CatalanULMFit.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "colab": {
      "name": "CatalanULMFit.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nWfh0QSIA24V",
        "colab_type": "text"
      },
      "source": [
        "## Instal·lant dependències"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "id": "rGBxlxrGA24V",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import os\n",
        "from fastai import *\n",
        "from fastai.text import *\n",
        "from gensim.corpora import WikiCorpus\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "\n",
        "from pathlib import Path\n",
        "from functools import partial"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7nv_BhdZxivY",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "outputId": "7b18abba-8e4a-4229-e6ef-eca7ad9231b6"
      },
      "source": [
        "!pip install sacremoses"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting sacremoses\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/7d/34/09d19aff26edcc8eb2a01bed8e98f13a1537005d31e95233fd48216eed10/sacremoses-0.0.43.tar.gz (883kB)\n",
            "\u001b[K     |████████████████████████████████| 890kB 6.6MB/s \n",
            "\u001b[?25hRequirement already satisfied: regex in /usr/local/lib/python3.6/dist-packages (from sacremoses) (2019.12.20)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from sacremoses) (1.15.0)\n",
            "Requirement already satisfied: click in /usr/local/lib/python3.6/dist-packages (from sacremoses) (7.1.2)\n",
            "Requirement already satisfied: joblib in /usr/local/lib/python3.6/dist-packages (from sacremoses) (0.16.0)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.6/dist-packages (from sacremoses) (4.41.1)\n",
            "Building wheels for collected packages: sacremoses\n",
            "  Building wheel for sacremoses (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for sacremoses: filename=sacremoses-0.0.43-cp36-none-any.whl size=893260 sha256=8d2f01e67e811af88ac648be384913e7ed04d25a9f8240769115f64da5ffc6f3\n",
            "  Stored in directory: /root/.cache/pip/wheels/29/3c/fd/7ce5c3f0666dab31a50123635e6fb5e19ceb42ce38d4e58f45\n",
            "Successfully built sacremoses\n",
            "Installing collected packages: sacremoses\n",
            "Successfully installed sacremoses-0.0.43\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-1IshZaaA24X",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 372
        },
        "outputId": "f6a7d0bc-5674-444b-c1cf-9aa936ca3fa5"
      },
      "source": [
        "!nvidia-smi"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Fri Aug 21 11:44:41 2020       \n",
            "+-----------------------------------------------------------------------------+\n",
            "| NVIDIA-SMI 450.57       Driver Version: 418.67       CUDA Version: 10.1     |\n",
            "|-------------------------------+----------------------+----------------------+\n",
            "| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |\n",
            "| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |\n",
            "|                               |                      |               MIG M. |\n",
            "|===============================+======================+======================|\n",
            "|   0  Tesla K80           Off  | 00000000:00:04.0 Off |                    0 |\n",
            "| N/A   72C    P8    35W / 149W |     11MiB / 11441MiB |      0%      Default |\n",
            "|                               |                      |                 ERR! |\n",
            "+-------------------------------+----------------------+----------------------+\n",
            "                                                                               \n",
            "+-----------------------------------------------------------------------------+\n",
            "| Processes:                                                                  |\n",
            "|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |\n",
            "|        ID   ID                                                   Usage      |\n",
            "|=============================================================================|\n",
            "|  No running processes found                                                 |\n",
            "+-----------------------------------------------------------------------------+\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NP84DdEwd39J",
        "colab_type": "text"
      },
      "source": [
