\"/O/O, \"over\"/O/O, \"a\"/O/O, \"year\"/O/O, \",\"/O/O, \"have\"/O/O, \"not\"/O/O, \"noticed\"/O/O, \"any\"/O/O, \"side\"/O/O, \"effects\"/O/O, \".\"/O/O, \"it\"/O/O, \"does\"/O/O, \"help\"/O/O, \"alot\"/O/O, \"i\"/O/O, \"noticed\"/O/O, \"that\"/O/O, \"when\"/O/O, \"there\"/O/O, \"are\"/O/O, \"times\"/O/O, \"when\"/O/O, \"i\"/O/O, \"forget\"/O/O, \"to\"/O/O, \"take\"/O/O, \"it\"/O/O, \"i\"/B-Symptom/B-Reduced_mobility, \"cant\"/I-Symptom/I-Reduced_mobility, \"stand\"/I-Symptom/I-Reduced_mobility, \"or\"/I-Symptom/I-Reduced_mobility, \"walk\"/I-Symptom/I-Reduced_mobility, \"for\"/I-Symptom/I-Reduced_mobility, \"any\"/I-Symptom/I-Reduced_mobility, \"lengths\"/I-Symptom/I-Reduced_mobility, \"of\"/I-Symptom/I-Reduced_mobility, \"time\"/E-Symptom/E-Reduced_mobility, \".\"/O/O],\n",
              " Sentence[89]: \"it seems as if bleeding forever is a side effect but doctors tell you that its not the drug . i have been on arthotec for three months i still have pain and stiffness . at first week into this drug i was in so much pain i sat on the floor at work and thought i was gaving a baby . i always feel weak and almost fainted a bunch of times . consider somthing that works i am switching today as this isnt in my head .\" → [\"it\"/O/O, \"seems\"/O/O, \"as\"/O/O, \"if\"/O/O, \"bleeding\"/S-ADR/S-Bleeding, \"forever\"/O/O, \"is\"/O/O, \"a\"/O/O, \"side\"/O/O, \"effect\"/O/O, \"but\"/O/O, \"doctors\"/O/O, \"tell\"/O/O, \"you\"/O/O, \"that\"/O/O, \"its\"/O/O, \"not\"/O/O, \"the\"/O/O, \"drug\"/O/O, \".\"/O/O, \"i\"/O/O, \"have\"/O/O, \"been\"/O/O, \"on\"/O/O, \"arthotec\"/S-Drug/S-Arthrotec, \"for\"/O/O, \"three\"/O/O, \"months\"/O/O, \"i\"/O/O, \"still\"/O/O, \"have\"/O/O, \"pain\"/B-Symptom/B-Pain, \"and\"/I-Symptom/I-Pain, \"stiffness\"/E-Symptom/E-Pain, \".\"/O/O, \"at\"/O/O, \"first\"/O/O, \"week\"/O/O, \"into\"/O/O, \"this\"/O/O, \"drug\"/O/O, \"i\"/O/O, \"was\"/O/O, \"in\"/O/O, \"so\"/B-ADR/B-Pain, \"much\"/I-ADR/I-Pain, \"pain\"/E-ADR/E-Pain, \"i\"/O/O, \"sat\"/O/O, \"on\"/O/O, \"the\"/O/O, \"floor\"/O/O, \"at\"/O/O, \"work\"/O/O, \"and\"/O/O, \"thought\"/O/O, \"i\"/O/O, \"was\"/O/O, \"gaving\"/O/O, \"a\"/O/O, \"baby\"/O/O, \".\"/O/O, \"i\"/O/O, \"always\"/O/O, \"feel\"/B-ADR/B-Malaise_and_fatigue, \"weak\"/I-ADR/I-Malaise_and_fatigue, \"and\"/I-ADR/I-Malaise_and_fatigue, \"almost\"/I-ADR/I-Malaise_and_fatigue, \"fainted\"/E-ADR/E-Malaise_and_fatigue, \"a\"/O/O, \"bunch\"/O/O, \"of\"/O/O, \"times\"/O/O, \".\"/O/O, \"consider\"/O/O, \"somthing\"/O/O, \"that\"/O/O, \"works\"/O/O, \"i\"/O/O, \"am\"/O/O, \"switching\"/O/O, \"today\"/O/O, \"as\"/O/O, \"this\"/O/O, \"isnt\"/O/O, \"in\"/O/O, \"my\"/O/O, \"head\"/O/O, \".\"/O/O],\n",
              " Sentence[56]: \"some diarrhea , some constipation , fatigue . it works very well for the heel ; i couldnt walk on it very well before taking it . i take 75mg . twice a day . also have severe osteoarthritis in the knees and hands , but it doesnt seem to relieve that pain at all .\" → [\"some\"/O/O, \"diarrhea\"/S-ADR/S-Diarrhoea, \",\"/O/O, \"some\"/O/O, \"constipation\"/S-ADR/S-Constipation, \",\"/O/O, \"fatigue\"/S-ADR/S-Fatigue, \".\"/O/O, \"it\"/O/O, \"works\"/O/O, \"very\"/O/O, \"well\"/O/O, \"for\"/O/O, \"the\"/B-Symptom/B-Walking_disability, \"heel\"/I-Symptom/I-Walking_disability, \";\"/O/O, \"i\"/I-Symptom/I-Walking_disability, \"couldnt\"/I-Symptom/I-Walking_disability, \"walk\"/I-Symptom/I-Walking_disability, \"on\"/I-Symptom/I-Walking_disability, \"it\"/O/O, \"very\"/I-Symptom/I-Walking_disability, \"well\"/E-Symptom/E-Walking_disability, \"before\"/O/O, \"taking\"/O/O, \"it\"/O/O, \".\"/O/O, \"i\"/O/O, \"take\"/O/O, \"75mg\"/O/O, \".\"/O/O, \"twice\"/O/O, \"a\"/O/O, \"day\"/O/O, \".\"/O/O, \"also\"/O/O, \"have\"/O/O, \"severe\"/B-Disease/B-Osteoarthritis_of_knee/B-Osteoarthrosis_of_hand, \"osteoarthritis\"/I-Disease/I-Osteoarthritis_of_knee/I-Osteoarthrosis_of_hand, \"in\"/I-Disease/I-Osteoarthritis_of_knee/I-Osteoarthrosis_of_hand, \"the\"/I-Disease/I-Osteoarthritis_of_knee/I-Osteoarthrosis_of_hand, \"knees\"/E-Disease/E-Osteoarthritis_of_knee, \"and\"/O/O, \"hands\"/E-Disease/E-Osteoarthrosis_of_hand, \",\"/O/O, \"but\"/O/O, \"it\"/O/O, \"doesnt\"/O/O, \"seem\"/O/O, \"to\"/O/O, \"relieve\"/O/O, \"that\"/O/O, \"pain\"/S-Symptom/S-Pain, \"at\"/O/O, \"all\"/O/O, \".\"/O/O],\n",
              " Sentence[84]: \"ive been on 75mg twice daily for 13 days , with recent body swelling , face , wrists , abdomen , thighs - clothes fit tight . insomina is much worse . i have forgetfulnes and confussion - really bad - almost dangerous . have felt \" empty stomach \" feeling , but not bothersome . have had more loose stools , but tolerable . im looking for a more natural therapy , checking into ha ( hyaluronic acid ) to replace arthrotec .\" → [\"ive\"/O/O, \"been\"/O/O, \"on\"/O/O, \"75mg\"/O/O, \"twice\"/O/O, \"daily\"/O/O, \"for\"/O/O, \"13\"/O/O, \"days\"/O/O, \",\"/O/O, \"with\"/O/O, \"recent\"/O/O, \"body\"/B-ADR/B-Swelling_of_body_region, \"swelling\"/B-ADR/E-ADR/B-Abdominal_swelling/B-Facial_swelling/E-Swelling_of_body_region/B-Swelling_of_lower_limb/B-Swelling_of_wrist_joint, \",\"/O/O, \"face\"/E-ADR/E-Facial_swelling, \",\"/O/O, \"wrists\"/E-ADR/E-Swelling_of_wrist_joint, \",\"/O/O, \"abdomen\"/E-ADR/E-Abdominal_swelling, \",\"/O/O, \"thighs\"/E-ADR/E-Swelling_of_lower_limb, \"-\"/O/O, \"clothes\"/O/O, \"fit\"/O/O, \"tight\"/O/O, \".\"/O/O, \"insomina\"/S-ADR/S-Insomnia, \"is\"/O/O, \"much\"/O/O, \"worse\"/O/O, \".\"/O/O, \"i\"/O/O, \"have\"/O/O, \"forgetfulnes\"/B-ADR/B-Forgetful, \"and\"/I-ADR/I-Forgetful, \"confussion\"/E-ADR/E-Forgetful, \"-\"/O/O, \"really\"/O/O, \"bad\"/O/O, \"-\"/O/O, \"almost\"/O/O, \"dangerous\"/O/O, \".\"/O/O, \"have\"/O/O, \"felt\"/O/O, \"\"\"/B-ADR/B-Feeling_empty, \"empty\"/I-ADR/I-Feeling_empty, \"stomach\"/I-ADR/I-Feeling_empty, \"\"\"/I-ADR/I-Feeling_empty, \"feeling\"/E-ADR/E-Feeling_empty, \",\"/O/O, \"but\"/O/O, \"not\"/O/O, \"bothersome\"/O/O, \".\"/O/O, \"have\"/O/O, \"had\"/O/O, \"more\"/O/O, \"loose\"/B-ADR/B-Loose_stool, \"stools\"/E-ADR/E-Loose_stool, \",\"/O/O, \"but\"/O/O, \"tolerable\"/O/O, \".\"/O/O, \"im\"/O/O, \"looking\"/O/O, \"for\"/O/O, \"a\"/O/O, \"more\"/O/O, \"natural\"/O/O, \"therapy\"/O/O, \",\"/O/O, \"checking\"/O/O, \"into\"/O/O, \"ha\"/B-Drug/B-CONCEPT_LESS, \"(\"/I-Drug/I-CONCEPT_LESS, \"hyaluronic\"/I-Drug/I-CONCEPT_LESS, \"acid\"/I-Drug/I-CONCEPT_LESS, \")\"/E-Drug/E-CONCEPT_LESS, \"to\"/O/O, \"replace\"/O/O, \"arthrotec\"/S-Drug/S-Arthrotec, \".\"/O/O],\n",
              " Sentence[68]: \"none that i can think of . i recently had bilateral knee replacement . prior to that , arthrotec was the only drug that helped . i tried every drug on the market , but , to no avail . i took 50 mg . 3 times a day of arthrotec and was almost pain free . i highly recommend it to anyone that suffers from arthritis .\" → [\"none\"/O/O, \"that\"/O/O, \"i\"/O/O, \"can\"/O/O, \"think\"/O/O, \"of\"/O/O, \".\"/O/O, \"i\"/O/O, \"recently\"/O/O, \"had\"/O/O, \"bilateral\"/O/O, \"knee\"/O/O, \"replacement\"/O/O, \".\"/O/O, \"prior\"/O/O, \"to\"/O/O, \"that\"/O/O, \",\"/O/O, \"arthrotec\"/S-Drug/S-Arthrotec, \"was\"/O/O, \"the\"/O/O, \"only\"/O/O, \"drug\"/O/O, \"that\"/O/O, \"helped\"/O/O, \".\"/O/O, \"i\"/O/O, \"tried\"/O/O, \"every\"/O/O, \"drug\"/O/O, \"on\"/O/O, \"the\"/O/O, \"market\"/O/O, \",\"/O/O, \"but\"/O/O, \",\"/O/O, \"to\"/O/O, \"no\"/O/O, \"avail\"/O/O, \".\"/O/O, \"i\"/O/O, \"took\"/O/O, \"50\"/O/O, \"mg\"/O/O, \".\"/O/O, \"3\"/O/O, \"times\"/O/O, \"a\"/O/O, \"day\"/O/O, \"of\"/O/O, \"arthrotec\"/S-Drug/S-Arthrotec, \"and\"/O/O, \"was\"/O/O, \"almost\"/O/O, \"pain\"/S-Symptom/S-Pain, \"free\"/O/O, \".\"/O/O, \"i\"/O/O, \"highly\"/O/O, \"recommend\"/O/O, \"it\"/O/O, \"to\"/O/O, \"anyone\"/O/O, \"that\"/O/O, \"suffers\"/O/O, \"from\"/O/O, \"arthritis\"/S-Finding/S-Arthritis, \".\"/O/O],\n",
              " Sentence[28]: \"hospitalization due extreme rectal bleed that required blood transfusion . i was extremely sick and initially felt i might not survive . should be removed from market .\" → [\"hospitalization\"/O/O, \"due\"/O/O, \"extreme\"/O/O, \"rectal\"/B-ADR/B-Rectal_haemorrhage, \"bleed\"/E-ADR/E-Rectal_haemorrhage, \"that\"/O/O, \"required\"/O/O, \"blood\"/O/O, \"transfusion\"/O/O, \".\"/O/O, \"i\"/O/O, \"was\"/O/O, \"extremely\"/B-ADR/B-Generally_unwell, \"sick\"/E-ADR/E-Generally_unwell, \"and\"/O/O, \"initially\"/O/O, \"felt\"/B-ADR/B-CONCEPT_LESS, \"i\"/I-ADR/I-CONCEPT_LESS, \"might\"/I-ADR/I-CONCEPT_LESS, \"not\"/I-ADR/I-CONCEPT_LESS, \"survive\"/E-ADR/E-CONCEPT_LESS, \".\"/O/O, \"should\"/O/O, \"be\"/O/O, \"removed\"/O/O, \"from\"/O/O, \"market\"/O/O, \".\"/O/O]]"
            ]
          },
          "execution_count": 8,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "with open(data_save_dir+'/cadec_flair_objects.pkl', 'rb') as f:\n",
        "    docs = pickle.load(f)\n",
        "docs[:10]"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sW10pPAPtGij",
        "outputId": "784bb1cf-1d16-4c99-acc3-01a6d2a9357e"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "1186"
            ]
          },
          "execution_count": 9,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "len(docs)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EMT5TsNStGij",
        "outputId": "fd57c767-128c-483f-f0fa-0337524eff49"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "Sentence[117]: \"i feel a bit drowsy & have a little blurred vision , so far no gastric problems . ive been on arthrotec 50 for over 10 years on and off , only taking it when i needed it . due to my arthritis getting progressively worse , to the point where i am in tears with the agony , gp ' s started me on 75 twice a day and i have to take it . every day for the next month to see how i get on , here goes . so far its been very good , pains almost gone , but i feel a bit weird , didnt have that when on 50 .\" → [\"i\"/O/O, \"feel\"/O/O, \"a\"/O/O, \"bit\"/B-ADR/B-Drowsy, \"drowsy\"/E-ADR/E-Drowsy, \"&\"/O/O, \"have\"/O/O, \"a\"/O/O, \"little\"/B-ADR/B-Blurred_vision_-_hazy, \"blurred\"/I-ADR/I-Blurred_vision_-_hazy, \"vision\"/E-ADR/E-Blurred_vision_-_hazy, \",\"/O/O, \"so\"/O/O, \"far\"/O/O, \"no\"/O/O, \"gastric\"/B-ADR/B-Excessive_upper_gastrointestinal_gas, \"problems\"/E-ADR/E-Excessive_upper_gastrointestinal_gas, \".\"/O/O, \"ive\"/O/O, \"been\"/O/O, \"on\"/O/O, \"arthrotec\"/S-Drug/S-Arthrotec, \"50\"/O/O, \"for\"/O/O, \"over\"/O/O, \"10\"/O/O, \"years\"/O/O, \"on\"/O/O, \"and\"/O/O, \"off\"/O/O, \",\"/O/O, \"only\"/O/O, \"taking\"/O/O, \"it\"/O/O, \"when\"/O/O, \"i\"/O/O, \"needed\"/O/O, \"it\"/O/O, \".\"/O/O, \"due\"/O/O, \"to\"/O/O, \"my\"/O/O, \"arthritis\"/S-Disease/S-Arthritis, \"getting\"/O/O, \"progressively\"/O/O, \"worse\"/O/O, \",\"/O/O, \"to\"/O/O, \"the\"/O/O, \"point\"/O/O, \"where\"/O/O, \"i\"/O/O, \"am\"/O/O, \"in\"/O/O, \"tears\"/O/O, \"with\"/O/O, \"the\"/O/O, \"agony\"/S-Symptom/S-Agony, \",\"/O/O, \"gp\"/O/O, \"'\"/O/O, \"s\"/O/O, \"started\"/O/O, \"me\"/O/O, \"on\"/O/O, \"75\"/O/O, \"twice\"/O/O, \"a\"/O/O, \"day\"/O/O, \"and\"/O/O, \"i\"/O/O, \"have\"/O/O, \"to\"/O/O, \"take\"/O/O, \"it\"/O/O, \".\"/O/O, \"every\"/O/O, \"day\"/O/O, \"for\"/O/O, \"the\"/O/O, \"next\"/O/O, \"month\"/O/O, \"to\"/O/O, \"see\"/O/O, \"how\"/O/O, \"i\"/O/O, \"get\"/O/O, \"on\"/O/O, \",\"/O/O, \"here\"/O/O, \"goes\"/O/O, \".\"/O/O, \"so\"/O/O, \"far\"/O/O, \"its\"/O/O, \"been\"/O/O, \"very\"/O/O, \"good\"/O/O, \",\"/O/O, \"pains\"/S-Symptom/S-Pain, \"almost\"/O/O, \"gone\"/O/O, \",\"/O/O, \"but\"/O/O, \"i\"/O/O, \"feel\"/B-ADR/B-Malaise, \"a\"/I-ADR/I-Malaise, \"bit\"/I-ADR/I-Malaise, \"weird\"/E-ADR/E-Malaise, \",\"/O/O, \"didnt\"/O/O, \"have\"/O/O, \"that\"/O/O, \"when\"/O/O, \"on\"/O/O, \"50\"/O/O, \".\"/O/O]"
            ]
          },
          "execution_count": 10,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "temp = docs[0]\n",
        "temp"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2EeTdfHstGij",
        "outputId": "2a1fc2b5-979c-4aec-87aa-5e3be68d90d7"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "{'ner': ['Token[0]: \"i\"'/'O' (1.0),\n",
              "  'Token[1]: \"feel\"'/'O' (1.0),\n",
              "  'Token[2]: \"a\"'/'O' (1.0),\n",
              "  'Token[3]: \"bit\"'/'B-ADR' (1.0),\n",
              "  'Token[4]: \"drowsy\"'/'E-ADR' (1.0),\n",
              "  'Token[5]: \"&\"'/'O' (1.0),\n",
              "  'Token[6]: \"have\"'/'O' (1.0),\n",
              "  'Token[7]: \"a\"'/'O' (1.0),\n",
              "  'Token[8]: \"little\"'/'B-ADR' (1.0),\n",
              "  'Token[9]: \"blurred\"'/'I-ADR' (1.0),\n",
              "  'Token[10]: \"vision\"'/'E-ADR' (1.0),\n",
              "  'Token[11]: \",\"'/'O' (1.0),\n",
              "  'Token[12]: \"so\"'/'O' (1.0),\n",
              "  'Token[13]: \"far\"'/'O' (1.0),\n",
              "  'Token[14]: \"no\"'/'O' (1.0),\n",
              "  'Token[15]: \"gastric\"'/'B-ADR' (1.0),\n",
              "  'Token[16]: \"problems\"'/'E-ADR' (1.0),\n",
              "  'Token[17]: \".\"'/'O' (1.0),\n",
              "  'Token[18]: \"ive\"'/'O' (1.0),\n",
              "  'Token[19]: \"been\"'/'O' (1.0),\n",
              "  'Token[20]: \"on\"'/'O' (1.0),\n",
              "  'Token[21]: \"arthrotec\"'/'S-Drug' (1.0),\n",
              "  'Token[22]: \"50\"'/'O' (1.0),\n",
              "  'Token[23]: \"for\"'/'O' (1.0),\n",
              "  'Token[24]: \"over\"'/'O' (1.0),\n",
              "  'Token[25]: \"10\"'/'O' (1.0),\n",
              "  'Token[26]: \"years\"'/'O' (1.0),\n",
              "  'Token[27]: \"on\"'/'O' (1.0),\n",
              "  'Token[28]: \"and\"'/'O' (1.0),\n",
              "  'Token[29]: \"off\"'/'O' (1.0),\n",
              "  'Token[30]: \",\"'/'O' (1.0),\n",
              "  'Token[31]: \"only\"'/'O' (1.0),\n",
              "  'Token[32]: \"taking\"'/'O' (1.0),\n",
              "  'Token[33]: \"it\"'/'O' (1.0),\n",
              "  'Token[34]: \"when\"'/'O' (1.0),\n",
              "  'Token[35]: \"i\"'/'O' (1.0),\n",
              "  'Token[36]: \"needed\"'/'O' (1.0),\n",
              "  'Token[37]: \"it\"'/'O' (1.0),\n",
              "  'Token[38]: \".\"'/'O' (1.0),\n",
              "  'Token[39]: \"due\"'/'O' (1.0),\n",
              "  'Token[40]: \"to\"'/'O' (1.0),\n",
              "  'Token[41]: \"my\"'/'O' (1.0),\n",
              "  'Token[42]: \"arthritis\"'/'S-Disease' (1.0),\n",
              "  'Token[43]: \"getting\"'/'O' (1.0),\n",
              "  'Token[44]: \"progressively\"'/'O' (1.0),\n",
              "  'Token[45]: \"worse\"'/'O' (1.0),\n",
              "  'Token[46]: \",\"'/'O' (1.0),\n",
              "  'Token[47]: \"to\"'/'O' (1.0),\n",
              "  'Token[48]: \"the\"'/'O' (1.0),\n",
              "  'Token[49]: \"point\"'/'O' (1.0),\n",
              "  'Token[50]: \"where\"'/'O' (1.0),\n",
              "  'Token[51]: \"i\"'/'O' (1.0),\n",
              "  'Token[52]: \"am\"'/'O' (1.0),\n",
              "  'Token[53]: \"in\"'/'O' (1.0),\n",
              "  'Token[54]: \"tears\"'/'O' (1.0),\n",
              "  'Token[55]: \"with\"'/'O' (1.0),\n",
              "  'Token[56]: \"the\"'/'O' (1.0),\n",
              "  'Token[57]: \"agony\"'/'S-Symptom' (1.0),\n",
              "  'Token[58]: \",\"'/'O' (1.0),\n",
              "  'Token[59]: \"gp\"'/'O' (1.0),\n",
              "  'Token[60]: \"'\"'/'O' (1.0),\n",
              "  'Token[61]: \"s\"'/'O' (1.0),\n",
              "  'Token[62]: \"started\"'/'O' (1.0),\n",
              "  'Token[63]: \"me\"'/'O' (1.0),\n",
              "  'Token[64]: \"on\"'/'O' (1.0),\n",
              "  'Token[65]: \"75\"'/'O' (1.0),\n",
              "  'Token[66]: \"twice\"'/'O' (1.0),\n",
              "  'Token[67]: \"a\"'/'O' (1.0),\n",
              "  'Token[68]: \"day\"'/'O' (1.0),\n",
              "  'Token[69]: \"and\"'/'O' (1.0),\n",
              "  'Token[70]: \"i\"'/'O' (1.0),\n",
              "  'Token[71]: \"have\"'/'O' (1.0),\n",
              "  'Token[72]: \"to\"'/'O' (1.0),\n",
              "  'Token[73]: \"take\"'/'O' (1.0),\n",
              "  'Token[74]: \"it\"'/'O' (1.0),\n",
              "  'Token[75]: \".\"'/'O' (1.0),\n",
              "  'Token[76]: \"every\"'/'O' (1.0),\n",
              "  'Token[77]: \"day\"'/'O' (1.0),\n",
              "  'Token[78]: \"for\"'/'O' (1.0),\n",
              "  'Token[79]: \"the\"'/'O' (1.0),\n",
              "  'Token[80]: \"next\"'/'O' (1.0),\n",
              "  'Token[81]: \"month\"'/'O' (1.0),\n",
              "  'Token[82]: \"to\"'/'O' (1.0),\n",
              "  'Token[83]: \"see\"'/'O' (1.0),\n",
              "  'Token[84]: \"how\"'/'O' (1.0),\n",
              "  'Token[85]: \"i\"'/'O' (1.0),\n",
              "  'Token[86]: \"get\"'/'O' (1.0),\n",
              "  'Token[87]: \"on\"'/'O' (1.0),\n",
              "  'Token[88]: \",\"'/'O' (1.0),\n",
              "  'Token[89]: \"here\"'/'O' (1.0),\n",
              "  'Token[90]: \"goes\"'/'O' (1.0),\n",
              "  'Token[91]: \".\"'/'O' (1.0),\n",
              "  'Token[92]: \"so\"'/'O' (1.0),\n",
              "  'Token[93]: \"far\"'/'O' (1.0),\n",