        "## Descarreguem el Wikipedia dump en català"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l3Ryq09yyB9X",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!rm -rf data/wiki_extr\n",
        "!rm -rf wikiextractor/ WikiExtractor.py"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UEMQy-CdNq5L",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 994
        },
        "outputId": "b172b012-8bab-48bc-cfec-bf3482f9b0e5"
      },
      "source": [
        "!sh get_wikimedia.sh"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saving data in data\n",
            "Chosen language: ca\n",
            "Starting download...\n",
            "--2020-08-21 11:45:17--  https://dumps.wikimedia.org/cawiki/latest/cawiki-latest-pages-articles.xml.bz2\n",
            "Resolving dumps.wikimedia.org (dumps.wikimedia.org)... 208.80.154.7, 2620:0:861:1:208:80:154:7\n",
            "Connecting to dumps.wikimedia.org (dumps.wikimedia.org)|208.80.154.7|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 899327837 (858M) [application/octet-stream]\n",
            "Saving to: 'data/wiki_dumps/cawiki-latest-pages-articles.xml.bz2'\n",
            "\n",
            "cawiki-latest-pages 100%[===================>] 857.67M  4.86MB/s    in 2m 57s  \n",
            "\n",
            "2020-08-21 11:48:14 (4.86 MB/s) - 'data/wiki_dumps/cawiki-latest-pages-articles.xml.bz2' saved [899327837/899327837]\n",
            "\n",
            "Cloning into 'wikiextractor'...\n",
            "remote: Enumerating objects: 4, done.\u001b[K\n",
            "remote: Counting objects: 100% (4/4), done.\u001b[K\n",
            "remote: Compressing objects: 100% (4/4), done.\u001b[K\n",
            "remote: Total 661 (delta 0), reused 0 (delta 0), pack-reused 657\u001b[K\n",
            "Receiving objects: 100% (661/661), 1.27 MiB | 2.92 MiB/s, done.\n",
            "Resolving deltas: 100% (374/374), done.\n",
            "Note: checking out '9cf2a2a883fc8e2146ff8df234d036e695df1be4'.\n",
            "\n",
            "You are in 'detached HEAD' state. You can look around, make experimental\n",
            "changes and commit them, and you can discard any commits you make in this\n",
            "state without impacting any branches by performing another checkout.\n",
            "\n",
            "If you want to create a new branch to retain commits you create, you may\n",
            "do so (now or later) by using -b with the checkout command again. Example:\n",
            "\n",
            "  git checkout -b <new-branch-name>\n",
            "\n",
            "HEAD is now at 9cf2a2a add feature filtering by category of wiki\n",
            "Extracting data/wiki_dumps/cawiki-latest-pages-articles.xml.bz2 to data/wiki_extr/ca...\n",
            "WARNING: Template errors in article 'P (complexitat)' (124631): title(2) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Joaquim Manuel de Moner i de Siscar' (653508): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Varegs' (669888): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Senyal horari' (670666): title(6) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Sensor CCD' (720483): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Far (automoci\\xf3)' (753943): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Norbert Schemansky' (803899): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Tortuga de Horsfield' (807705): title(3) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Prov\\xedncia d'\\xd6v\\xf6rkhangai' (829778): title(5) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'R\\xe8tol de ne\\xf3' (966345): title(2) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Once Upon a Time' (968977): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'B\\xe0squet als Jocs Ol\\xedmpics d'estiu de 2012' (1004177): title(115) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Albert Girard' (1161376): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Escacat (&quot;Taquejat jaqu\\xe8s&quot;)' (1283589): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Segellat de temps' (1298846): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'LSE (llenguatge)' (1358601): title(3) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'F\\xf3rmula de Stirling' (1438250): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'AngularJS' (1500632): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'BQP (complexitat)' (1630407): title(2) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Llei sobre la Bandera i l'Himne Nacional (Jap\\xf3)' (1678260): title(1) recursion(0, 0, 0)\n",