              "  'Token[94]: \"its\"'/'O' (1.0),\n",
              "  'Token[95]: \"been\"'/'O' (1.0),\n",
              "  'Token[96]: \"very\"'/'O' (1.0),\n",
              "  'Token[97]: \"good\"'/'O' (1.0),\n",
              "  'Token[98]: \",\"'/'O' (1.0),\n",
              "  'Token[99]: \"pains\"'/'S-Symptom' (1.0),\n",
              "  'Token[100]: \"almost\"'/'O' (1.0),\n",
              "  'Token[101]: \"gone\"'/'O' (1.0),\n",
              "  'Token[102]: \",\"'/'O' (1.0),\n",
              "  'Token[103]: \"but\"'/'O' (1.0),\n",
              "  'Token[104]: \"i\"'/'O' (1.0),\n",
              "  'Token[105]: \"feel\"'/'B-ADR' (1.0),\n",
              "  'Token[106]: \"a\"'/'I-ADR' (1.0),\n",
              "  'Token[107]: \"bit\"'/'I-ADR' (1.0),\n",
              "  'Token[108]: \"weird\"'/'E-ADR' (1.0),\n",
              "  'Token[109]: \",\"'/'O' (1.0),\n",
              "  'Token[110]: \"didnt\"'/'O' (1.0),\n",
              "  'Token[111]: \"have\"'/'O' (1.0),\n",
              "  'Token[112]: \"that\"'/'O' (1.0),\n",
              "  'Token[113]: \"when\"'/'O' (1.0),\n",
              "  'Token[114]: \"on\"'/'O' (1.0),\n",
              "  'Token[115]: \"50\"'/'O' (1.0),\n",
              "  'Token[116]: \".\"'/'O' (1.0)],\n",
              " 'meddra': ['Token[0]: \"i\"'/'O' (1.0),\n",
              "  'Token[1]: \"feel\"'/'O' (1.0),\n",
              "  'Token[2]: \"a\"'/'O' (1.0),\n",
              "  'Token[3]: \"bit\"'/'B-Drowsy' (1.0),\n",
              "  'Token[4]: \"drowsy\"'/'E-Drowsy' (1.0),\n",
              "  'Token[5]: \"&\"'/'O' (1.0),\n",
              "  'Token[6]: \"have\"'/'O' (1.0),\n",
              "  'Token[7]: \"a\"'/'O' (1.0),\n",
              "  'Token[8]: \"little\"'/'B-Blurred_vision_-_hazy' (1.0),\n",
              "  'Token[9]: \"blurred\"'/'I-Blurred_vision_-_hazy' (1.0),\n",
              "  'Token[10]: \"vision\"'/'E-Blurred_vision_-_hazy' (1.0),\n",
              "  'Token[11]: \",\"'/'O' (1.0),\n",
              "  'Token[12]: \"so\"'/'O' (1.0),\n",
              "  'Token[13]: \"far\"'/'O' (1.0),\n",
              "  'Token[14]: \"no\"'/'O' (1.0),\n",
              "  'Token[15]: \"gastric\"'/'B-Excessive_upper_gastrointestinal_gas' (1.0),\n",
              "  'Token[16]: \"problems\"'/'E-Excessive_upper_gastrointestinal_gas' (1.0),\n",
              "  'Token[17]: \".\"'/'O' (1.0),\n",
              "  'Token[18]: \"ive\"'/'O' (1.0),\n",
              "  'Token[19]: \"been\"'/'O' (1.0),\n",
              "  'Token[20]: \"on\"'/'O' (1.0),\n",
              "  'Token[21]: \"arthrotec\"'/'S-Arthrotec' (1.0),\n",
              "  'Token[22]: \"50\"'/'O' (1.0),\n",
              "  'Token[23]: \"for\"'/'O' (1.0),\n",
              "  'Token[24]: \"over\"'/'O' (1.0),\n",
              "  'Token[25]: \"10\"'/'O' (1.0),\n",
              "  'Token[26]: \"years\"'/'O' (1.0),\n",
              "  'Token[27]: \"on\"'/'O' (1.0),\n",
              "  'Token[28]: \"and\"'/'O' (1.0),\n",
              "  'Token[29]: \"off\"'/'O' (1.0),\n",
              "  'Token[30]: \",\"'/'O' (1.0),\n",
              "  'Token[31]: \"only\"'/'O' (1.0),\n",
              "  'Token[32]: \"taking\"'/'O' (1.0),\n",
              "  'Token[33]: \"it\"'/'O' (1.0),\n",
              "  'Token[34]: \"when\"'/'O' (1.0),\n",
              "  'Token[35]: \"i\"'/'O' (1.0),\n",
              "  'Token[36]: \"needed\"'/'O' (1.0),\n",
              "  'Token[37]: \"it\"'/'O' (1.0),\n",
              "  'Token[38]: \".\"'/'O' (1.0),\n",
              "  'Token[39]: \"due\"'/'O' (1.0),\n",
              "  'Token[40]: \"to\"'/'O' (1.0),\n",
              "  'Token[41]: \"my\"'/'O' (1.0),\n",
              "  'Token[42]: \"arthritis\"'/'S-Arthritis' (1.0),\n",
              "  'Token[43]: \"getting\"'/'O' (1.0),\n",
              "  'Token[44]: \"progressively\"'/'O' (1.0),\n",
              "  'Token[45]: \"worse\"'/'O' (1.0),\n",
              "  'Token[46]: \",\"'/'O' (1.0),\n",
              "  'Token[47]: \"to\"'/'O' (1.0),\n",
              "  'Token[48]: \"the\"'/'O' (1.0),\n",
              "  'Token[49]: \"point\"'/'O' (1.0),\n",
              "  'Token[50]: \"where\"'/'O' (1.0),\n",
              "  'Token[51]: \"i\"'/'O' (1.0),\n",
              "  'Token[52]: \"am\"'/'O' (1.0),\n",
              "  'Token[53]: \"in\"'/'O' (1.0),\n",
              "  'Token[54]: \"tears\"'/'O' (1.0),\n",
              "  'Token[55]: \"with\"'/'O' (1.0),\n",
              "  'Token[56]: \"the\"'/'O' (1.0),\n",
              "  'Token[57]: \"agony\"'/'S-Agony' (1.0),\n",
              "  'Token[58]: \",\"'/'O' (1.0),\n",
              "  'Token[59]: \"gp\"'/'O' (1.0),\n",
              "  'Token[60]: \"'\"'/'O' (1.0),\n",
              "  'Token[61]: \"s\"'/'O' (1.0),\n",
              "  'Token[62]: \"started\"'/'O' (1.0),\n",
              "  'Token[63]: \"me\"'/'O' (1.0),\n",
              "  'Token[64]: \"on\"'/'O' (1.0),\n",
              "  'Token[65]: \"75\"'/'O' (1.0),\n",
              "  'Token[66]: \"twice\"'/'O' (1.0),\n",
              "  'Token[67]: \"a\"'/'O' (1.0),\n",
              "  'Token[68]: \"day\"'/'O' (1.0),\n",
              "  'Token[69]: \"and\"'/'O' (1.0),\n",
              "  'Token[70]: \"i\"'/'O' (1.0),\n",
              "  'Token[71]: \"have\"'/'O' (1.0),\n",
              "  'Token[72]: \"to\"'/'O' (1.0),\n",
              "  'Token[73]: \"take\"'/'O' (1.0),\n",
              "  'Token[74]: \"it\"'/'O' (1.0),\n",
              "  'Token[75]: \".\"'/'O' (1.0),\n",
              "  'Token[76]: \"every\"'/'O' (1.0),\n",
              "  'Token[77]: \"day\"'/'O' (1.0),\n",
              "  'Token[78]: \"for\"'/'O' (1.0),\n",
              "  'Token[79]: \"the\"'/'O' (1.0),\n",
              "  'Token[80]: \"next\"'/'O' (1.0),\n",
              "  'Token[81]: \"month\"'/'O' (1.0),\n",
              "  'Token[82]: \"to\"'/'O' (1.0),\n",
              "  'Token[83]: \"see\"'/'O' (1.0),\n",
              "  'Token[84]: \"how\"'/'O' (1.0),\n",
              "  'Token[85]: \"i\"'/'O' (1.0),\n",
              "  'Token[86]: \"get\"'/'O' (1.0),\n",
              "  'Token[87]: \"on\"'/'O' (1.0),\n",
              "  'Token[88]: \",\"'/'O' (1.0),\n",
              "  'Token[89]: \"here\"'/'O' (1.0),\n",
              "  'Token[90]: \"goes\"'/'O' (1.0),\n",
              "  'Token[91]: \".\"'/'O' (1.0),\n",
              "  'Token[92]: \"so\"'/'O' (1.0),\n",
              "  'Token[93]: \"far\"'/'O' (1.0),\n",
              "  'Token[94]: \"its\"'/'O' (1.0),\n",
              "  'Token[95]: \"been\"'/'O' (1.0),\n",
              "  'Token[96]: \"very\"'/'O' (1.0),\n",
              "  'Token[97]: \"good\"'/'O' (1.0),\n",
              "  'Token[98]: \",\"'/'O' (1.0),\n",
              "  'Token[99]: \"pains\"'/'S-Pain' (1.0),\n",
              "  'Token[100]: \"almost\"'/'O' (1.0),\n",
              "  'Token[101]: \"gone\"'/'O' (1.0),\n",
              "  'Token[102]: \",\"'/'O' (1.0),\n",
              "  'Token[103]: \"but\"'/'O' (1.0),\n",
              "  'Token[104]: \"i\"'/'O' (1.0),\n",
              "  'Token[105]: \"feel\"'/'B-Malaise' (1.0),\n",
              "  'Token[106]: \"a\"'/'I-Malaise' (1.0),\n",
              "  'Token[107]: \"bit\"'/'I-Malaise' (1.0),\n",
              "  'Token[108]: \"weird\"'/'E-Malaise' (1.0),\n",
              "  'Token[109]: \",\"'/'O' (1.0),\n",
              "  'Token[110]: \"didnt\"'/'O' (1.0),\n",
              "  'Token[111]: \"have\"'/'O' (1.0),\n",
              "  'Token[112]: \"that\"'/'O' (1.0),\n",
              "  'Token[113]: \"when\"'/'O' (1.0),\n",
              "  'Token[114]: \"on\"'/'O' (1.0),\n",
              "  'Token[115]: \"50\"'/'O' (1.0),\n",
              "  'Token[116]: \".\"'/'O' (1.0)]}"
            ]
          },
          "execution_count": 11,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "temp.annotation_layers"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "id": "4CsGKKB6tGik"
      },
      "outputs": [],
      "source": [
        "\n",
        "# define your NER tags\n",
        "label_dict = {\n",
        "    'ADR': 'ADR',\n",
        "    'Drug': 'Drug',\n",
        "    'Finding': 'Finding',\n",
        "    'Symptom': 'Symptom',\n",
        "    'Disease': 'Disease'\n",
        "}"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QmNsy8AltGik",
        "outputId": "6345a9b1-43e8-40e7-b7f1-a36c4c287799"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 14:55:55,778 Filtering empty sentences\n",
            "2023-04-21 14:55:55,812 Corpus: 948 train + 119 dev + 119 test sentences\n",
            "2023-04-21 14:55:55,814 Computing label dictionary. Progress:\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "948it [00:00, 6369.44it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 14:55:55,971 Dictionary created for label 'ner' with 22 values: O (seen 79057 times), I-ADR (seen 4449 times), E-ADR (seen 3506 times), B-ADR (seen 3349 times), S-ADR (seen 1363 times), S-Drug (seen 1286 times), B-Finding (seen 200 times), E-Finding (seen 200 times), I-Finding (seen 149 times), B-Disease (seen 119 times), E-Disease (seen 119 times), E-Symptom (seen 114 times), S-Disease (seen 112 times), S-Symptom (seen 112 times), E-Drug (seen 111 times), B-Symptom (seen 110 times), B-Drug (seen 110 times), S-Finding (seen 109 times), I-Symptom (seen 104 times), I-Drug (seen 50 times)\n",
            "False\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "True True False 1024\n",
            "{'is_token_embedding': True, 'is_document_embedding': False, 'allow_long_sentences': False, 'tokenizer': BertTokenizerFast(name_or_path='SpanBERT/spanbert-large-cased', vocab_size=28996, model_max_length=512, is_fast=True, padding_side='right', truncation_side='right', special_tokens={'unk_token': '[UNK]', 'sep_token': '[SEP]', 'pad_token': '[PAD]', 'cls_token': '[CLS]', 'mask_token': '[MASK]', 'additional_special_tokens': ['[FLERT]']}, clean_up_tokenization_spaces=True), 'context_length': 64, 'context_dropout': 0.5, 'respect_document_boundaries': True, 'truncate': True, 'stride': 0, 'embedding_length': 1024, 'name': 'transformer-SpanBERT/spanbert-large-cased', 'fine_tune': True, 'use_lang_emb': False, 'force_max_length': True, 'feature_extractor': None, 'use_context_separator': True, 'needs_manual_ocr': False}\n",
            "BertTokenizerFast(name_or_path='SpanBERT/spanbert-large-cased', vocab_size=28996, model_max_length=512, is_fast=True, padding_side='right', truncation_side='right', special_tokens={'unk_token': '[UNK]', 'sep_token': '[SEP]', 'pad_token': '[PAD]', 'cls_token': '[CLS]', 'mask_token': '[MASK]', 'additional_special_tokens': ['[FLERT]']}, clean_up_tokenization_spaces=True)\n",
            "2023-04-21 14:56:03,820 SequenceTagger predicts: Dictionary with 22 tags: <unk>, O, I-ADR, E-ADR, B-ADR, S-ADR, S-Drug, B-Finding, E-Finding, I-Finding, B-Disease, E-Disease, E-Symptom, S-Disease, S-Symptom, E-Drug, B-Symptom, B-Drug, S-Finding, I-Symptom, I-Drug, I-Disease\n",
            "True\n",
            "False\n"
          ]
        }
      ],
      "source": [
        "from flair.data import Corpus\n",
        "\n",
        "from flair.embeddings import WordEmbeddings, FlairEmbeddings, StackedEmbeddings, TransformerWordEmbeddings\n",
        "from flair.models import SequenceTagger\n",
        "from flair.trainers import ModelTrainer\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "from torch.optim import AdamW, SGD\n",
        "\n",
        "from flair.visual.training_curves import Plotter\n",
        "plotter = Plotter()\n",
        "\n",
        "\n",
        "\n",
        "train_docs, test_docs = train_test_split(docs, test_size=0.2, random_state=321)\n",
        "dev_docs, test_docs = train_test_split(test_docs, test_size=0.5, random_state=123)\n",
        "\n",
        "corpus = Corpus(train=train_docs, dev=dev_docs, test=test_docs) # Corpus: 948 train + 119 dev + 119 test sentences\n",
        "corpus.filter_empty_sentences()\n",
        "# # create a tag dictionary\n",
        "labels = ['ADR', 'Drug', 'Finding', 'Symptom', 'Disease']\n",
        "# tag_dictionary = Dictionary(add_unk=True)\n",
        "# for label in labels:\n",
        "#     tag_dictionary.add_item(label)\n",
        "tag_dictionary = corpus.make_label_dictionary(label_type='ner')\n",
        "print(tag_dictionary.span_labels)\n",
        "\n",
        "# create a stacked embedding using GloVe and Flair embeddings\n",
        "# stacked_embeddings = StackedEmbeddings([\n",
        "#     WordEmbeddings('glove',\n",
        "#                    force_cpu=False,\n",
        "#                 #    embedding_length=1024,\n",
        "#                   #  fine_tune=True,\n",
        "#                    ),\n",
        "#     # TransformerWordEmbeddings('emilyalsentzer/Bio_ClinicalBERT', layers=\"-1\",\n",
        "#     #                                    subtoken_pooling=\"first\",\n",
        "#     #                                    is_document_embedding=True,\n",
        "#     #                                    fine_tune=True,\n",
        "#     #                                    use_context=True,\n",
        "#     #                                    max_seq_length=1024,\n",
        "#     #                                    ),\n",
        "#     FlairEmbeddings('news-forward',\n",
        "#                     # fine_tune=True\n",
        "#                     ),\n",
        "#     FlairEmbeddings('news-backward',\n",
        "#                     # fine_tune=True\n",
        "#                     ),\n",
        "# ])\n",
        "# https://github.com/flairNLP/flair/blob/master/resources/docs/embeddings/FLAIR_EMBEDDINGS.md\n",
        "\n",
        "\n",
        "embeddings = TransformerWordEmbeddings(model='SpanBERT/spanbert-large-cased',\n",
        "                                       layers=\"-1\",\n",
        "                                       subtoken_pooling=\"first\",\n",
        "                                       fine_tune=True,\n",
        "                                       use_context=True,\n",
        "                                       force_max_length=True,\n",
        "                                      #  truncate=True, # not accepted here\n",
        "                                      # #  is_document_embedding=True, # False is default\n",
        "                                      #  allow_long_sentences=True, # True is default; doesnt go through\n",
        "                                      #  embedding_length=512,\n",
        "                                       )\n",
        "# embeddings.allow_long_sentences = False\n",
        "embeddings.truncate = True\n",
        "# embeddings.embedding_length_internal = 512\n",
        "print(embeddings.force_max_length,embeddings.truncate,embeddings.allow_long_sentences,embeddings.embedding_length_internal)\n",
        "embeddings.tokenizer.model_max_length=512\n",
        "print(embeddings.to_args())\n",
        "print(embeddings.tokenizer)\n",
        "\n",
        "\n",
        "# create a SequenceTagger object and specify the tag dictionary and embeddings to use\n",
        "tagger = SequenceTagger(hidden_size=256,\n",
        "                        # embeddings=stacked_embeddings,\n",
        "                        embeddings=embeddings,\n",
        "                        tag_dictionary=tag_dictionary,\n",
        "                        tag_type='ner',\n",
        "                        use_crf=False,\n",
        "                        use_rnn=False,\n",
        "                        # reproject_embeddings=False\n",
        "                        )\n",
        "tagger.predict_spans = False\n",
        "\n",
        "print(tagger._determine_if_span_prediction_problem(tagger.label_dictionary))\n",
        "print(tagger.predict_spans)\n",
        "\n",
        "# https://github.com/flairNLP/flair/blob/master/flair/trainers/trainer.py\n",
        "# create a ModelTrainer object and train the tagger\n",
        "trainer = ModelTrainer(tagger, corpus)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "id": "fh0zU2jHAA4J"
      },
      "outputs": [],
      "source": [
        "# !tensorboard --logdir=runs\n",
        "# http://localhost:6006/"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UKTMzxV8zF9o",
        "outputId": "26302b7d-82a5-48f5-9b72-f5dcd664d63e"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 14:56:03,858 tensorboard logging path is runs/tb_logs_v2_spanbert-large-cased\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.9/dist-packages/flair/trainers/trainer.py:69: UserWarning: There should be no best model saved at epoch 1 except there is a model from previous trainings in your training folder. All previous best models will be deleted.\n",
            "  warnings.warn(\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 14:56:04,404 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 14:56:04,409 Model: \"SequenceTagger(\n",
            "  (embeddings): TransformerWordEmbeddings(\n",
            "    (model): BertModel(\n",
            "      (embeddings): BertEmbeddings(\n",
            "        (word_embeddings): Embedding(28997, 1024)\n",
            "        (position_embeddings): Embedding(512, 1024)\n",
            "        (token_type_embeddings): Embedding(2, 1024)\n",
            "        (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
            "        (dropout): Dropout(p=0.1, inplace=False)\n",
            "      )\n",
            "      (encoder): BertEncoder(\n",
            "        (layer): ModuleList(\n",
            "          (0-23): 24 x BertLayer(\n",
            "            (attention): BertAttention(\n",
            "              (self): BertSelfAttention(\n",
            "                (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
            "                (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
            "                (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
            "                (dropout): Dropout(p=0.1, inplace=False)\n",
            "              )\n",
            "              (output): BertSelfOutput(\n",
            "                (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
            "                (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
            "                (dropout): Dropout(p=0.1, inplace=False)\n",
            "              )\n",
            "            )\n",
            "            (intermediate): BertIntermediate(\n",
            "              (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
            "              (intermediate_act_fn): GELUActivation()\n",
            "            )\n",
            "            (output): BertOutput(\n",
            "              (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
            "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
            "              (dropout): Dropout(p=0.1, inplace=False)\n",
            "            )\n",
            "          )\n",
            "        )\n",
            "      )\n",
            "      (pooler): BertPooler(\n",
            "        (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
            "        (activation): Tanh()\n",
            "      )\n",
            "    )\n",
            "  )\n",
            "  (locked_dropout): LockedDropout(p=0.5)\n",
            "  (embedding2nn): Linear(in_features=1024, out_features=1024, bias=True)\n",
            "  (linear): Linear(in_features=1024, out_features=22, bias=True)\n",
            "  (loss_function): CrossEntropyLoss()\n",
            ")\"\n",
            "2023-04-21 14:56:04,411 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 14:56:04,412 Corpus: \"Corpus: 948 train + 119 dev + 119 test sentences\"\n",
            "2023-04-21 14:56:04,414 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 14:56:04,415 Parameters:\n",
            "2023-04-21 14:56:04,416  - learning_rate: \"0.000005\"\n",
            "2023-04-21 14:56:04,418  - mini_batch_size: \"15\"\n",
            "2023-04-21 14:56:04,420  - patience: \"1\"\n",
            "2023-04-21 14:56:04,422  - anneal_factor: \"0.5\"\n",
            "2023-04-21 14:56:04,423  - max_epochs: \"100\"\n",
            "2023-04-21 14:56:04,425  - shuffle: \"True\"\n",
            "2023-04-21 14:56:04,427  - train_with_dev: \"False\"\n",
            "2023-04-21 14:56:04,428  - batch_growth_annealing: \"False\"\n",
            "2023-04-21 14:56:04,430 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 14:56:04,432 Model training base path: \"flair_ner_model_v2_spanbert-large-cased\"\n",
            "2023-04-21 14:56:04,434 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 14:56:04,435 Device: cuda:0\n",
            "2023-04-21 14:56:04,437 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 14:56:04,438 Embeddings storage mode: gpu\n",
            "2023-04-21 14:56:04,440 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 14:56:11,919 epoch 1 - iter 6/64 - loss 2.95545405 - time (sec): 7.48 - samples/sec: 1283.94 - lr: 0.000005\n",
            "2023-04-21 14:56:18,563 epoch 1 - iter 12/64 - loss 2.78797174 - time (sec): 14.12 - samples/sec: 1258.58 - lr: 0.000005\n",
            "2023-04-21 14:56:25,230 epoch 1 - iter 18/64 - loss 2.62722416 - time (sec): 20.79 - samples/sec: 1267.13 - lr: 0.000005\n",
            "2023-04-21 14:56:31,934 epoch 1 - iter 24/64 - loss 2.48559876 - time (sec): 27.49 - samples/sec: 1281.12 - lr: 0.000005\n",
            "2023-04-21 14:56:38,571 epoch 1 - iter 30/64 - loss 2.36355950 - time (sec): 34.13 - samples/sec: 1280.74 - lr: 0.000005\n",
            "2023-04-21 14:56:45,301 epoch 1 - iter 36/64 - loss 2.26519843 - time (sec): 40.86 - samples/sec: 1293.32 - lr: 0.000005\n",
            "2023-04-21 14:56:52,147 epoch 1 - iter 42/64 - loss 2.17359554 - time (sec): 47.71 - samples/sec: 1311.39 - lr: 0.000005\n",
            "2023-04-21 14:56:58,841 epoch 1 - iter 48/64 - loss 2.09276437 - time (sec): 54.40 - samples/sec: 1312.88 - lr: 0.000005\n",
            "2023-04-21 14:57:05,520 epoch 1 - iter 54/64 - loss 2.02546519 - time (sec): 61.08 - samples/sec: 1316.65 - lr: 0.000005\n",
            "2023-04-21 14:57:12,226 epoch 1 - iter 60/64 - loss 1.96351024 - time (sec): 67.78 - samples/sec: 1319.16 - lr: 0.000005\n",
            "2023-04-21 14:57:15,935 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 14:57:15,936 EPOCH 1 done: loss 1.9307 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:25<00:00,  2.51it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 14:57:41,554 Evaluating as a multi-label problem: True\n",
            "2023-04-21 14:57:45,261 TRAIN : loss 1.3106995820999146 - f1-score (weighted avg)  0.0001\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.51it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 14:57:48,607 Evaluating as a multi-label problem: False\n",
            "2023-04-21 14:57:48,669 DEV : loss 1.305672526359558 - f1-score (weighted avg)  0.0\n",
            "2023-04-21 14:57:48,690 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 14:57:48,694 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 14:57:57,982 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 14:58:04,798 epoch 2 - iter 6/64 - loss 1.33603308 - time (sec): 6.81 - samples/sec: 1341.90 - lr: 0.000005\n",
            "2023-04-21 14:58:11,556 epoch 2 - iter 12/64 - loss 1.28680588 - time (sec): 13.57 - samples/sec: 1371.24 - lr: 0.000005\n",
            "2023-04-21 14:58:18,213 epoch 2 - iter 18/64 - loss 1.29745374 - time (sec): 20.23 - samples/sec: 1304.91 - lr: 0.000005\n",
            "2023-04-21 14:58:24,893 epoch 2 - iter 24/64 - loss 1.26851612 - time (sec): 26.91 - samples/sec: 1297.57 - lr: 0.000005\n",
            "2023-04-21 14:58:31,584 epoch 2 - iter 30/64 - loss 1.24326455 - time (sec): 33.60 - samples/sec: 1303.06 - lr: 0.000005\n",
            "2023-04-21 14:58:38,439 epoch 2 - iter 36/64 - loss 1.23533549 - time (sec): 40.45 - samples/sec: 1323.91 - lr: 0.000005\n",
            "2023-04-21 14:58:45,129 epoch 2 - iter 42/64 - loss 1.22119349 - time (sec): 47.14 - samples/sec: 1328.16 - lr: 0.000005\n",
            "2023-04-21 14:58:51,916 epoch 2 - iter 48/64 - loss 1.20769671 - time (sec): 53.93 - samples/sec: 1334.22 - lr: 0.000005\n",
            "2023-04-21 14:58:58,649 epoch 2 - iter 54/64 - loss 1.18915053 - time (sec): 60.66 - samples/sec: 1344.97 - lr: 0.000005\n",
            "2023-04-21 14:59:05,299 epoch 2 - iter 60/64 - loss 1.17406988 - time (sec): 67.31 - samples/sec: 1343.96 - lr: 0.000005\n",
            "2023-04-21 14:59:08,902 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 14:59:08,904 EPOCH 2 done: loss 1.1659 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.59it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 14:59:33,712 Evaluating as a multi-label problem: True\n",
            "2023-04-21 14:59:36,339 TRAIN : loss 0.9801037907600403 - f1-score (weighted avg)  0.0001\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 14:59:39,726 Evaluating as a multi-label problem: False\n",
            "2023-04-21 14:59:39,792 DEV : loss 0.9744789004325867 - f1-score (weighted avg)  0.0\n",
            "2023-04-21 14:59:39,813 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 14:59:39,818 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 14:59:50,306 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 14:59:56,970 epoch 3 - iter 6/64 - loss 0.98570872 - time (sec): 6.66 - samples/sec: 1211.88 - lr: 0.000005\n",
            "2023-04-21 15:00:03,657 epoch 3 - iter 12/64 - loss 0.94653296 - time (sec): 13.35 - samples/sec: 1255.32 - lr: 0.000005\n",
            "2023-04-21 15:00:10,322 epoch 3 - iter 18/64 - loss 0.92393274 - time (sec): 20.01 - samples/sec: 1264.04 - lr: 0.000005\n",
            "2023-04-21 15:00:17,050 epoch 3 - iter 24/64 - loss 0.92115932 - time (sec): 26.74 - samples/sec: 1329.25 - lr: 0.000005\n",
            "2023-04-21 15:00:23,839 epoch 3 - iter 30/64 - loss 0.92733796 - time (sec): 33.53 - samples/sec: 1324.63 - lr: 0.000005\n",
            "2023-04-21 15:00:30,756 epoch 3 - iter 36/64 - loss 0.94415302 - time (sec): 40.45 - samples/sec: 1347.97 - lr: 0.000005\n",
            "2023-04-21 15:00:37,473 epoch 3 - iter 42/64 - loss 0.93546052 - time (sec): 47.16 - samples/sec: 1372.13 - lr: 0.000005\n",
            "2023-04-21 15:00:44,293 epoch 3 - iter 48/64 - loss 0.93753027 - time (sec): 53.98 - samples/sec: 1370.30 - lr: 0.000005\n",
            "2023-04-21 15:00:50,936 epoch 3 - iter 54/64 - loss 0.93455612 - time (sec): 60.63 - samples/sec: 1353.51 - lr: 0.000005\n",
            "2023-04-21 15:00:57,640 epoch 3 - iter 60/64 - loss 0.93562043 - time (sec): 67.33 - samples/sec: 1341.46 - lr: 0.000005\n",
            "2023-04-21 15:01:01,261 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:01:01,263 EPOCH 3 done: loss 0.9359 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:25<00:00,  2.48it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:01:27,144 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:01:29,785 TRAIN : loss 0.8688046932220459 - f1-score (weighted avg)  0.0001\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:01:33,166 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:01:33,233 DEV : loss 0.8662349581718445 - f1-score (weighted avg)  0.0\n",
            "2023-04-21 15:01:33,255 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:01:33,260 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:01:43,527 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:01:50,294 epoch 4 - iter 6/64 - loss 0.83476886 - time (sec): 6.76 - samples/sec: 1352.25 - lr: 0.000005\n",
            "2023-04-21 15:01:57,006 epoch 4 - iter 12/64 - loss 0.85781116 - time (sec): 13.48 - samples/sec: 1288.42 - lr: 0.000005\n",
            "2023-04-21 15:02:03,805 epoch 4 - iter 18/64 - loss 0.86118018 - time (sec): 20.28 - samples/sec: 1318.61 - lr: 0.000005\n",
            "2023-04-21 15:02:10,526 epoch 4 - iter 24/64 - loss 0.84471599 - time (sec): 27.00 - samples/sec: 1336.41 - lr: 0.000005\n",
            "2023-04-21 15:02:17,366 epoch 4 - iter 30/64 - loss 0.85396684 - time (sec): 33.84 - samples/sec: 1344.77 - lr: 0.000005\n",
            "2023-04-21 15:02:24,021 epoch 4 - iter 36/64 - loss 0.84239807 - time (sec): 40.49 - samples/sec: 1326.85 - lr: 0.000005\n",
            "2023-04-21 15:02:30,729 epoch 4 - iter 42/64 - loss 0.82701857 - time (sec): 47.20 - samples/sec: 1322.50 - lr: 0.000005\n",
            "2023-04-21 15:02:37,449 epoch 4 - iter 48/64 - loss 0.80933841 - time (sec): 53.92 - samples/sec: 1314.56 - lr: 0.000005\n",
            "2023-04-21 15:02:44,155 epoch 4 - iter 54/64 - loss 0.79338685 - time (sec): 60.63 - samples/sec: 1327.87 - lr: 0.000005\n",
            "2023-04-21 15:02:50,886 epoch 4 - iter 60/64 - loss 0.77967875 - time (sec): 67.36 - samples/sec: 1339.34 - lr: 0.000005\n",
            "2023-04-21 15:02:54,493 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:02:54,495 EPOCH 4 done: loss 0.7731 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:25<00:00,  2.50it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:03:20,258 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:03:22,898 TRAIN : loss 0.6235568523406982 - f1-score (weighted avg)  0.0001\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.49it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:03:26,268 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:03:26,342 DEV : loss 0.6301484107971191 - f1-score (weighted avg)  0.0\n",
            "2023-04-21 15:03:26,363 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:03:26,367 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:03:37,587 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:03:44,305 epoch 5 - iter 6/64 - loss 0.64189917 - time (sec): 6.72 - samples/sec: 1271.25 - lr: 0.000005\n",
            "2023-04-21 15:03:51,071 epoch 5 - iter 12/64 - loss 0.64165897 - time (sec): 13.48 - samples/sec: 1321.05 - lr: 0.000005\n",
            "2023-04-21 15:03:57,915 epoch 5 - iter 18/64 - loss 0.67020753 - time (sec): 20.33 - samples/sec: 1272.20 - lr: 0.000005\n",
            "2023-04-21 15:04:04,706 epoch 5 - iter 24/64 - loss 0.65889479 - time (sec): 27.12 - samples/sec: 1319.02 - lr: 0.000005\n",
            "2023-04-21 15:04:11,412 epoch 5 - iter 30/64 - loss 0.64833348 - time (sec): 33.82 - samples/sec: 1306.16 - lr: 0.000005\n",
            "2023-04-21 15:04:18,099 epoch 5 - iter 36/64 - loss 0.63512917 - time (sec): 40.51 - samples/sec: 1314.62 - lr: 0.000005\n",
            "2023-04-21 15:04:24,792 epoch 5 - iter 42/64 - loss 0.61736846 - time (sec): 47.20 - samples/sec: 1342.86 - lr: 0.000005\n",
            "2023-04-21 15:04:31,554 epoch 5 - iter 48/64 - loss 0.60767246 - time (sec): 53.97 - samples/sec: 1364.82 - lr: 0.000005\n",
            "2023-04-21 15:04:38,267 epoch 5 - iter 54/64 - loss 0.60187066 - time (sec): 60.68 - samples/sec: 1358.83 - lr: 0.000005\n",
            "2023-04-21 15:04:44,937 epoch 5 - iter 60/64 - loss 0.59980848 - time (sec): 67.35 - samples/sec: 1336.55 - lr: 0.000005\n",
            "2023-04-21 15:04:48,576 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:04:48,577 EPOCH 5 done: loss 0.6000 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.59it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:05:13,411 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:05:17,203 TRAIN : loss 0.527245819568634 - f1-score (weighted avg)  0.2602\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.47it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:05:20,607 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:05:20,679 DEV : loss 0.5404419302940369 - f1-score (weighted avg)  0.2488\n",
            "2023-04-21 15:05:20,701 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:05:20,705 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:05:30,835 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:05:37,558 epoch 6 - iter 6/64 - loss 0.56890558 - time (sec): 6.72 - samples/sec: 1175.72 - lr: 0.000005\n",
            "2023-04-21 15:05:44,214 epoch 6 - iter 12/64 - loss 0.55692723 - time (sec): 13.38 - samples/sec: 1284.38 - lr: 0.000005\n",
            "2023-04-21 15:05:51,008 epoch 6 - iter 18/64 - loss 0.53666869 - time (sec): 20.17 - samples/sec: 1324.53 - lr: 0.000005\n",
            "2023-04-21 15:05:57,708 epoch 6 - iter 24/64 - loss 0.52640795 - time (sec): 26.87 - samples/sec: 1360.73 - lr: 0.000005\n",
            "2023-04-21 15:06:04,458 epoch 6 - iter 30/64 - loss 0.52323205 - time (sec): 33.62 - samples/sec: 1386.41 - lr: 0.000005\n",
            "2023-04-21 15:06:11,228 epoch 6 - iter 36/64 - loss 0.51630565 - time (sec): 40.39 - samples/sec: 1386.08 - lr: 0.000005\n",
            "2023-04-21 15:06:17,973 epoch 6 - iter 42/64 - loss 0.51275128 - time (sec): 47.13 - samples/sec: 1372.84 - lr: 0.000005\n",
            "2023-04-21 15:06:24,629 epoch 6 - iter 48/64 - loss 0.51667890 - time (sec): 53.79 - samples/sec: 1345.11 - lr: 0.000005\n",
            "2023-04-21 15:06:31,363 epoch 6 - iter 54/64 - loss 0.51743633 - time (sec): 60.52 - samples/sec: 1337.82 - lr: 0.000005\n",
            "2023-04-21 15:06:38,223 epoch 6 - iter 60/64 - loss 0.52199920 - time (sec): 67.38 - samples/sec: 1336.70 - lr: 0.000005\n",
            "2023-04-21 15:06:41,845 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:06:41,847 EPOCH 6 done: loss 0.5207 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.59it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:07:06,714 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:07:10,556 TRAIN : loss 0.4584510326385498 - f1-score (weighted avg)  0.4274\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:07:13,948 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:07:14,014 DEV : loss 0.48442408442497253 - f1-score (weighted avg)  0.397\n",