            "WARNING: Template errors in article 'Festival de la Can\\xe7\\xf3 d'Eurovisi\\xf3 2020' (1692867): title(1) recursion(0, 0, 0)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JPQn4ydbQZJl",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "name = 'CatalanULMFit'\n",
        "model_dir = 'models'\n",
        "model_dir = Path(model_dir)\n",
        "model_dir.mkdir(exist_ok=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zIUK4zt2eLCR",
        "colab_type": "text"
      },
      "source": [
        "Creem el corpus"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ooYbrQWaoePg",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "c9e2c908-a655-4183-aa0f-238e28c4940e"
      },
      "source": [
        "!ls data/wiki_extr/ca"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "AA  AB\tAC  AD\tAE  AF\tAG  AH\tAI  AJ\tAK  AL\tAM\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "i4knTPiEpiCU",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 550
        },
        "outputId": "d65043dd-0d4f-43ad-9400-52692037a0ed"
      },
      "source": [
        "!python create_wikiText.py -i data/wiki_extr/ca -o data/wiki -l ca"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Writing to data/wiki/ca-2/ca.wiki.train.tokens...\n",
            "data/wiki/ca-2/ca.wiki.train.tokens. # documents: 6,499. # tokens: 2,000,114.\n",
            "Copying data/wiki/ca-2/ca.wiki.train.tokens to data/wiki/ca-100/ca.wiki.train.tokens.\n",
            "Writing to data/wiki/ca-2/ca.wiki.valid.tokens...\n",
            "data/wiki/ca-2/ca.wiki.valid.tokens. # documents: 788. # tokens: 200,027.\n",
            "Copying data/wiki/ca-2/ca.wiki.valid.tokens to data/wiki/ca-100/ca.wiki.valid.tokens.\n",
            "Writing to data/wiki/ca-2/ca.wiki.test.tokens...\n",
            "data/wiki/ca-2/ca.wiki.test.tokens. # documents: 640. # tokens: 200,178.\n",
            "Copying data/wiki/ca-2/ca.wiki.test.tokens to data/wiki/ca-100/ca.wiki.test.tokens.\n",
            "Writing to data/wiki/ca-100/ca.wiki.train.tokens...\n",
            "Processed 10,000 documents. Total # tokens: 3,198,401.\n",
            "Processed 30,000 documents. Total # tokens: 9,673,988.\n",
            "Processed 40,000 documents. Total # tokens: 13,008,520.\n",
            "Processed 50,000 documents. Total # tokens: 16,326,644.\n",
            "Processed 60,000 documents. Total # tokens: 19,503,057.\n",
            "Processed 90,000 documents. Total # tokens: 28,315,001.\n",
            "Processed 100,000 documents. Total # tokens: 31,066,394.\n",
            "Processed 120,000 documents. Total # tokens: 37,067,559.\n",
            "Processed 130,000 documents. Total # tokens: 41,744,310.\n",
            "Processed 140,000 documents. Total # tokens: 46,043,140.\n",
            "Processed 150,000 documents. Total # tokens: 50,467,852.\n",
            "Processed 160,000 documents. Total # tokens: 55,042,194.\n",
            "Processed 170,000 documents. Total # tokens: 59,093,965.\n",
            "Processed 180,000 documents. Total # tokens: 63,094,839.\n",
            "Processed 200,000 documents. Total # tokens: 70,725,290.\n",
            "Processed 210,000 documents. Total # tokens: 74,439,177.\n",
            "Processed 220,000 documents. Total # tokens: 79,584,731.\n",
            "Processed 230,000 documents. Total # tokens: 85,836,158.\n",
            "Processed 250,000 documents. Total # tokens: 97,355,077.\n",
            "data/wiki/ca-100/ca.wiki.train.tokens. # documents: 252,016. # tokens: 98,000,238.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5SBXoriPxLlL",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 407
        },
        "outputId": "f60b523e-be16-4240-b1fb-42a834f0d193"
      },
      "source": [
        " !python postprocess_wikiText.py -i data/wiki -l ca"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Replacing numbers in data/wiki/ca-2/ca.wiki.train.tokens. Writing to data/wiki/ca-2/ca.wiki.train.tokens.unk.\n",
            "Replacing numbers in data/wiki/ca-2/ca.wiki.valid.tokens. Writing to data/wiki/ca-2/ca.wiki.valid.tokens.unk.\n",
            "Replacing numbers in data/wiki/ca-2/ca.wiki.test.tokens. Writing to data/wiki/ca-2/ca.wiki.test.tokens.unk.\n",
            "Replacing numbers in data/wiki/ca-100/ca.wiki.train.tokens. Writing to data/wiki/ca-100/ca.wiki.train.tokens.unk.\n",
            "Replacing numbers in data/wiki/ca-100/ca.wiki.valid.tokens. Writing to data/wiki/ca-100/ca.wiki.valid.tokens.unk.\n",
            "Replacing numbers in data/wiki/ca-100/ca.wiki.test.tokens. Writing to data/wiki/ca-100/ca.wiki.test.tokens.unk.\n",
            "OOV ratio: 0.0376.\n",
            "ca-2 vocab size: 32420\n",
            "OOV ratio: 0.0086.\n",
            "ca-100 vocab size: 400161\n",
            "Limiting vocab in data/wiki/ca-2/ca.wiki.train.tokens.unk. Writing to data/wiki/ca-2/ca.wiki.train.tokens.unk.\n",
            "ca.wiki.train.tokens.unk. # of tokens: 2054248\n",
            "Limiting vocab in data/wiki/ca-2/ca.wiki.valid.tokens.unk. Writing to data/wiki/ca-2/ca.wiki.valid.tokens.unk.\n",
            "ca.wiki.valid.tokens.unk. # of tokens: 207431\n",
            "Limiting vocab in data/wiki/ca-2/ca.wiki.test.tokens.unk. Writing to data/wiki/ca-2/ca.wiki.test.tokens.unk.\n",
            "ca.wiki.test.tokens.unk. # of tokens: 225004\n",
            "Limiting vocab in data/wiki/ca-100/ca.wiki.train.tokens.unk. Writing to data/wiki/ca-100/ca.wiki.train.tokens.unk.\n",
            "ca.wiki.train.tokens.unk. # of tokens: 101145977\n",
            "Limiting vocab in data/wiki/ca-100/ca.wiki.valid.tokens.unk. Writing to data/wiki/ca-100/ca.wiki.valid.tokens.unk.\n",
            "ca.wiki.valid.tokens.unk. # of tokens: 207431\n",
            "Limiting vocab in data/wiki/ca-100/ca.wiki.test.tokens.unk. Writing to data/wiki/ca-100/ca.wiki.test.tokens.unk.\n",
            "ca.wiki.test.tokens.unk. # of tokens: 225004\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "U_GLMGTOfKgO",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "trn_path = Path('data/wiki/ca-2/ca.wiki.train.tokens')\n",
        "val_path = Path('data/wiki/ca-2/ca.wiki.valid.tokens')\n",
        "\n",
        "with open(trn_path, encoding='utf8') as f:\n",
        "  train = f.readlines()\n",
        "\n",
        "df_train = pd.DataFrame(\n",
        "  {'content': train},\n",
        "  columns=['content'])\n",
        "\n",
        "df_train['is_valid'] = 0\n",
        "\n",
        "with open(val_path, encoding='utf8') as f:\n",
        "      val = f.readlines()\n",
        "\n",
        "df_valid = pd.DataFrame(\n",
        "  {'content': val},\n",
        "  columns=['content'])\n",
        "\n",
        "df_valid['is_valid'] = 1\n",
        "\n",
        "\n",
        "df_regroup = pd.concat([df_train, df_valid])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dJ8Qm1fjZu_N",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 197
        },
        "outputId": "d79334c0-d778-482e-fbc6-770c7d43dc43"
      },
      "source": [
        "df_regroup.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>content</th>\n",
              "      <th>is_valid</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Milledgeville ( Illinois )\\n</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>\\n</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Milledgeville és una població dels Estats Unit...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>\\n</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>Demografia .\\n</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                             content  is_valid\n",
              "0                       Milledgeville ( Illinois )\\n         0\n",
              "1                                                 \\n         0\n",
              "2  Milledgeville és una població dels Estats Unit...         0\n",
              "3                                                 \\n         0\n",
              "4                                     Demografia .\\n         0"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dI1boonXZxTR",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 197
        },
        "outputId": "8d090447-0235-4e6c-c171-23c72531815f"
      },
      "source": [
        "df_regroup = df_regroup[df_regroup.content != '\\n']\n",
        "df_regroup.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>content</th>\n",
              "      <th>is_valid</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Milledgeville ( Illinois )\\n</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Milledgeville és una població dels Estats Unit...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>Demografia .\\n</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5</th>\n",
              "      <td>Segons el cens del 2000 , Milledgeville tenia ...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7</th>\n",
              "      <td>Dels 448 habitatges en un 27,5 % hi vivien nen...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                             content  is_valid\n",
              "0                       Milledgeville ( Illinois )\\n         0\n",
              "2  Milledgeville és una població dels Estats Unit...         0\n",
              "4                                     Demografia .\\n         0\n",