            "2023-04-21 15:07:14,037 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:07:14,041 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:07:25,754 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:07:32,466 epoch 7 - iter 6/64 - loss 0.46197731 - time (sec): 6.71 - samples/sec: 1288.18 - lr: 0.000005\n",
            "2023-04-21 15:07:39,204 epoch 7 - iter 12/64 - loss 0.45065659 - time (sec): 13.45 - samples/sec: 1259.86 - lr: 0.000005\n",
            "2023-04-21 15:07:45,906 epoch 7 - iter 18/64 - loss 0.46680728 - time (sec): 20.15 - samples/sec: 1283.22 - lr: 0.000005\n",
            "2023-04-21 15:07:52,718 epoch 7 - iter 24/64 - loss 0.46761208 - time (sec): 26.96 - samples/sec: 1329.89 - lr: 0.000005\n",
            "2023-04-21 15:07:59,399 epoch 7 - iter 30/64 - loss 0.46536453 - time (sec): 33.64 - samples/sec: 1310.42 - lr: 0.000005\n",
            "2023-04-21 15:08:06,158 epoch 7 - iter 36/64 - loss 0.45927458 - time (sec): 40.40 - samples/sec: 1321.46 - lr: 0.000005\n",
            "2023-04-21 15:08:13,024 epoch 7 - iter 42/64 - loss 0.46790830 - time (sec): 47.27 - samples/sec: 1345.88 - lr: 0.000005\n",
            "2023-04-21 15:08:19,737 epoch 7 - iter 48/64 - loss 0.45891154 - time (sec): 53.98 - samples/sec: 1364.59 - lr: 0.000005\n",
            "2023-04-21 15:08:26,439 epoch 7 - iter 54/64 - loss 0.45257839 - time (sec): 60.68 - samples/sec: 1361.11 - lr: 0.000005\n",
            "2023-04-21 15:08:33,132 epoch 7 - iter 60/64 - loss 0.45354723 - time (sec): 67.38 - samples/sec: 1347.65 - lr: 0.000005\n",
            "2023-04-21 15:08:36,753 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:08:36,754 EPOCH 7 done: loss 0.4545 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.59it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:09:01,613 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:09:04,469 TRAIN : loss 0.3886544406414032 - f1-score (weighted avg)  0.5102\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:09:07,869 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:09:07,936 DEV : loss 0.4294235408306122 - f1-score (weighted avg)  0.4372\n",
            "2023-04-21 15:09:07,957 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:09:07,961 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:09:18,245 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:09:24,959 epoch 8 - iter 6/64 - loss 0.40626142 - time (sec): 6.71 - samples/sec: 1379.26 - lr: 0.000005\n",
            "2023-04-21 15:09:31,643 epoch 8 - iter 12/64 - loss 0.40624815 - time (sec): 13.40 - samples/sec: 1347.69 - lr: 0.000005\n",
            "2023-04-21 15:09:38,582 epoch 8 - iter 18/64 - loss 0.42924466 - time (sec): 20.34 - samples/sec: 1373.17 - lr: 0.000005\n",
            "2023-04-21 15:09:45,270 epoch 8 - iter 24/64 - loss 0.41517193 - time (sec): 27.02 - samples/sec: 1336.39 - lr: 0.000005\n",
            "2023-04-21 15:09:52,045 epoch 8 - iter 30/64 - loss 0.41377540 - time (sec): 33.80 - samples/sec: 1338.91 - lr: 0.000005\n",
            "2023-04-21 15:09:58,715 epoch 8 - iter 36/64 - loss 0.41083631 - time (sec): 40.47 - samples/sec: 1314.62 - lr: 0.000005\n",
            "2023-04-21 15:10:05,410 epoch 8 - iter 42/64 - loss 0.40334432 - time (sec): 47.16 - samples/sec: 1304.47 - lr: 0.000005\n",
            "2023-04-21 15:10:12,119 epoch 8 - iter 48/64 - loss 0.39945385 - time (sec): 53.87 - samples/sec: 1298.38 - lr: 0.000005\n",
            "2023-04-21 15:10:18,867 epoch 8 - iter 54/64 - loss 0.39454222 - time (sec): 60.62 - samples/sec: 1319.47 - lr: 0.000005\n",
            "2023-04-21 15:10:25,604 epoch 8 - iter 60/64 - loss 0.39155544 - time (sec): 67.36 - samples/sec: 1327.88 - lr: 0.000005\n",
            "2023-04-21 15:10:29,289 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:10:29,291 EPOCH 8 done: loss 0.3867 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:25<00:00,  2.48it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:10:55,214 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:10:59,087 TRAIN : loss 0.3294545114040375 - f1-score (weighted avg)  0.6105\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.50it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:11:02,455 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:11:02,529 DEV : loss 0.3896329402923584 - f1-score (weighted avg)  0.5163\n",
            "2023-04-21 15:11:02,550 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:11:02,555 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:11:12,270 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:11:18,968 epoch 9 - iter 6/64 - loss 0.34897522 - time (sec): 6.70 - samples/sec: 1214.03 - lr: 0.000005\n",
            "2023-04-21 15:11:25,752 epoch 9 - iter 12/64 - loss 0.34330558 - time (sec): 13.48 - samples/sec: 1359.85 - lr: 0.000005\n",
            "2023-04-21 15:11:32,596 epoch 9 - iter 18/64 - loss 0.36722092 - time (sec): 20.32 - samples/sec: 1397.39 - lr: 0.000005\n",
            "2023-04-21 15:11:39,320 epoch 9 - iter 24/64 - loss 0.36030415 - time (sec): 27.05 - samples/sec: 1345.56 - lr: 0.000005\n",
            "2023-04-21 15:11:46,047 epoch 9 - iter 30/64 - loss 0.34910217 - time (sec): 33.77 - samples/sec: 1371.14 - lr: 0.000005\n",
            "2023-04-21 15:11:52,758 epoch 9 - iter 36/64 - loss 0.34559488 - time (sec): 40.49 - samples/sec: 1359.88 - lr: 0.000005\n",
            "2023-04-21 15:11:59,559 epoch 9 - iter 42/64 - loss 0.34846104 - time (sec): 47.29 - samples/sec: 1365.81 - lr: 0.000005\n",
            "2023-04-21 15:12:06,256 epoch 9 - iter 48/64 - loss 0.34823649 - time (sec): 53.98 - samples/sec: 1354.15 - lr: 0.000005\n",
            "2023-04-21 15:12:12,918 epoch 9 - iter 54/64 - loss 0.34470516 - time (sec): 60.65 - samples/sec: 1344.78 - lr: 0.000005\n",
            "2023-04-21 15:12:19,605 epoch 9 - iter 60/64 - loss 0.34082423 - time (sec): 67.33 - samples/sec: 1344.41 - lr: 0.000005\n",
            "2023-04-21 15:12:23,222 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:12:23,224 EPOCH 9 done: loss 0.3399 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.57it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:12:48,278 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:12:52,290 TRAIN : loss 0.28942999243736267 - f1-score (weighted avg)  0.7048\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.49it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:12:55,658 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:12:55,728 DEV : loss 0.3671266436576843 - f1-score (weighted avg)  0.604\n",
            "2023-04-21 15:12:55,749 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:12:55,753 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:13:05,346 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:13:12,067 epoch 10 - iter 6/64 - loss 0.31070939 - time (sec): 6.72 - samples/sec: 1343.28 - lr: 0.000005\n",
            "2023-04-21 15:13:18,850 epoch 10 - iter 12/64 - loss 0.31347606 - time (sec): 13.50 - samples/sec: 1400.77 - lr: 0.000005\n",
            "2023-04-21 15:13:25,508 epoch 10 - iter 18/64 - loss 0.31279389 - time (sec): 20.16 - samples/sec: 1344.68 - lr: 0.000005\n",
            "2023-04-21 15:13:32,178 epoch 10 - iter 24/64 - loss 0.30872141 - time (sec): 26.83 - samples/sec: 1322.83 - lr: 0.000005\n",
            "2023-04-21 15:13:38,847 epoch 10 - iter 30/64 - loss 0.29998883 - time (sec): 33.50 - samples/sec: 1316.90 - lr: 0.000005\n",
            "2023-04-21 15:13:45,678 epoch 10 - iter 36/64 - loss 0.31661349 - time (sec): 40.33 - samples/sec: 1327.75 - lr: 0.000005\n",
            "2023-04-21 15:13:52,347 epoch 10 - iter 42/64 - loss 0.31182225 - time (sec): 47.00 - samples/sec: 1324.88 - lr: 0.000005\n",
            "2023-04-21 15:13:59,155 epoch 10 - iter 48/64 - loss 0.31310675 - time (sec): 53.81 - samples/sec: 1332.87 - lr: 0.000005\n",
            "2023-04-21 15:14:05,850 epoch 10 - iter 54/64 - loss 0.30902653 - time (sec): 60.50 - samples/sec: 1323.63 - lr: 0.000005\n",
            "2023-04-21 15:14:12,592 epoch 10 - iter 60/64 - loss 0.30474333 - time (sec): 67.24 - samples/sec: 1323.75 - lr: 0.000005\n",
            "2023-04-21 15:14:16,237 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:14:16,238 EPOCH 10 done: loss 0.3027 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.57it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:14:41,230 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:14:43,972 TRAIN : loss 0.2574145793914795 - f1-score (weighted avg)  0.7427\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.50it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:14:47,338 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:14:47,408 DEV : loss 0.35706833004951477 - f1-score (weighted avg)  0.6324\n",
            "2023-04-21 15:14:47,428 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:14:47,432 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:14:57,441 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:15:04,306 epoch 11 - iter 6/64 - loss 0.30564832 - time (sec): 6.86 - samples/sec: 1455.17 - lr: 0.000005\n",
            "2023-04-21 15:15:11,018 epoch 11 - iter 12/64 - loss 0.28194098 - time (sec): 13.57 - samples/sec: 1344.14 - lr: 0.000005\n",
            "2023-04-21 15:15:17,679 epoch 11 - iter 18/64 - loss 0.27977207 - time (sec): 20.24 - samples/sec: 1320.58 - lr: 0.000005\n",
            "2023-04-21 15:15:24,343 epoch 11 - iter 24/64 - loss 0.28029822 - time (sec): 26.90 - samples/sec: 1334.91 - lr: 0.000005\n",
            "2023-04-21 15:15:31,077 epoch 11 - iter 30/64 - loss 0.27001259 - time (sec): 33.63 - samples/sec: 1345.72 - lr: 0.000005\n",
            "2023-04-21 15:15:37,801 epoch 11 - iter 36/64 - loss 0.26540711 - time (sec): 40.36 - samples/sec: 1332.73 - lr: 0.000005\n",
            "2023-04-21 15:15:45,521 epoch 11 - iter 42/64 - loss 0.26706012 - time (sec): 48.08 - samples/sec: 1310.09 - lr: 0.000005\n",
            "2023-04-21 15:15:52,216 epoch 11 - iter 48/64 - loss 0.26577149 - time (sec): 54.77 - samples/sec: 1325.16 - lr: 0.000005\n",
            "2023-04-21 15:15:59,090 epoch 11 - iter 54/64 - loss 0.27888925 - time (sec): 61.65 - samples/sec: 1334.66 - lr: 0.000005\n",
            "2023-04-21 15:16:05,770 epoch 11 - iter 60/64 - loss 0.27481982 - time (sec): 68.33 - samples/sec: 1320.35 - lr: 0.000005\n",
            "2023-04-21 15:16:09,403 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:16:09,404 EPOCH 11 done: loss 0.2729 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:16:34,360 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:16:38,244 TRAIN : loss 0.23602625727653503 - f1-score (weighted avg)  0.7522\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.47it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:16:41,643 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:16:41,714 DEV : loss 0.3434412181377411 - f1-score (weighted avg)  0.6371\n",
            "2023-04-21 15:16:41,737 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:16:41,741 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:16:51,187 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:16:57,899 epoch 12 - iter 6/64 - loss 0.23931080 - time (sec): 6.71 - samples/sec: 1292.80 - lr: 0.000005\n",
            "2023-04-21 15:17:04,583 epoch 12 - iter 12/64 - loss 0.23948579 - time (sec): 13.39 - samples/sec: 1247.28 - lr: 0.000005\n",
            "2023-04-21 15:17:11,540 epoch 12 - iter 18/64 - loss 0.30055898 - time (sec): 20.35 - samples/sec: 1296.71 - lr: 0.000005\n",
            "2023-04-21 15:17:18,215 epoch 12 - iter 24/64 - loss 0.29330624 - time (sec): 27.03 - samples/sec: 1274.22 - lr: 0.000005\n",
            "2023-04-21 15:17:25,006 epoch 12 - iter 30/64 - loss 0.28472687 - time (sec): 33.82 - samples/sec: 1294.48 - lr: 0.000005\n",
            "2023-04-21 15:17:31,691 epoch 12 - iter 36/64 - loss 0.27282329 - time (sec): 40.50 - samples/sec: 1314.88 - lr: 0.000005\n",
            "2023-04-21 15:17:38,435 epoch 12 - iter 42/64 - loss 0.26293625 - time (sec): 47.25 - samples/sec: 1328.82 - lr: 0.000005\n",
            "2023-04-21 15:17:45,156 epoch 12 - iter 48/64 - loss 0.25648813 - time (sec): 53.97 - samples/sec: 1348.39 - lr: 0.000005\n",
            "2023-04-21 15:17:51,866 epoch 12 - iter 54/64 - loss 0.25329595 - time (sec): 60.68 - samples/sec: 1352.69 - lr: 0.000005\n",
            "2023-04-21 15:17:58,612 epoch 12 - iter 60/64 - loss 0.25098320 - time (sec): 67.42 - samples/sec: 1346.70 - lr: 0.000005\n",
            "2023-04-21 15:18:02,222 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:18:02,223 EPOCH 12 done: loss 0.2499 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:18:27,176 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:18:31,130 TRAIN : loss 0.21483080089092255 - f1-score (weighted avg)  0.7673\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.47it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:18:34,529 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:18:34,600 DEV : loss 0.33488979935646057 - f1-score (weighted avg)  0.6326\n",
            "2023-04-21 15:18:34,621 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:18:34,625 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:18:44,903 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:18:51,595 epoch 13 - iter 6/64 - loss 0.20847284 - time (sec): 6.69 - samples/sec: 1360.34 - lr: 0.000005\n",
            "2023-04-21 15:18:58,409 epoch 13 - iter 12/64 - loss 0.21190554 - time (sec): 13.50 - samples/sec: 1390.51 - lr: 0.000005\n",
            "2023-04-21 15:19:05,075 epoch 13 - iter 18/64 - loss 0.21141206 - time (sec): 20.17 - samples/sec: 1362.21 - lr: 0.000005\n",
            "2023-04-21 15:19:11,786 epoch 13 - iter 24/64 - loss 0.21950593 - time (sec): 26.88 - samples/sec: 1333.47 - lr: 0.000005\n",
            "2023-04-21 15:19:18,462 epoch 13 - iter 30/64 - loss 0.21810185 - time (sec): 33.56 - samples/sec: 1305.93 - lr: 0.000005\n",
            "2023-04-21 15:19:25,141 epoch 13 - iter 36/64 - loss 0.21779225 - time (sec): 40.24 - samples/sec: 1285.75 - lr: 0.000005\n",
            "2023-04-21 15:19:31,867 epoch 13 - iter 42/64 - loss 0.21725054 - time (sec): 46.96 - samples/sec: 1309.86 - lr: 0.000005\n",
            "2023-04-21 15:19:38,688 epoch 13 - iter 48/64 - loss 0.22308326 - time (sec): 53.78 - samples/sec: 1339.21 - lr: 0.000005\n",
            "2023-04-21 15:19:45,576 epoch 13 - iter 54/64 - loss 0.23427603 - time (sec): 60.67 - samples/sec: 1350.44 - lr: 0.000005\n",
            "2023-04-21 15:19:52,264 epoch 13 - iter 60/64 - loss 0.23158435 - time (sec): 67.36 - samples/sec: 1331.13 - lr: 0.000005\n",
            "2023-04-21 15:19:55,889 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:19:55,891 EPOCH 13 done: loss 0.2307 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.57it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:20:20,915 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:20:23,627 TRAIN : loss 0.20050732791423798 - f1-score (weighted avg)  0.7741\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.47it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:20:28,029 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:20:28,099 DEV : loss 0.3598255515098572 - f1-score (weighted avg)  0.6272\n",
            "2023-04-21 15:20:28,121 BAD EPOCHS (no improvement): 1\n",
            "2023-04-21 15:20:28,125 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:20:34,875 epoch 14 - iter 6/64 - loss 0.22765681 - time (sec): 6.75 - samples/sec: 1340.33 - lr: 0.000005\n",
            "2023-04-21 15:20:41,581 epoch 14 - iter 12/64 - loss 0.20539010 - time (sec): 13.45 - samples/sec: 1341.77 - lr: 0.000005\n",
            "2023-04-21 15:20:48,265 epoch 14 - iter 18/64 - loss 0.21240241 - time (sec): 20.14 - samples/sec: 1322.82 - lr: 0.000005\n",
            "2023-04-21 15:20:54,958 epoch 14 - iter 24/64 - loss 0.21182007 - time (sec): 26.83 - samples/sec: 1325.92 - lr: 0.000005\n",
            "2023-04-21 15:21:01,647 epoch 14 - iter 30/64 - loss 0.20986812 - time (sec): 33.52 - samples/sec: 1323.85 - lr: 0.000005\n",
            "2023-04-21 15:21:08,388 epoch 14 - iter 36/64 - loss 0.21239919 - time (sec): 40.26 - samples/sec: 1307.28 - lr: 0.000005\n",
            "2023-04-21 15:21:15,163 epoch 14 - iter 42/64 - loss 0.21391425 - time (sec): 47.04 - samples/sec: 1315.50 - lr: 0.000005\n",
            "2023-04-21 15:21:21,844 epoch 14 - iter 48/64 - loss 0.21103679 - time (sec): 53.72 - samples/sec: 1307.08 - lr: 0.000005\n",
            "2023-04-21 15:21:28,565 epoch 14 - iter 54/64 - loss 0.20774361 - time (sec): 60.44 - samples/sec: 1332.59 - lr: 0.000005\n",
            "2023-04-21 15:21:35,418 epoch 14 - iter 60/64 - loss 0.21908082 - time (sec): 67.29 - samples/sec: 1323.26 - lr: 0.000005\n",
            "2023-04-21 15:21:39,067 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:21:39,069 EPOCH 14 done: loss 0.2172 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:22:04,052 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:22:08,006 TRAIN : loss 0.18805454671382904 - f1-score (weighted avg)  0.7824\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.47it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:22:11,406 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:22:11,475 DEV : loss 0.33287736773490906 - f1-score (weighted avg)  0.6272\n",