              "5  Segons el cens del 2000 , Milledgeville tenia ...         0\n",
              "7  Dels 448 habitatges en un 27,5 % hi vivien nen...         0"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bbiV2q_cZtmS",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df_regroup.to_csv('data/wiki/full_train.csv', header=None, index=None)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OJUo_MUY6au6",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 16
        },
        "outputId": "a3bcbc25-5e22-4a13-a6d5-161616a22802"
      },
      "source": [
        "tokenizer = Tokenizer(lang='ca', n_cpus=6)\n",
        "data_lm_full = (TextList.from_csv('data/wiki', csv_name='full_train.csv', cols=0, processor=[TokenizeProcessor(tokenizer=tokenizer), NumericalizeProcessor(max_vocab=60000)])\n",
        "           #Inputs: all the text files in path\n",
        "            .split_from_df(col=1)\n",
        "           #We may have other temp folders that contain text files so we only keep what's in train and test\n",
        "            .label_for_lm()           \n",
        "           #We want to do a language model so we label accordingly\n",
        "            .databunch(bs=32))\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "A3iAkSbc6b5f",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        },
        "outputId": "f934f1b3-605c-45a1-9c14-9a59d60a09ed"
      },
      "source": [
        "itos = data_lm_full.train_ds.vocab.itos\n",
        "print('Size of vocabulary:', len(itos))\n",
        "print('First 10 words in vocab:', ', '.join([itos[i] for i in range(10)]))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Size of vocabulary: 30656\n",
            "First 10 words in vocab: xxunk, xxpad, xxbos, xxeos, xxfld, xxmaj, xxup, xxrep, xxwrep, de\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8BOsNVQ1j87N",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "data_lm_full.save('full_lm_30k')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lninUmS0A24a",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "d4f4d278-4e55-4e45-862a-7a1e830007a1"
      },
      "source": [
        "torch.cuda.is_available()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "True"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "i6MYpkF9A24d",
        "colab_type": "text"
      },
      "source": [
        "# Entrenar el Language Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WtXtjb_kA24d",
        "colab_type": "text"
      },
      "source": [
        "Hem usat com a dataset tot el text que s'ha trobat en el repositori https://github.com/Softcatala/ca-text-corpus. També hem provat d'entrenar amb un dataset creat a partir del script **./get_wikimedia.sh** però ens tornava un dataset massa brut. "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5RScd6CFA241",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "learn = language_model_learner(data_lm_full,  arch=AWD_LSTM, drop_mult=0, callback_fns=ShowGraph)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_l-U9Aq9A243",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "83fb9f24-a329-40e2-ea1c-005a4a45cb65"
      },
      "source": [
        "# save vocabulary\n",
        "print('Saving vocabulary...')\n",
        "with open(os.path.join('data', 'itos.pkl'), 'wb') as f:\n",
        "    pickle.dump(itos, f)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saving vocabulary...\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "V8f1QFA-A249",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 370
        },
        "outputId": "66b36bef-345e-48b6-b83e-6014fcb6fba2"
      },
      "source": [
        "data_lm_full.show_batch()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/fastai/text/data.py:339: UserWarning: This overload of nonzero is deprecated:\n",
            "\tnonzero()\n",
            "Consider using one of the following signatures instead:\n",
            "\tnonzero(*, bool as_tuple) (Triggered internally at  /pytorch/torch/csrc/utils/python_arg_parser.cpp:766.)\n",
            "  idx_min = (t != self.pad_idx).nonzero().min()\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th>idx</th>\n",