            "2023-04-21 15:22:11,496 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:22:11,500 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:22:21,192 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:22:27,936 epoch 15 - iter 6/64 - loss 0.19823804 - time (sec): 6.74 - samples/sec: 1254.69 - lr: 0.000005\n",
            "2023-04-21 15:22:34,831 epoch 15 - iter 12/64 - loss 0.21445966 - time (sec): 13.64 - samples/sec: 1303.03 - lr: 0.000005\n",
            "2023-04-21 15:22:41,535 epoch 15 - iter 18/64 - loss 0.20060552 - time (sec): 20.34 - samples/sec: 1278.93 - lr: 0.000005\n",
            "2023-04-21 15:22:48,446 epoch 15 - iter 24/64 - loss 0.22524695 - time (sec): 27.25 - samples/sec: 1313.30 - lr: 0.000005\n",
            "2023-04-21 15:22:55,143 epoch 15 - iter 30/64 - loss 0.22088385 - time (sec): 33.95 - samples/sec: 1290.53 - lr: 0.000005\n",
            "2023-04-21 15:23:01,852 epoch 15 - iter 36/64 - loss 0.21542389 - time (sec): 40.66 - samples/sec: 1289.88 - lr: 0.000005\n",
            "2023-04-21 15:23:08,580 epoch 15 - iter 42/64 - loss 0.21216370 - time (sec): 47.39 - samples/sec: 1326.27 - lr: 0.000005\n",
            "2023-04-21 15:23:15,269 epoch 15 - iter 48/64 - loss 0.20702645 - time (sec): 54.08 - samples/sec: 1319.61 - lr: 0.000005\n",
            "2023-04-21 15:23:22,014 epoch 15 - iter 54/64 - loss 0.20750851 - time (sec): 60.82 - samples/sec: 1323.78 - lr: 0.000005\n",
            "2023-04-21 15:23:28,742 epoch 15 - iter 60/64 - loss 0.20513266 - time (sec): 67.55 - samples/sec: 1321.39 - lr: 0.000005\n",
            "2023-04-21 15:23:32,369 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:23:32,371 EPOCH 15 done: loss 0.2066 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.57it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:23:57,402 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:24:01,352 TRAIN : loss 0.18198886513710022 - f1-score (weighted avg)  0.7836\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:24:04,740 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:24:04,813 DEV : loss 0.3354378342628479 - f1-score (weighted avg)  0.6286\n",
            "2023-04-21 15:24:04,834 BAD EPOCHS (no improvement): 1\n",
            "2023-04-21 15:24:04,839 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:24:11,545 epoch 16 - iter 6/64 - loss 0.18717252 - time (sec): 6.70 - samples/sec: 1332.29 - lr: 0.000005\n",
            "2023-04-21 15:24:18,275 epoch 16 - iter 12/64 - loss 0.19987973 - time (sec): 13.43 - samples/sec: 1329.22 - lr: 0.000005\n",
            "2023-04-21 15:24:25,112 epoch 16 - iter 18/64 - loss 0.22905987 - time (sec): 20.27 - samples/sec: 1325.12 - lr: 0.000005\n",
            "2023-04-21 15:24:31,836 epoch 16 - iter 24/64 - loss 0.21463242 - time (sec): 27.00 - samples/sec: 1372.20 - lr: 0.000005\n",
            "2023-04-21 15:24:38,612 epoch 16 - iter 30/64 - loss 0.21857972 - time (sec): 33.77 - samples/sec: 1368.32 - lr: 0.000005\n",
            "2023-04-21 15:24:45,304 epoch 16 - iter 36/64 - loss 0.21282021 - time (sec): 40.46 - samples/sec: 1346.24 - lr: 0.000005\n",
            "2023-04-21 15:24:51,981 epoch 16 - iter 42/64 - loss 0.20849913 - time (sec): 47.14 - samples/sec: 1338.82 - lr: 0.000005\n",
            "2023-04-21 15:24:58,691 epoch 16 - iter 48/64 - loss 0.20254008 - time (sec): 53.85 - samples/sec: 1343.47 - lr: 0.000005\n",
            "2023-04-21 15:25:05,407 epoch 16 - iter 54/64 - loss 0.19622723 - time (sec): 60.57 - samples/sec: 1346.93 - lr: 0.000005\n",
            "2023-04-21 15:25:12,135 epoch 16 - iter 60/64 - loss 0.19685994 - time (sec): 67.29 - samples/sec: 1351.67 - lr: 0.000005\n",
            "2023-04-21 15:25:15,736 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:25:15,738 EPOCH 16 done: loss 0.1952 - lr 0.000005\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:25:40,642 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:25:43,364 TRAIN : loss 0.1684892773628235 - f1-score (weighted avg)  0.7916\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:04<00:00,  1.89it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:25:47,770 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:25:47,840 DEV : loss 0.3369958996772766 - f1-score (weighted avg)  0.6381\n",
            "2023-04-21 15:25:47,862 Epoch    16: reducing learning rate of group 0 to 2.5000e-06.\n",
            "2023-04-21 15:25:47,864 BAD EPOCHS (no improvement): 2\n",
            "2023-04-21 15:25:47,868 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:25:54,526 epoch 17 - iter 6/64 - loss 0.16495951 - time (sec): 6.66 - samples/sec: 1206.02 - lr: 0.000003\n",
            "2023-04-21 15:26:01,300 epoch 17 - iter 12/64 - loss 0.16613855 - time (sec): 13.43 - samples/sec: 1348.09 - lr: 0.000003\n",
            "2023-04-21 15:26:08,035 epoch 17 - iter 18/64 - loss 0.16684907 - time (sec): 20.17 - samples/sec: 1364.06 - lr: 0.000003\n",
            "2023-04-21 15:26:14,720 epoch 17 - iter 24/64 - loss 0.16697916 - time (sec): 26.85 - samples/sec: 1341.27 - lr: 0.000003\n",
            "2023-04-21 15:26:21,422 epoch 17 - iter 30/64 - loss 0.16840034 - time (sec): 33.55 - samples/sec: 1321.31 - lr: 0.000003\n",
            "2023-04-21 15:26:28,108 epoch 17 - iter 36/64 - loss 0.16634961 - time (sec): 40.24 - samples/sec: 1293.49 - lr: 0.000003\n",
            "2023-04-21 15:26:34,799 epoch 17 - iter 42/64 - loss 0.16592353 - time (sec): 46.93 - samples/sec: 1293.93 - lr: 0.000003\n",
            "2023-04-21 15:26:41,521 epoch 17 - iter 48/64 - loss 0.16532292 - time (sec): 53.65 - samples/sec: 1307.66 - lr: 0.000003\n",
            "2023-04-21 15:26:48,236 epoch 17 - iter 54/64 - loss 0.16666040 - time (sec): 60.37 - samples/sec: 1315.01 - lr: 0.000003\n",
            "2023-04-21 15:26:55,034 epoch 17 - iter 60/64 - loss 0.17048448 - time (sec): 67.17 - samples/sec: 1332.88 - lr: 0.000003\n",
            "2023-04-21 15:26:58,834 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:26:58,835 EPOCH 17 done: loss 0.1823 - lr 0.000003\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:27:23,736 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:27:27,616 TRAIN : loss 0.16132107377052307 - f1-score (weighted avg)  0.7948\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:27:30,998 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:27:31,066 DEV : loss 0.328365683555603 - f1-score (weighted avg)  0.6335\n",
            "2023-04-21 15:27:31,087 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:27:31,091 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:27:41,038 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:27:47,745 epoch 18 - iter 6/64 - loss 0.18048938 - time (sec): 6.70 - samples/sec: 1191.81 - lr: 0.000003\n",
            "2023-04-21 15:27:54,460 epoch 18 - iter 12/64 - loss 0.16760785 - time (sec): 13.42 - samples/sec: 1265.14 - lr: 0.000003\n",
            "2023-04-21 15:28:01,148 epoch 18 - iter 18/64 - loss 0.16707931 - time (sec): 20.11 - samples/sec: 1265.70 - lr: 0.000003\n",
            "2023-04-21 15:28:07,859 epoch 18 - iter 24/64 - loss 0.16179826 - time (sec): 26.82 - samples/sec: 1296.92 - lr: 0.000003\n",
            "2023-04-21 15:28:14,679 epoch 18 - iter 30/64 - loss 0.16717468 - time (sec): 33.64 - samples/sec: 1351.62 - lr: 0.000003\n",
            "2023-04-21 15:28:21,397 epoch 18 - iter 36/64 - loss 0.16393035 - time (sec): 40.36 - samples/sec: 1353.16 - lr: 0.000003\n",
            "2023-04-21 15:28:28,273 epoch 18 - iter 42/64 - loss 0.17973566 - time (sec): 47.23 - samples/sec: 1338.07 - lr: 0.000003\n",
            "2023-04-21 15:28:34,990 epoch 18 - iter 48/64 - loss 0.18070871 - time (sec): 53.95 - samples/sec: 1336.00 - lr: 0.000003\n",
            "2023-04-21 15:28:41,762 epoch 18 - iter 54/64 - loss 0.17883435 - time (sec): 60.72 - samples/sec: 1326.56 - lr: 0.000003\n",
            "2023-04-21 15:28:48,487 epoch 18 - iter 60/64 - loss 0.17962788 - time (sec): 67.45 - samples/sec: 1320.98 - lr: 0.000003\n",
            "2023-04-21 15:28:52,147 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:28:52,149 EPOCH 18 done: loss 0.1770 - lr 0.000003\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:25<00:00,  2.56it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:29:17,322 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:29:21,445 TRAIN : loss 0.15461769700050354 - f1-score (weighted avg)  0.7998\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.47it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:29:24,855 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:29:24,927 DEV : loss 0.33131614327430725 - f1-score (weighted avg)  0.6339\n",
            "2023-04-21 15:29:24,949 BAD EPOCHS (no improvement): 1\n",
            "2023-04-21 15:29:24,954 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:29:31,667 epoch 19 - iter 6/64 - loss 0.16640238 - time (sec): 6.71 - samples/sec: 1127.11 - lr: 0.000003\n",
            "2023-04-21 15:29:38,359 epoch 19 - iter 12/64 - loss 0.16096480 - time (sec): 13.40 - samples/sec: 1200.65 - lr: 0.000003\n",
            "2023-04-21 15:29:45,106 epoch 19 - iter 18/64 - loss 0.15055039 - time (sec): 20.15 - samples/sec: 1282.26 - lr: 0.000003\n",
            "2023-04-21 15:29:51,887 epoch 19 - iter 24/64 - loss 0.15291720 - time (sec): 26.93 - samples/sec: 1322.24 - lr: 0.000003\n",
            "2023-04-21 15:29:58,562 epoch 19 - iter 30/64 - loss 0.15255188 - time (sec): 33.61 - samples/sec: 1291.73 - lr: 0.000003\n",
            "2023-04-21 15:30:05,275 epoch 19 - iter 36/64 - loss 0.15586052 - time (sec): 40.32 - samples/sec: 1329.92 - lr: 0.000003\n",
            "2023-04-21 15:30:11,942 epoch 19 - iter 42/64 - loss 0.15605982 - time (sec): 46.99 - samples/sec: 1306.99 - lr: 0.000003\n",
            "2023-04-21 15:30:18,749 epoch 19 - iter 48/64 - loss 0.16083071 - time (sec): 53.79 - samples/sec: 1328.06 - lr: 0.000003\n",
            "2023-04-21 15:30:25,611 epoch 19 - iter 54/64 - loss 0.17529423 - time (sec): 60.66 - samples/sec: 1341.54 - lr: 0.000003\n",
            "2023-04-21 15:30:32,312 epoch 19 - iter 60/64 - loss 0.17214415 - time (sec): 67.36 - samples/sec: 1342.84 - lr: 0.000003\n",
            "2023-04-21 15:30:35,910 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:30:35,912 EPOCH 19 done: loss 0.1707 - lr 0.000003\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.59it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:31:00,784 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:31:04,681 TRAIN : loss 0.15391214191913605 - f1-score (weighted avg)  0.7992\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.47it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:31:08,089 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:31:08,159 DEV : loss 0.32777702808380127 - f1-score (weighted avg)  0.6338\n",
            "2023-04-21 15:31:08,181 BAD EPOCHS (no improvement): 0\n",
            "2023-04-21 15:31:08,185 saving best model\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:31:18,204 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:31:24,960 epoch 20 - iter 6/64 - loss 0.14271126 - time (sec): 6.75 - samples/sec: 1306.94 - lr: 0.000003\n",
            "2023-04-21 15:31:31,639 epoch 20 - iter 12/64 - loss 0.15585020 - time (sec): 13.43 - samples/sec: 1278.05 - lr: 0.000003\n",
            "2023-04-21 15:31:38,307 epoch 20 - iter 18/64 - loss 0.15594884 - time (sec): 20.10 - samples/sec: 1283.08 - lr: 0.000003\n",
            "2023-04-21 15:31:44,988 epoch 20 - iter 24/64 - loss 0.15455928 - time (sec): 26.78 - samples/sec: 1300.83 - lr: 0.000003\n",
            "2023-04-21 15:31:51,726 epoch 20 - iter 30/64 - loss 0.14662873 - time (sec): 33.52 - samples/sec: 1309.15 - lr: 0.000003\n",
            "2023-04-21 15:31:58,563 epoch 20 - iter 36/64 - loss 0.15358836 - time (sec): 40.36 - samples/sec: 1328.99 - lr: 0.000003\n",
            "2023-04-21 15:32:05,317 epoch 20 - iter 42/64 - loss 0.15448412 - time (sec): 47.11 - samples/sec: 1326.00 - lr: 0.000003\n",
            "2023-04-21 15:32:12,033 epoch 20 - iter 48/64 - loss 0.15307218 - time (sec): 53.83 - samples/sec: 1346.24 - lr: 0.000003\n",
            "2023-04-21 15:32:18,699 epoch 20 - iter 54/64 - loss 0.15185725 - time (sec): 60.49 - samples/sec: 1325.74 - lr: 0.000003\n",
            "2023-04-21 15:32:25,547 epoch 20 - iter 60/64 - loss 0.16494715 - time (sec): 67.34 - samples/sec: 1325.23 - lr: 0.000003\n",
            "2023-04-21 15:32:29,201 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:32:29,203 EPOCH 20 done: loss 0.1648 - lr 0.000003\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.57it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:32:54,279 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:32:57,007 TRAIN : loss 0.14464759826660156 - f1-score (weighted avg)  0.8127\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:33:00,420 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:33:00,489 DEV : loss 0.3325453996658325 - f1-score (weighted avg)  0.6441\n",
            "2023-04-21 15:33:00,510 BAD EPOCHS (no improvement): 1\n",
            "2023-04-21 15:33:00,514 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:33:07,365 epoch 21 - iter 6/64 - loss 0.17217422 - time (sec): 6.85 - samples/sec: 1402.68 - lr: 0.000003\n",
            "2023-04-21 15:33:14,058 epoch 21 - iter 12/64 - loss 0.15335210 - time (sec): 13.54 - samples/sec: 1343.61 - lr: 0.000003\n",
            "2023-04-21 15:33:20,752 epoch 21 - iter 18/64 - loss 0.16032620 - time (sec): 20.24 - samples/sec: 1292.54 - lr: 0.000003\n",
            "2023-04-21 15:33:27,534 epoch 21 - iter 24/64 - loss 0.16199702 - time (sec): 27.02 - samples/sec: 1312.96 - lr: 0.000003\n",
            "2023-04-21 15:33:34,214 epoch 21 - iter 30/64 - loss 0.15932650 - time (sec): 33.70 - samples/sec: 1297.45 - lr: 0.000003\n",
            "2023-04-21 15:33:40,918 epoch 21 - iter 36/64 - loss 0.15510179 - time (sec): 40.40 - samples/sec: 1313.69 - lr: 0.000003\n",
            "2023-04-21 15:33:47,602 epoch 21 - iter 42/64 - loss 0.15181310 - time (sec): 47.09 - samples/sec: 1307.75 - lr: 0.000003\n",
            "2023-04-21 15:33:55,478 epoch 21 - iter 48/64 - loss 0.16738824 - time (sec): 54.96 - samples/sec: 1303.93 - lr: 0.000003\n",
            "2023-04-21 15:34:02,258 epoch 21 - iter 54/64 - loss 0.16491422 - time (sec): 61.74 - samples/sec: 1319.00 - lr: 0.000003\n",
            "2023-04-21 15:34:08,975 epoch 21 - iter 60/64 - loss 0.16135538 - time (sec): 68.46 - samples/sec: 1310.51 - lr: 0.000003\n",
            "2023-04-21 15:34:12,614 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:34:12,616 EPOCH 21 done: loss 0.1594 - lr 0.000003\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.56it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:34:37,741 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:34:41,915 TRAIN : loss 0.14014171063899994 - f1-score (weighted avg)  0.8244\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.42it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:34:45,393 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:34:45,468 DEV : loss 0.3306526839733124 - f1-score (weighted avg)  0.6489\n",
            "2023-04-21 15:34:45,490 Epoch    21: reducing learning rate of group 0 to 1.2500e-06.\n",
            "2023-04-21 15:34:45,492 BAD EPOCHS (no improvement): 2\n",
            "2023-04-21 15:34:45,497 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:34:52,203 epoch 22 - iter 6/64 - loss 0.11353155 - time (sec): 6.70 - samples/sec: 1242.85 - lr: 0.000001\n",
            "2023-04-21 15:34:58,902 epoch 22 - iter 12/64 - loss 0.13448182 - time (sec): 13.40 - samples/sec: 1227.41 - lr: 0.000001\n",
            "2023-04-21 15:35:05,708 epoch 22 - iter 18/64 - loss 0.14159776 - time (sec): 20.21 - samples/sec: 1338.95 - lr: 0.000001\n",
            "2023-04-21 15:35:12,525 epoch 22 - iter 24/64 - loss 0.14168050 - time (sec): 27.03 - samples/sec: 1335.16 - lr: 0.000001\n",
            "2023-04-21 15:35:19,390 epoch 22 - iter 30/64 - loss 0.16542412 - time (sec): 33.89 - samples/sec: 1333.94 - lr: 0.000001\n",
            "2023-04-21 15:35:26,115 epoch 22 - iter 36/64 - loss 0.16091051 - time (sec): 40.62 - samples/sec: 1334.13 - lr: 0.000001\n",
            "2023-04-21 15:35:32,825 epoch 22 - iter 42/64 - loss 0.15711232 - time (sec): 47.33 - samples/sec: 1358.22 - lr: 0.000001\n",
            "2023-04-21 15:35:39,553 epoch 22 - iter 48/64 - loss 0.15634466 - time (sec): 54.06 - samples/sec: 1361.37 - lr: 0.000001\n",
            "2023-04-21 15:35:46,238 epoch 22 - iter 54/64 - loss 0.15706366 - time (sec): 60.74 - samples/sec: 1335.41 - lr: 0.000001\n",
            "2023-04-21 15:35:52,952 epoch 22 - iter 60/64 - loss 0.15703066 - time (sec): 67.45 - samples/sec: 1341.75 - lr: 0.000001\n",
            "2023-04-21 15:35:56,545 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:35:56,547 EPOCH 22 done: loss 0.1557 - lr 0.000001\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.59it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:36:21,442 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:36:25,401 TRAIN : loss 0.1375254988670349 - f1-score (weighted avg)  0.828\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:36:28,796 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:36:28,865 DEV : loss 0.3337743282318115 - f1-score (weighted avg)  0.6521\n",
            "2023-04-21 15:36:28,886 BAD EPOCHS (no improvement): 1\n",
            "2023-04-21 15:36:28,891 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:36:35,552 epoch 23 - iter 6/64 - loss 0.13558789 - time (sec): 6.66 - samples/sec: 1312.54 - lr: 0.000001\n",
            "2023-04-21 15:36:42,221 epoch 23 - iter 12/64 - loss 0.12713098 - time (sec): 13.33 - samples/sec: 1327.09 - lr: 0.000001\n",
            "2023-04-21 15:36:48,927 epoch 23 - iter 18/64 - loss 0.13801037 - time (sec): 20.03 - samples/sec: 1331.78 - lr: 0.000001\n",
            "2023-04-21 15:36:55,745 epoch 23 - iter 24/64 - loss 0.16672700 - time (sec): 26.85 - samples/sec: 1357.05 - lr: 0.000001\n",
            "2023-04-21 15:37:02,444 epoch 23 - iter 30/64 - loss 0.15977061 - time (sec): 33.55 - samples/sec: 1335.01 - lr: 0.000001\n",
            "2023-04-21 15:37:09,142 epoch 23 - iter 36/64 - loss 0.15520714 - time (sec): 40.25 - samples/sec: 1326.90 - lr: 0.000001\n",
            "2023-04-21 15:37:15,821 epoch 23 - iter 42/64 - loss 0.15361032 - time (sec): 46.93 - samples/sec: 1309.55 - lr: 0.000001\n",
            "2023-04-21 15:37:22,530 epoch 23 - iter 48/64 - loss 0.15088705 - time (sec): 53.64 - samples/sec: 1325.79 - lr: 0.000001\n",
            "2023-04-21 15:37:29,291 epoch 23 - iter 54/64 - loss 0.15074181 - time (sec): 60.40 - samples/sec: 1332.77 - lr: 0.000001\n",
            "2023-04-21 15:37:36,078 epoch 23 - iter 60/64 - loss 0.15116510 - time (sec): 67.19 - samples/sec: 1335.19 - lr: 0.000001\n",
            "2023-04-21 15:37:39,820 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:37:39,821 EPOCH 23 done: loss 0.1532 - lr 0.000001\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:38:04,762 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:38:07,485 TRAIN : loss 0.13586443662643433 - f1-score (weighted avg)  0.83\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:38:10,870 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:38:10,938 DEV : loss 0.3327925503253937 - f1-score (weighted avg)  0.6515\n",
            "2023-04-21 15:38:10,960 Epoch    23: reducing learning rate of group 0 to 6.2500e-07.\n",
            "2023-04-21 15:38:10,962 BAD EPOCHS (no improvement): 2\n",
            "2023-04-21 15:38:10,966 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:38:17,845 epoch 24 - iter 6/64 - loss 0.25127253 - time (sec): 6.88 - samples/sec: 1427.47 - lr: 0.000001\n",
            "2023-04-21 15:38:24,572 epoch 24 - iter 12/64 - loss 0.19208961 - time (sec): 13.60 - samples/sec: 1379.04 - lr: 0.000001\n",
            "2023-04-21 15:38:31,243 epoch 24 - iter 18/64 - loss 0.17137901 - time (sec): 20.28 - samples/sec: 1324.48 - lr: 0.000001\n",
            "2023-04-21 15:38:38,007 epoch 24 - iter 24/64 - loss 0.16074100 - time (sec): 27.04 - samples/sec: 1356.01 - lr: 0.000001\n",
            "2023-04-21 15:38:44,683 epoch 24 - iter 30/64 - loss 0.15324728 - time (sec): 33.72 - samples/sec: 1336.07 - lr: 0.000001\n",
            "2023-04-21 15:38:51,396 epoch 24 - iter 36/64 - loss 0.15127211 - time (sec): 40.43 - samples/sec: 1347.02 - lr: 0.000001\n",
            "2023-04-21 15:38:58,122 epoch 24 - iter 42/64 - loss 0.14790780 - time (sec): 47.15 - samples/sec: 1349.43 - lr: 0.000001\n",
            "2023-04-21 15:39:04,937 epoch 24 - iter 48/64 - loss 0.15455465 - time (sec): 53.97 - samples/sec: 1353.74 - lr: 0.000001\n",
            "2023-04-21 15:39:11,626 epoch 24 - iter 54/64 - loss 0.15328476 - time (sec): 60.66 - samples/sec: 1347.80 - lr: 0.000001\n",
            "2023-04-21 15:39:18,348 epoch 24 - iter 60/64 - loss 0.15321095 - time (sec): 67.38 - samples/sec: 1340.58 - lr: 0.000001\n",
            "2023-04-21 15:39:21,975 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:39:21,976 EPOCH 24 done: loss 0.1518 - lr 0.000001\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:25<00:00,  2.48it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:39:47,911 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:39:51,818 TRAIN : loss 0.13387742638587952 - f1-score (weighted avg)  0.833\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.49it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:39:55,192 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:39:55,260 DEV : loss 0.3336173892021179 - f1-score (weighted avg)  0.6512\n",
            "2023-04-21 15:39:55,282 BAD EPOCHS (no improvement): 1\n",
            "2023-04-21 15:39:55,286 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:40:02,020 epoch 25 - iter 6/64 - loss 0.12142794 - time (sec): 6.73 - samples/sec: 1479.73 - lr: 0.000001\n",
            "2023-04-21 15:40:08,724 epoch 25 - iter 12/64 - loss 0.13350003 - time (sec): 13.44 - samples/sec: 1381.95 - lr: 0.000001\n",
            "2023-04-21 15:40:15,443 epoch 25 - iter 18/64 - loss 0.13331750 - time (sec): 20.16 - samples/sec: 1369.07 - lr: 0.000001\n",
            "2023-04-21 15:40:22,259 epoch 25 - iter 24/64 - loss 0.14511150 - time (sec): 26.97 - samples/sec: 1366.18 - lr: 0.000001\n",
            "2023-04-21 15:40:29,167 epoch 25 - iter 30/64 - loss 0.16453028 - time (sec): 33.88 - samples/sec: 1389.93 - lr: 0.000001\n",
            "2023-04-21 15:40:35,844 epoch 25 - iter 36/64 - loss 0.15999685 - time (sec): 40.56 - samples/sec: 1366.03 - lr: 0.000001\n",
            "2023-04-21 15:40:42,499 epoch 25 - iter 42/64 - loss 0.15601516 - time (sec): 47.21 - samples/sec: 1335.29 - lr: 0.000001\n",
            "2023-04-21 15:40:49,173 epoch 25 - iter 48/64 - loss 0.15673578 - time (sec): 53.89 - samples/sec: 1329.21 - lr: 0.000001\n",
            "2023-04-21 15:40:55,923 epoch 25 - iter 54/64 - loss 0.15255153 - time (sec): 60.64 - samples/sec: 1319.11 - lr: 0.000001\n",
            "2023-04-21 15:41:02,644 epoch 25 - iter 60/64 - loss 0.14974444 - time (sec): 67.36 - samples/sec: 1339.45 - lr: 0.000001\n",
            "2023-04-21 15:41:06,260 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:41:06,262 EPOCH 25 done: loss 0.1498 - lr 0.000001\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:41:31,223 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:41:33,942 TRAIN : loss 0.13351976871490479 - f1-score (weighted avg)  0.833\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:04<00:00,  1.89it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:41:38,350 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:41:38,423 DEV : loss 0.3345133662223816 - f1-score (weighted avg)  0.6535\n",
            "2023-04-21 15:41:38,445 Epoch    25: reducing learning rate of group 0 to 3.1250e-07.\n",
            "2023-04-21 15:41:38,447 BAD EPOCHS (no improvement): 2\n",
            "2023-04-21 15:41:38,450 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:41:45,159 epoch 26 - iter 6/64 - loss 0.11773994 - time (sec): 6.71 - samples/sec: 1360.33 - lr: 0.000000\n",
            "2023-04-21 15:41:51,871 epoch 26 - iter 12/64 - loss 0.12587114 - time (sec): 13.42 - samples/sec: 1415.30 - lr: 0.000000\n",
            "2023-04-21 15:41:58,590 epoch 26 - iter 18/64 - loss 0.12769392 - time (sec): 20.14 - samples/sec: 1433.45 - lr: 0.000000\n",
            "2023-04-21 15:42:05,282 epoch 26 - iter 24/64 - loss 0.13214790 - time (sec): 26.83 - samples/sec: 1407.39 - lr: 0.000000\n",
            "2023-04-21 15:42:11,985 epoch 26 - iter 30/64 - loss 0.13731815 - time (sec): 33.53 - samples/sec: 1386.15 - lr: 0.000000\n",
            "2023-04-21 15:42:18,653 epoch 26 - iter 36/64 - loss 0.13903807 - time (sec): 40.20 - samples/sec: 1333.46 - lr: 0.000000\n",
            "2023-04-21 15:42:25,346 epoch 26 - iter 42/64 - loss 0.13575056 - time (sec): 46.89 - samples/sec: 1334.46 - lr: 0.000000\n",
            "2023-04-21 15:42:32,361 epoch 26 - iter 48/64 - loss 0.15653867 - time (sec): 53.91 - samples/sec: 1358.36 - lr: 0.000000\n",
            "2023-04-21 15:42:39,081 epoch 26 - iter 54/64 - loss 0.15189844 - time (sec): 60.63 - samples/sec: 1341.22 - lr: 0.000000\n",
            "2023-04-21 15:42:45,813 epoch 26 - iter 60/64 - loss 0.14850387 - time (sec): 67.36 - samples/sec: 1337.98 - lr: 0.000000\n",
            "2023-04-21 15:42:49,442 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:42:49,444 EPOCH 26 done: loss 0.1485 - lr 0.000000\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:43:14,409 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:43:18,319 TRAIN : loss 0.13248226046562195 - f1-score (weighted avg)  0.8361\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.49it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:43:21,701 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:43:21,770 DEV : loss 0.3357246220111847 - f1-score (weighted avg)  0.6544\n",
            "2023-04-21 15:43:21,791 BAD EPOCHS (no improvement): 1\n",
            "2023-04-21 15:43:21,796 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:43:28,513 epoch 27 - iter 6/64 - loss 0.12384726 - time (sec): 6.72 - samples/sec: 1268.38 - lr: 0.000000\n",
            "2023-04-21 15:43:35,216 epoch 27 - iter 12/64 - loss 0.11927437 - time (sec): 13.42 - samples/sec: 1283.95 - lr: 0.000000\n",
            "2023-04-21 15:43:42,095 epoch 27 - iter 18/64 - loss 0.16818964 - time (sec): 20.30 - samples/sec: 1281.74 - lr: 0.000000\n",
            "2023-04-21 15:43:48,798 epoch 27 - iter 24/64 - loss 0.15824031 - time (sec): 27.00 - samples/sec: 1267.20 - lr: 0.000000\n",
            "2023-04-21 15:43:55,578 epoch 27 - iter 30/64 - loss 0.16413943 - time (sec): 33.78 - samples/sec: 1300.19 - lr: 0.000000\n",
            "2023-04-21 15:44:02,274 epoch 27 - iter 36/64 - loss 0.15790644 - time (sec): 40.48 - samples/sec: 1319.63 - lr: 0.000000\n",
            "2023-04-21 15:44:08,972 epoch 27 - iter 42/64 - loss 0.15184415 - time (sec): 47.17 - samples/sec: 1323.18 - lr: 0.000000\n",
            "2023-04-21 15:44:15,721 epoch 27 - iter 48/64 - loss 0.14950490 - time (sec): 53.92 - samples/sec: 1324.43 - lr: 0.000000\n",
            "2023-04-21 15:44:22,459 epoch 27 - iter 54/64 - loss 0.14713808 - time (sec): 60.66 - samples/sec: 1333.19 - lr: 0.000000\n",
            "2023-04-21 15:44:29,173 epoch 27 - iter 60/64 - loss 0.14756096 - time (sec): 67.38 - samples/sec: 1341.04 - lr: 0.000000\n",
            "2023-04-21 15:44:32,795 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:44:32,797 EPOCH 27 done: loss 0.1478 - lr 0.000000\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:44:57,710 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:45:01,642 TRAIN : loss 0.13194435834884644 - f1-score (weighted avg)  0.8369\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:45:05,032 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:45:05,103 DEV : loss 0.3356753885746002 - f1-score (weighted avg)  0.6564\n",
            "2023-04-21 15:45:05,126 Epoch    27: reducing learning rate of group 0 to 1.5625e-07.\n",
            "2023-04-21 15:45:05,128 BAD EPOCHS (no improvement): 2\n",
            "2023-04-21 15:45:05,133 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:45:11,940 epoch 28 - iter 6/64 - loss 0.16046613 - time (sec): 6.81 - samples/sec: 1417.28 - lr: 0.000000\n",
            "2023-04-21 15:45:18,642 epoch 28 - iter 12/64 - loss 0.15260810 - time (sec): 13.51 - samples/sec: 1415.45 - lr: 0.000000\n",
            "2023-04-21 15:45:25,363 epoch 28 - iter 18/64 - loss 0.14031779 - time (sec): 20.23 - samples/sec: 1363.18 - lr: 0.000000\n",
            "2023-04-21 15:45:32,257 epoch 28 - iter 24/64 - loss 0.16723186 - time (sec): 27.12 - samples/sec: 1381.96 - lr: 0.000000\n",
            "2023-04-21 15:45:38,990 epoch 28 - iter 30/64 - loss 0.16047278 - time (sec): 33.86 - samples/sec: 1366.28 - lr: 0.000000\n",
            "2023-04-21 15:45:45,650 epoch 28 - iter 36/64 - loss 0.15653009 - time (sec): 40.52 - samples/sec: 1337.97 - lr: 0.000000\n",
            "2023-04-21 15:45:52,353 epoch 28 - iter 42/64 - loss 0.15327988 - time (sec): 47.22 - samples/sec: 1328.16 - lr: 0.000000\n",
            "2023-04-21 15:45:59,065 epoch 28 - iter 48/64 - loss 0.15216819 - time (sec): 53.93 - samples/sec: 1322.45 - lr: 0.000000\n",
            "2023-04-21 15:46:05,787 epoch 28 - iter 54/64 - loss 0.14951125 - time (sec): 60.65 - samples/sec: 1331.77 - lr: 0.000000\n",
            "2023-04-21 15:46:12,542 epoch 28 - iter 60/64 - loss 0.14889327 - time (sec): 67.41 - samples/sec: 1339.92 - lr: 0.000000\n",
            "2023-04-21 15:46:16,173 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:46:16,175 EPOCH 28 done: loss 0.1462 - lr 0.000000\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:46:41,088 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:46:45,056 TRAIN : loss 0.13082623481750488 - f1-score (weighted avg)  0.8388\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:46:48,441 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:46:48,513 DEV : loss 0.33689722418785095 - f1-score (weighted avg)  0.6546\n",
            "2023-04-21 15:46:48,533 BAD EPOCHS (no improvement): 1\n",
            "2023-04-21 15:46:48,537 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:46:55,236 epoch 29 - iter 6/64 - loss 0.11365887 - time (sec): 6.70 - samples/sec: 1160.16 - lr: 0.000000\n",
            "2023-04-21 15:47:01,981 epoch 29 - iter 12/64 - loss 0.12197672 - time (sec): 13.44 - samples/sec: 1349.30 - lr: 0.000000\n",
            "2023-04-21 15:47:08,696 epoch 29 - iter 18/64 - loss 0.12068983 - time (sec): 20.16 - samples/sec: 1409.28 - lr: 0.000000\n",
            "2023-04-21 15:47:15,379 epoch 29 - iter 24/64 - loss 0.12583134 - time (sec): 26.84 - samples/sec: 1381.07 - lr: 0.000000\n",
            "2023-04-21 15:47:22,050 epoch 29 - iter 30/64 - loss 0.12798782 - time (sec): 33.51 - samples/sec: 1345.48 - lr: 0.000000\n",
            "2023-04-21 15:47:28,783 epoch 29 - iter 36/64 - loss 0.13028769 - time (sec): 40.24 - samples/sec: 1337.37 - lr: 0.000000\n",
            "2023-04-21 15:47:35,589 epoch 29 - iter 42/64 - loss 0.14016421 - time (sec): 47.05 - samples/sec: 1348.81 - lr: 0.000000\n",
            "2023-04-21 15:47:42,298 epoch 29 - iter 48/64 - loss 0.13534457 - time (sec): 53.76 - samples/sec: 1343.27 - lr: 0.000000\n",
            "2023-04-21 15:47:48,979 epoch 29 - iter 54/64 - loss 0.13288270 - time (sec): 60.44 - samples/sec: 1339.71 - lr: 0.000000\n",
            "2023-04-21 15:47:55,846 epoch 29 - iter 60/64 - loss 0.14731449 - time (sec): 67.31 - samples/sec: 1339.15 - lr: 0.000000\n",
            "2023-04-21 15:47:59,497 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:47:59,499 EPOCH 29 done: loss 0.1461 - lr 0.000000\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:48:24,464 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:48:27,195 TRAIN : loss 0.13075421750545502 - f1-score (weighted avg)  0.8393\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.46it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:48:30,603 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:48:30,675 DEV : loss 0.3382260501384735 - f1-score (weighted avg)  0.654\n",
            "2023-04-21 15:48:30,697 Epoch    29: reducing learning rate of group 0 to 7.8125e-08.\n",
            "2023-04-21 15:48:30,699 BAD EPOCHS (no improvement): 2\n",
            "2023-04-21 15:48:30,702 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:48:37,446 epoch 30 - iter 6/64 - loss 0.12883740 - time (sec): 6.74 - samples/sec: 1453.66 - lr: 0.000000\n",
            "2023-04-21 15:48:44,137 epoch 30 - iter 12/64 - loss 0.12830416 - time (sec): 13.43 - samples/sec: 1419.84 - lr: 0.000000\n",
            "2023-04-21 15:48:50,849 epoch 30 - iter 18/64 - loss 0.13461667 - time (sec): 20.14 - samples/sec: 1398.62 - lr: 0.000000\n",
            "2023-04-21 15:48:57,514 epoch 30 - iter 24/64 - loss 0.13301284 - time (sec): 26.81 - samples/sec: 1347.04 - lr: 0.000000\n",
            "2023-04-21 15:49:04,231 epoch 30 - iter 30/64 - loss 0.13227422 - time (sec): 33.53 - samples/sec: 1340.15 - lr: 0.000000\n",
            "2023-04-21 15:49:12,088 epoch 30 - iter 36/64 - loss 0.15462808 - time (sec): 41.38 - samples/sec: 1305.84 - lr: 0.000000\n",
            "2023-04-21 15:49:18,800 epoch 30 - iter 42/64 - loss 0.15011043 - time (sec): 48.10 - samples/sec: 1303.11 - lr: 0.000000\n",
            "2023-04-21 15:49:25,634 epoch 30 - iter 48/64 - loss 0.15119787 - time (sec): 54.93 - samples/sec: 1318.69 - lr: 0.000000\n",
            "2023-04-21 15:49:32,338 epoch 30 - iter 54/64 - loss 0.14912407 - time (sec): 61.63 - samples/sec: 1306.54 - lr: 0.000000\n",