              "      <th>text</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>km ² . \\n  xxbos xxmaj dels 448 habitatges en un 27,5 % hi vivien nens de menys de 18 anys , en un 55,6 % hi vivien parelles casades , en un 7,4 % dones solteres , i en un 34,6 % no eren unitats familiars . xxmaj en el 31,9 % dels habitatges hi vivien persones soles el 18,1 % de les quals corresponia a persones de</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>la seva xxmaj conferència , amb 106 punts . xxmaj després de guanyar en la fase final als xxmaj new xxmaj york xxmaj xxunk , els xxmaj montreal xxmaj canadiens i els xxmaj philadelphia xxmaj xxunk , els xxmaj tampa xxmaj bay xxmaj lightning es van proclamar campions de la xxmaj conferència xxmaj est i va aconseguir arribar per primera vegada en la seva història a la final de la</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>de les famílies i el 9,5 % de la població estaven per davall del llindar de pobresa . \\n  xxbos xxmaj poblacions properes . \\n  xxbos xxmaj el següent diagrama mostra les poblacions més properes . \\n  xxbos xxmaj simpson ( xxmaj illinois ) \\n  xxbos xxmaj simpson és una població dels xxmaj estats xxmaj units a l ' estat d ' xxmaj illinois . xxmaj</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>% entre 25 i 44 , un 18,9 % de 45 a 60 i un 9,8 % 65 anys o més . \\n  xxbos l ' edat mediana era de 32 anys . xxmaj per cada 100 dones de 18 o més anys hi havia 98,5 homes . \\n  xxbos xxmaj la renda mediana per habitatge era de xxunk $ i la renda mediana per família de xxunk</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>i el 12,3 % de la població estaven per davall del llindar de pobresa . \\n  xxbos xxmaj poblacions més properes . \\n  xxbos xxmaj el següent diagrama mostra les poblacions més properes . \\n  xxbos xxmaj crooksville \\n  xxbos xxmaj crooksville és una població dels xxmaj estats xxmaj units a l ' estat d ' xxmaj ohio . xxmaj segons el cens del 2000 tenia</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "id": "Vh0ppCLoA25B",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 112
        },
        "outputId": "28ed01e1-5186-4997-8622-4c3001398ef8"
      },
      "source": [
        "learn.lr_find()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='0' class='' max='1' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      0.00% [0/1 00:00<00:00]\n",
              "    </div>\n",
              "    \n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "  </tbody>\n",
              "</table><p>\n",
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='99' class='' max='1012' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      9.78% [99/1012 00:35<05:24 9.8540]\n",
              "    </div>\n",
              "    "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Rj8hap8HA25F",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 278
        },
        "outputId": "893cfbb7-c706-439c-80fc-39b3e52ab15d"
      },
      "source": [
        "learn.recorder.plot(skip_start=0)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IQfRakLqA25H",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 505
        },
        "outputId": "7822d796-f754-4856-8fbe-4fe67004baaf"
      },
      "source": [
        "learn.fit_one_cycle(7, 3e-2, moms=(0.8, 0.7))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>3.251645</td>\n",
              "      <td>3.121398</td>\n",
              "      <td>0.464436</td>\n",
              "      <td>06:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>3.198326</td>\n",
              "      <td>3.160502</td>\n",
              "      <td>0.464463</td>\n",
              "      <td>06:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>3.011875</td>\n",
              "      <td>3.129758</td>\n",
              "      <td>0.465249</td>\n",
              "      <td>06:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>2.760864</td>\n",
              "      <td>3.074570</td>\n",
              "      <td>0.470408</td>\n",
              "      <td>06:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>2.441473</td>\n",
              "      <td>3.031145</td>\n",
              "      <td>0.472679</td>\n",
              "      <td>06:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>2.088003</td>\n",
              "      <td>3.039873</td>\n",
              "      <td>0.474470</td>\n",
              "      <td>06:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.830366</td>\n",
              "      <td>3.079516</td>\n",
              "      <td>0.471844</td>\n",
              "      <td>06:12</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "F6zngdsXRPTE",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "learn.save('model_60k-vocab-2')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xZXLZcSSA25J",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "21745bed-5138-4a0e-e7a5-46e7482787ad"
      },
      "source": [
        "np.exp(3.07)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "21.54190267500241"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 43
        }
      ]
    }
  ]
}