            "2023-04-21 15:49:39,081 epoch 30 - iter 60/64 - loss 0.14677669 - time (sec): 68.38 - samples/sec: 1314.05 - lr: 0.000000\n",
            "2023-04-21 15:49:42,703 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:49:42,704 EPOCH 30 done: loss 0.1454 - lr 0.000000\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:50:07,687 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:50:11,602 TRAIN : loss 0.13063673675060272 - f1-score (weighted avg)  0.839\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.46it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:50:15,018 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:50:15,087 DEV : loss 0.33797186613082886 - f1-score (weighted avg)  0.6548\n",
            "2023-04-21 15:50:15,108 BAD EPOCHS (no improvement): 1\n",
            "2023-04-21 15:50:15,113 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:50:21,798 epoch 31 - iter 6/64 - loss 0.09897079 - time (sec): 6.68 - samples/sec: 1363.76 - lr: 0.000000\n",
            "2023-04-21 15:50:28,512 epoch 31 - iter 12/64 - loss 0.11426901 - time (sec): 13.40 - samples/sec: 1277.09 - lr: 0.000000\n",
            "2023-04-21 15:50:35,297 epoch 31 - iter 18/64 - loss 0.13729169 - time (sec): 20.18 - samples/sec: 1304.91 - lr: 0.000000\n",
            "2023-04-21 15:50:41,999 epoch 31 - iter 24/64 - loss 0.13821290 - time (sec): 26.88 - samples/sec: 1306.11 - lr: 0.000000\n",
            "2023-04-21 15:50:48,707 epoch 31 - iter 30/64 - loss 0.13753054 - time (sec): 33.59 - samples/sec: 1294.34 - lr: 0.000000\n",
            "2023-04-21 15:50:55,390 epoch 31 - iter 36/64 - loss 0.14088950 - time (sec): 40.28 - samples/sec: 1289.02 - lr: 0.000000\n",
            "2023-04-21 15:51:02,283 epoch 31 - iter 42/64 - loss 0.16273451 - time (sec): 47.17 - samples/sec: 1305.29 - lr: 0.000000\n",
            "2023-04-21 15:51:09,042 epoch 31 - iter 48/64 - loss 0.15667424 - time (sec): 53.93 - samples/sec: 1320.53 - lr: 0.000000\n",
            "2023-04-21 15:51:15,752 epoch 31 - iter 54/64 - loss 0.15135294 - time (sec): 60.64 - samples/sec: 1349.55 - lr: 0.000000\n",
            "2023-04-21 15:51:22,434 epoch 31 - iter 60/64 - loss 0.14755317 - time (sec): 67.32 - samples/sec: 1345.78 - lr: 0.000000\n",
            "2023-04-21 15:51:26,046 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:51:26,048 EPOCH 31 done: loss 0.1465 - lr 0.000000\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:51:50,980 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:51:54,929 TRAIN : loss 0.1305588036775589 - f1-score (weighted avg)  0.8395\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:51:58,322 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:51:58,391 DEV : loss 0.3375723361968994 - f1-score (weighted avg)  0.657\n",
            "2023-04-21 15:51:58,412 Epoch    31: reducing learning rate of group 0 to 3.9063e-08.\n",
            "2023-04-21 15:51:58,414 BAD EPOCHS (no improvement): 2\n",
            "2023-04-21 15:51:58,417 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:52:05,115 epoch 32 - iter 6/64 - loss 0.11941881 - time (sec): 6.70 - samples/sec: 1425.99 - lr: 0.000000\n",
            "2023-04-21 15:52:11,835 epoch 32 - iter 12/64 - loss 0.12052562 - time (sec): 13.42 - samples/sec: 1418.66 - lr: 0.000000\n",
            "2023-04-21 15:52:18,531 epoch 32 - iter 18/64 - loss 0.12396488 - time (sec): 20.11 - samples/sec: 1359.37 - lr: 0.000000\n",
            "2023-04-21 15:52:25,223 epoch 32 - iter 24/64 - loss 0.12340630 - time (sec): 26.80 - samples/sec: 1363.27 - lr: 0.000000\n",
            "2023-04-21 15:52:32,104 epoch 32 - iter 30/64 - loss 0.14970966 - time (sec): 33.68 - samples/sec: 1357.34 - lr: 0.000000\n",
            "2023-04-21 15:52:38,788 epoch 32 - iter 36/64 - loss 0.14444917 - time (sec): 40.37 - samples/sec: 1359.72 - lr: 0.000000\n",
            "2023-04-21 15:52:45,595 epoch 32 - iter 42/64 - loss 0.14551454 - time (sec): 47.18 - samples/sec: 1360.38 - lr: 0.000000\n",
            "2023-04-21 15:52:52,467 epoch 32 - iter 48/64 - loss 0.14838597 - time (sec): 54.05 - samples/sec: 1364.88 - lr: 0.000000\n",
            "2023-04-21 15:52:59,148 epoch 32 - iter 54/64 - loss 0.14776232 - time (sec): 60.73 - samples/sec: 1351.60 - lr: 0.000000\n",
            "2023-04-21 15:53:05,823 epoch 32 - iter 60/64 - loss 0.14764939 - time (sec): 67.40 - samples/sec: 1341.63 - lr: 0.000000\n",
            "2023-04-21 15:53:09,456 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:53:09,458 EPOCH 32 done: loss 0.1456 - lr 0.000000\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:53:34,386 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:53:37,110 TRAIN : loss 0.13050147891044617 - f1-score (weighted avg)  0.8396\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.49it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:53:40,486 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:53:40,556 DEV : loss 0.3376908600330353 - f1-score (weighted avg)  0.657\n",
            "2023-04-21 15:53:40,577 BAD EPOCHS (no improvement): 1\n",
            "2023-04-21 15:53:40,582 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:53:47,278 epoch 33 - iter 6/64 - loss 0.10902645 - time (sec): 6.69 - samples/sec: 1409.26 - lr: 0.000000\n",
            "2023-04-21 15:53:53,997 epoch 33 - iter 12/64 - loss 0.12133482 - time (sec): 13.41 - samples/sec: 1427.33 - lr: 0.000000\n",
            "2023-04-21 15:54:00,698 epoch 33 - iter 18/64 - loss 0.12594820 - time (sec): 20.11 - samples/sec: 1414.91 - lr: 0.000000\n",
            "2023-04-21 15:54:07,408 epoch 33 - iter 24/64 - loss 0.12855389 - time (sec): 26.82 - samples/sec: 1374.70 - lr: 0.000000\n",
            "2023-04-21 15:54:14,158 epoch 33 - iter 30/64 - loss 0.12826825 - time (sec): 33.57 - samples/sec: 1375.71 - lr: 0.000000\n",
            "2023-04-21 15:54:21,868 epoch 33 - iter 36/64 - loss 0.13206107 - time (sec): 41.28 - samples/sec: 1346.49 - lr: 0.000000\n",
            "2023-04-21 15:54:28,577 epoch 33 - iter 42/64 - loss 0.13273690 - time (sec): 47.99 - samples/sec: 1328.81 - lr: 0.000000\n",
            "2023-04-21 15:54:35,267 epoch 33 - iter 48/64 - loss 0.13267821 - time (sec): 54.68 - samples/sec: 1321.00 - lr: 0.000000\n",
            "2023-04-21 15:54:42,037 epoch 33 - iter 54/64 - loss 0.13675329 - time (sec): 61.45 - samples/sec: 1308.47 - lr: 0.000000\n",
            "2023-04-21 15:54:48,932 epoch 33 - iter 60/64 - loss 0.14678085 - time (sec): 68.35 - samples/sec: 1313.51 - lr: 0.000000\n",
            "2023-04-21 15:54:52,585 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:54:52,587 EPOCH 33 done: loss 0.1451 - lr 0.000000\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.58it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:55:17,549 Evaluating as a multi-label problem: True\n",
            "2023-04-21 15:55:21,508 TRAIN : loss 0.13037079572677612 - f1-score (weighted avg)  0.8397\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.48it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:55:24,907 Evaluating as a multi-label problem: False\n",
            "2023-04-21 15:55:24,978 DEV : loss 0.3376230001449585 - f1-score (weighted avg)  0.6567\n",
            "2023-04-21 15:55:24,999 Epoch    33: reducing learning rate of group 0 to 1.9531e-08.\n",
            "2023-04-21 15:55:25,001 BAD EPOCHS (no improvement): 2\n",
            "2023-04-21 15:55:25,017 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:55:25,019 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:55:25,020 learning rate too small - quitting training!\n",
            "2023-04-21 15:55:25,021 ----------------------------------------------------------------------------------------------------\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:55:37,474 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:55:48,075 SequenceTagger predicts: Dictionary with 22 tags: <unk>, O, I-ADR, E-ADR, B-ADR, S-ADR, S-Drug, B-Finding, E-Finding, I-Finding, B-Disease, E-Disease, E-Symptom, S-Disease, S-Symptom, E-Drug, B-Symptom, B-Drug, S-Finding, I-Symptom, I-Drug, I-Disease\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.53it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:55:51,643 Evaluating as a multi-label problem: True\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:55:51,987 0.6996\t0.719\t0.7091\t0.9279\n",
            "2023-04-21 15:55:51,988 \n",
            "Results:\n",
            "- F-score (micro) 0.7091\n",
            "- F-score (macro) 0.1898\n",
            "- Accuracy 0.9279\n",
            "\n",
            "By class:\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "       I-ADR     0.6293    0.7600    0.6885       525\n",
            "       E-ADR     0.7752    0.8190    0.7965       442\n",
            "       B-ADR     0.7602    0.8493    0.8023       418\n",
            "      S-Drug     0.6800    0.9740    0.8009       192\n",
            "       S-ADR     0.6318    0.8038    0.7075       158\n",
            "   S-Finding     0.0000    0.0000    0.0000        28\n",
            "   E-Finding     0.0000    0.0000    0.0000        27\n",
            "      B-Drug     0.0000    0.0000    0.0000        27\n",
            "      E-Drug     0.0000    0.0000    0.0000        27\n",
            "      I-Drug     0.0000    0.0000    0.0000        25\n",
            "   B-Finding     0.0000    0.0000    0.0000        24\n",
            "   I-Symptom     0.0000    0.0000    0.0000        20\n",
            "   B-Symptom     0.0000    0.0000    0.0000        13\n",
            "   E-Symptom     0.0000    0.0000    0.0000        13\n",
            "   S-Disease     0.0000    0.0000    0.0000        12\n",
            "   E-Disease     0.0000    0.0000    0.0000         9\n",
            "   I-Finding     0.0000    0.0000    0.0000         9\n",
            "   S-Symptom     0.0000    0.0000    0.0000         9\n",
            "   B-Disease     0.0000    0.0000    0.0000         8\n",
            "   I-Disease     0.0000    0.0000    0.0000         3\n",
            "\n",
            "   micro avg     0.6996    0.7190    0.7091      1989\n",
            "   macro avg     0.1738    0.2103    0.1898      1989\n",
            "weighted avg     0.6140    0.7190    0.6608      1989\n",
            " samples avg     0.1208    0.1208    0.1208      1989\n",
            "\n",
            "2023-04-21 15:55:51,989 ----------------------------------------------------------------------------------------------------\n"
          ]
        }
      ],
      "source": [
        "hist = trainer.train('flair_ner_model'+suf,\n",
        "              embeddings_storage_mode='gpu',\n",
        "              gold_label_dictionary_for_eval=tag_dictionary,\n",
        "              mini_batch_size = 15,\n",
        "              eval_batch_size = 15,\n",
        "              patience=1,\n",
        "              anneal_factor = 0.5,\n",
        "              anneal_against_dev_loss=True,\n",
        "              main_evaluation_metric = (\"weighted avg\", \"f1-score\"),\n",
        "              metrics_for_tensorboard=[(\"weighted avg\", 'precision'), (\"weighted avg\", 'recall'), (\"macro avg\", 'f1-score'), (\"macro avg\", 'precision'), (\"macro avg\", 'recall')],\n",
        "              use_tensorboard=True,\n",
        "              tensorboard_comment='Flair NER',\n",
        "              tensorboard_log_dir='runs/tb_logs'+suf,\n",
        "              min_learning_rate=3e-8,\n",
        "              optimizer=AdamW,\n",
        "              learning_rate=5.0e-6,\n",
        "              weight_decay=0.001,\n",
        "              max_epochs=100,\n",
        "              monitor_train=True,\n",
        "              save_optimizer_state=True,)\n",
        "# hist"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "id": "ExiLRreqXsDJ"
      },
      "outputs": [],
      "source": [
        "# print(hist.keys())\n",
        "# hist"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "P5Uu5qkN0pia",
        "outputId": "2dc910a6-ad18-4515-8f1b-4442fca4ccee"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "_v2_spanbert-large-cased\n",
            "2023-04-21 15:55:52,012 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:55:52,013 WARNING: No LOSS found for test split in this data.\n",
            "2023-04-21 15:55:52,014 Are you sure you want to plot LOSS and not another value?\n",
            "2023-04-21 15:55:52,015 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:55:52,039 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:55:52,040 WARNING: No F1 found for test split in this data.\n",
            "2023-04-21 15:55:52,040 Are you sure you want to plot F1 and not another value?\n",
            "2023-04-21 15:55:52,041 ----------------------------------------------------------------------------------------------------\n",
            "2023-04-21 15:55:53,128 Loss and F1 plots are saved in flair_ner_model_v2_spanbert-large-cased/training.png\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1500x1000 with 2 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 64/64 [00:24<00:00,  2.56it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:56:18,571 Evaluating as a multi-label problem: True\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 8/8 [00:03<00:00,  2.50it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:56:25,680 Evaluating as a multi-label problem: False\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n",
            "100%|██████████| 8/8 [00:03<00:00,  2.54it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-04-21 15:56:28,925 Evaluating as a multi-label problem: True\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "data": {
            "text/html": [
              "\n",
              "  <div id=\"df-3db668d1-19af-4ccc-912a-bf5e6b16f194\">\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>split</th>\n",
              "      <th>index</th>\n",
              "      <th>I-ADR</th>\n",
              "      <th>E-ADR</th>\n",
              "      <th>B-ADR</th>\n",
              "      <th>S-ADR</th>\n",
              "      <th>S-Drug</th>\n",
              "      <th>B-Finding</th>\n",
              "      <th>E-Finding</th>\n",
              "      <th>I-Finding</th>\n",
              "      <th>...</th>\n",
              "      <th>B-Symptom</th>\n",
              "      <th>B-Drug</th>\n",
              "      <th>S-Finding</th>\n",
              "      <th>I-Symptom</th>\n",
              "      <th>I-Drug</th>\n",
              "      <th>I-Disease</th>\n",
              "      <th>micro avg</th>\n",
              "      <th>macro avg</th>\n",
              "      <th>weighted avg</th>\n",
              "      <th>samples avg</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>train</td>\n",
              "      <td>precision</td>\n",
              "      <td>0.873157</td>\n",
              "      <td>0.871623</td>\n",
              "      <td>0.856192</td>\n",
              "      <td>0.736933</td>\n",
              "      <td>0.727064</td>\n",
              "      <td>0.391304</td>\n",
              "      <td>0.384615</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.836392</td>\n",
              "      <td>0.242044</td>\n",
              "      <td>0.757067</td>\n",
              "      <td>0.142355</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>train</td>\n",
              "      <td>recall</td>\n",
              "      <td>0.971679</td>\n",
              "      <td>0.966343</td>\n",
              "      <td>0.959988</td>\n",
              "      <td>0.941306</td>\n",
              "      <td>0.986003</td>\n",
              "      <td>0.045000</td>\n",
              "      <td>0.025000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.858097</td>\n",
              "      <td>0.244766</td>\n",
              "      <td>0.858097</td>\n",
              "      <td>0.142318</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>train</td>\n",
              "      <td>f1-score</td>\n",
              "      <td>0.919787</td>\n",
              "      <td>0.916543</td>\n",
              "      <td>0.905124</td>\n",
              "      <td>0.826675</td>\n",
              "      <td>0.836964</td>\n",
              "      <td>0.080717</td>\n",
              "      <td>0.046948</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.847105</td>\n",
              "      <td>0.226638</td>\n",
              "      <td>0.799225</td>\n",
              "      <td>0.142331</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>train</td>\n",
              "      <td>support</td>\n",
              "      <td>4449.000000</td>\n",
              "      <td>3506.000000</td>\n",
              "      <td>3349.000000</td>\n",
              "      <td>1363.000000</td>\n",
              "      <td>1286.000000</td>\n",
              "      <td>200.000000</td>\n",
              "      <td>200.000000</td>\n",
              "      <td>149.0</td>\n",
              "      <td>...</td>\n",
              "      <td>110.0</td>\n",
              "      <td>110.0</td>\n",
              "      <td>109.0</td>\n",
              "      <td>104.0</td>\n",
              "      <td>50.0</td>\n",
              "      <td>50.0</td>\n",
              "      <td>15722.000000</td>\n",
              "      <td>15722.000000</td>\n",
              "      <td>15722.000000</td>\n",
              "      <td>15722.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>dev</td>\n",
              "      <td>precision</td>\n",
              "      <td>0.580189</td>\n",
              "      <td>0.730693</td>\n",
              "      <td>0.739044</td>\n",
              "      <td>0.678571</td>\n",
              "      <td>0.654472</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.671795</td>\n",
              "      <td>0.169148</td>\n",
              "      <td>0.582696</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5</th>\n",
              "      <td>dev</td>\n",
              "      <td>recall</td>\n",
              "      <td>0.713733</td>\n",
              "      <td>0.821826</td>\n",
              "      <td>0.839367</td>\n",
              "      <td>0.818182</td>\n",
              "      <td>0.981707</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.697483</td>\n",
              "      <td>0.208741</td>\n",
              "      <td>0.697483</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>6</th>\n",
              "      <td>dev</td>\n",
              "      <td>f1-score</td>\n",
              "      <td>0.640069</td>\n",
              "      <td>0.773585</td>\n",
              "      <td>0.786017</td>\n",
              "      <td>0.741866</td>\n",
              "      <td>0.785366</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.684398</td>\n",
              "      <td>0.186345</td>\n",
              "      <td>0.633846</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7</th>\n",
              "      <td>dev</td>\n",
              "      <td>support</td>\n",
              "      <td>517.000000</td>\n",
              "      <td>449.000000</td>\n",
              "      <td>442.000000</td>\n",
              "      <td>209.000000</td>\n",
              "      <td>164.000000</td>\n",
              "      <td>39.000000</td>\n",
              "      <td>34.000000</td>\n",
              "      <td>20.0</td>\n",
              "      <td>...</td>\n",
              "      <td>12.0</td>\n",
              "      <td>20.0</td>\n",
              "      <td>25.0</td>\n",
              "      <td>23.0</td>\n",
              "      <td>7.0</td>\n",
              "      <td>13.0</td>\n",
              "      <td>2066.000000</td>\n",
              "      <td>2066.000000</td>\n",
              "      <td>2066.000000</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>8</th>\n",
              "      <td>test</td>\n",
              "      <td>precision</td>\n",
              "      <td>0.629338</td>\n",
              "      <td>0.775161</td>\n",
              "      <td>0.760171</td>\n",
              "      <td>0.631841</td>\n",
              "      <td>0.680000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.699609</td>\n",
              "      <td>0.173826</td>\n",
              "      <td>0.613960</td>\n",
              "      <td>0.120838</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9</th>\n",
              "      <td>test</td>\n",
              "      <td>recall</td>\n",
              "      <td>0.760000</td>\n",
              "      <td>0.819005</td>\n",
              "      <td>0.849282</td>\n",
              "      <td>0.803797</td>\n",
              "      <td>0.973958</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.718954</td>\n",
              "      <td>0.210302</td>\n",
              "      <td>0.718954</td>\n",
              "      <td>0.120754</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10</th>\n",
              "      <td>test</td>\n",
              "      <td>f1-score</td>\n",
              "      <td>0.688525</td>\n",
              "      <td>0.796480</td>\n",
              "      <td>0.802260</td>\n",
              "      <td>0.707521</td>\n",
              "      <td>0.800857</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.709150</td>\n",
              "      <td>0.189782</td>\n",
              "      <td>0.660843</td>\n",
              "      <td>0.120782</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>11</th>\n",
              "      <td>test</td>\n",
              "      <td>support</td>\n",
              "      <td>525.000000</td>\n",
              "      <td>442.000000</td>\n",
              "      <td>418.000000</td>\n",
              "      <td>158.000000</td>\n",
              "      <td>192.000000</td>\n",
              "      <td>24.000000</td>\n",
              "      <td>27.000000</td>\n",
              "      <td>9.0</td>\n",
              "      <td>...</td>\n",
              "      <td>13.0</td>\n",
              "      <td>27.0</td>\n",
              "      <td>28.0</td>\n",
              "      <td>20.0</td>\n",
              "      <td>25.0</td>\n",
              "      <td>3.0</td>\n",
              "      <td>1989.000000</td>\n",
              "      <td>1989.000000</td>\n",
              "      <td>1989.000000</td>\n",
              "      <td>1989.000000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>12 rows × 26 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-3db668d1-19af-4ccc-912a-bf5e6b16f194')\"\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-3db668d1-19af-4ccc-912a-bf5e6b16f194 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-3db668d1-19af-4ccc-912a-bf5e6b16f194');\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",
              "  "
            ],
            "text/plain": [
              "    split      index        I-ADR        E-ADR        B-ADR        S-ADR  \\\n",
              "0   train  precision     0.873157     0.871623     0.856192     0.736933   \n",
              "1   train     recall     0.971679     0.966343     0.959988     0.941306   \n",
              "2   train   f1-score     0.919787     0.916543     0.905124     0.826675   \n",
              "3   train    support  4449.000000  3506.000000  3349.000000  1363.000000   \n",
              "4     dev  precision     0.580189     0.730693     0.739044     0.678571   \n",
              "5     dev     recall     0.713733     0.821826     0.839367     0.818182   \n",
              "6     dev   f1-score     0.640069     0.773585     0.786017     0.741866   \n",
              "7     dev    support   517.000000   449.000000   442.000000   209.000000   \n",
              "8    test  precision     0.629338     0.775161     0.760171     0.631841   \n",
              "9    test     recall     0.760000     0.819005     0.849282     0.803797   \n",
              "10   test   f1-score     0.688525     0.796480     0.802260     0.707521   \n",
              "11   test    support   525.000000   442.000000   418.000000   158.000000   \n",
              "\n",
              "         S-Drug   B-Finding   E-Finding  I-Finding  ...  B-Symptom  B-Drug  \\\n",
              "0      0.727064    0.391304    0.384615        0.0  ...        0.0     0.0   \n",
              "1      0.986003    0.045000    0.025000        0.0  ...        0.0     0.0   \n",
              "2      0.836964    0.080717    0.046948        0.0  ...        0.0     0.0   \n",
              "3   1286.000000  200.000000  200.000000      149.0  ...      110.0   110.0   \n",
              "4      0.654472    0.000000    0.000000        0.0  ...        0.0     0.0   \n",
              "5      0.981707    0.000000    0.000000        0.0  ...        0.0     0.0   \n",
              "6      0.785366    0.000000    0.000000        0.0  ...        0.0     0.0   \n",
              "7    164.000000   39.000000   34.000000       20.0  ...       12.0    20.0   \n",
              "8      0.680000    0.000000    0.000000        0.0  ...        0.0     0.0   \n",
              "9      0.973958    0.000000    0.000000        0.0  ...        0.0     0.0   \n",
              "10     0.800857    0.000000    0.000000        0.0  ...        0.0     0.0   \n",
              "11   192.000000   24.000000   27.000000        9.0  ...       13.0    27.0   \n",
              "\n",
              "    S-Finding  I-Symptom  I-Drug  I-Disease     micro avg     macro avg  \\\n",
              "0         0.0        0.0     0.0        0.0      0.836392      0.242044   \n",
              "1         0.0        0.0     0.0        0.0      0.858097      0.244766   \n",
              "2         0.0        0.0     0.0        0.0      0.847105      0.226638   \n",
              "3       109.0      104.0    50.0       50.0  15722.000000  15722.000000   \n",
              "4         0.0        0.0     0.0        0.0      0.671795      0.169148   \n",
              "5         0.0        0.0     0.0        0.0      0.697483      0.208741   \n",
              "6         0.0        0.0     0.0        0.0      0.684398      0.186345   \n",
              "7        25.0       23.0     7.0       13.0   2066.000000   2066.000000   \n",
              "8         0.0        0.0     0.0        0.0      0.699609      0.173826   \n",
              "9         0.0        0.0     0.0        0.0      0.718954      0.210302   \n",
              "10        0.0        0.0     0.0        0.0      0.709150      0.189782   \n",
              "11       28.0       20.0    25.0        3.0   1989.000000   1989.000000   \n",
              "\n",
              "    weighted avg   samples avg  \n",
              "0       0.757067      0.142355  \n",
              "1       0.858097      0.142318  \n",
              "2       0.799225      0.142331  \n",
              "3   15722.000000  15722.000000  \n",
              "4       0.582696           NaN  \n",
              "5       0.697483           NaN  \n",
              "6       0.633846           NaN  \n",
              "7    2066.000000           NaN  \n",
              "8       0.613960      0.120838  \n",
              "9       0.718954      0.120754  \n",
              "10      0.660843      0.120782  \n",
              "11   1989.000000   1989.000000  \n",
              "\n",
              "[12 rows x 26 columns]"
            ]
          },
          "execution_count": 17,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "print(suf)\n",
        "\n",
        "os.mkdir('flair_ner_model'+suf+'/classification_report')\n",
        "plotter.plot_training_curves('flair_ner_model'+suf+'/loss.tsv')\n",
        "\n",
        "# evaluate the tagger on the test set\n",
        "result = tagger.evaluate(corpus.train, \n",
        "                         embeddings_storage_mode='gpu',\n",
        "                         mini_batch_size=15,\n",
        "                         main_evaluation_metric = (\"weighted avg\", \"f1-score\"),\n",
        "                         gold_label_type='ner',\n",
        "                         gold_label_dictionary=tag_dictionary)\n",
        "open('flair_ner_model'+suf+'/classification_report/train.txt', 'w').write(str(result))\n",
        "res1 = pd.DataFrame.from_dict(result.classification_report).reset_index()\n",
        "res1.insert(loc=0, column='split', value='train')\n",
        "\n",
        "result = tagger.evaluate(corpus.dev, \n",
        "                         embeddings_storage_mode='gpu',\n",
        "                         mini_batch_size=15,\n",
        "                         main_evaluation_metric = (\"weighted avg\", \"f1-score\"),\n",
        "                         gold_label_type='ner',\n",
        "                         gold_label_dictionary=tag_dictionary)\n",
        "open('flair_ner_model'+suf+'/classification_report/dev.txt', 'w').write(str(result))\n",
        "res2 = pd.DataFrame.from_dict(result.classification_report).reset_index()\n",
        "res2.insert(loc=0, column='split', value='dev')\n",
        "\n",
        "result = tagger.evaluate(corpus.test, \n",
        "                         embeddings_storage_mode='gpu',\n",
        "                         mini_batch_size=15,\n",
        "                         main_evaluation_metric = (\"weighted avg\", \"f1-score\"),\n",
        "                         gold_label_type='ner',\n",
        "                         gold_label_dictionary=tag_dictionary)\n",
        "open('flair_ner_model'+suf+'/classification_report/test.txt', 'w').write(str(result))\n",
        "res3 = pd.DataFrame.from_dict(result.classification_report).reset_index()\n",
        "res3.insert(loc=0, column='split', value='test')\n",
        "\n",
        "res = pd.concat([res1,res2,res3],ignore_index=True)\n",
        "res.to_csv('flair_ner_model'+suf+'/classification_report.csv',index=False)\n",
        "res"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 18,
      "metadata": {
        "id": "VvIDCtYJkY-G"
      },
      "outputs": [],
      "source": [
        "# import time\n",
        "# start_time = time.time()\n",
        "# while time.time() - start_time < 1:\n",
        "#     time.sleep(1)\n",
        "# print('Session has timed out (2 minutes). Disconnecting...')\n",
        "# !kill -9 -1"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tGILL4RFtGil"
      },
      "source": [
        "Layering can be a good idea if the NER tags in the first layer have their own set of dependent values in the second layer. This can help capture more information about the entities being recognized and provide more context for downstream tasks. However, it also increases the complexity of the annotation and the training process, so it should be used judiciously. In general, it's a good idea to start with a simple tag scheme and gradually increase complexity as needed."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JoZ5D6aEtGin"
      },
      "source": [
        "    does it really matter if u use the E header\n",
        "\n",
        "It depends on the specific task and the requirements of the downstream model. In some cases, using the E tag can help disambiguate between overlapping entities, while in other cases it may not be necessary or even desired. Ultimately, it's up to the developer to determine the appropriate tagging scheme for their specific use case."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BGt2RhM6tGin"
      },
      "source": [
        "    would adding POS as an extra NER layer help model performance or not\n",
        "\n",
        "Adding POS as an extra NER layer may or may not help model performance, depending on the specific use case and data being analyzed.\n",
        "\n",
        "On one hand, incorporating POS tags could provide additional context to the model and help it better distinguish between named entities and other parts of speech. For example, knowing that a word is a noun or verb could help the model determine if it is likely to be a named entity or not.\n",
        "\n",
        "On the other hand, adding an additional layer could also increase model complexity and training time without necessarily improving performance. It may also require additional manual annotation efforts to create a labeled dataset with both NER and POS tags.\n",
        "\n",
        "Ultimately, whether to include POS as an additional layer will depend on the specific requirements and goals of the project, as well as the available data and resources."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6zpEn1bgtGin"
      },
      "source": [
        "    besides bioes ordering (which is up-to-down), it doesn't matter what order the tags in the same domain appear in left-to-right?\n",
        "\n",
        "Yes, that is correct. As long as the BIOES order is maintained vertically, the order of tags in the same domain from left-to-right should not impact training in flai"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yrmL98VatGin"
      },
      "source": [
        "    different typenames must be treated as separate tasks?\n",
        "\n",
        "Yes, that's correct. Different typenames should be treated as separate tasks because they represent different types of annotations that you want to predict. For example, in a biomedical NER task, you might want to predict the entity types of both drugs and diseases, which would be represented by two different typenames in your annotation data. Predicting both types of entities would require different models and possibly different evaluation metrics."
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "gpuType": "A100",
      "provenance": []
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "flair",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.9.16"
    },
    "vscode": {
      "interpreter": {
        "hash": "381c8abc3392c5f20473858f182d9ac4a53ef053fb1f276fcb4936938616c3f3"
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}