7 - XNLI_turkish.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "XNLI-turkish.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "rN4MX_O-3TJc"
      },
      "source": [
        "from fastai import *\n",
        "from fastai.text import * \n",
        "def random_seed(seed_value, use_cuda):\n",
        "    np.random.seed(seed_value) # cpu vars\n",
        "    torch.manual_seed(seed_value) # cpu  vars\n",
        "    random.seed(seed_value) # Python\n",
        "    if use_cuda: \n",
        "        torch.cuda.manual_seed(seed_value)\n",
        "        torch.cuda.manual_seed_all(seed_value) # gpu vars\n",
        "        torch.backends.cudnn.deterministic = True  #needed\n",
        "        torch.backends.cudnn.benchmark = False\n",
        "        \n",
        "random_seed(42, True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TZH-tZxA3aea",
        "outputId": "6bc2f00f-09b1-4314-8ab6-4af7edec3022"
      },
      "source": [
        "# bring en files (all then filter)\n",
        "!wget https://raw.githubusercontent.com/abedkhooli/ds2/master/ulmfit2/prepare_xnli.sh "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2021-06-13 10:15:07--  https://raw.githubusercontent.com/abedkhooli/ds2/master/ulmfit2/prepare_xnli.sh\n",
            "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
            "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.108.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 772 [text/plain]\n",
            "Saving to: ‘prepare_xnli.sh’\n",
            "\n",
            "\rprepare_xnli.sh       0%[                    ]       0  --.-KB/s               \rprepare_xnli.sh     100%[===================>]     772  --.-KB/s    in 0s      \n",
            "\n",
            "2021-06-13 10:15:07 (55.0 MB/s) - ‘prepare_xnli.sh’ saved [772/772]\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2zyQ_7gJ3jY-",
        "outputId": "65223f4c-d630-4188-ca7a-91b71bc6c1bf"
      },
      "source": [
        "filename_dev = 'xnli.dev.tsv'\n",
        "filename_test = 'xnli.test.tsv'\n",
        "df_dev = pd.read_csv(filename_dev, delimiter='\\t', quoting=csv.QUOTE_NONE)\n",
        "df_test = pd.read_csv(filename_test, delimiter='\\t', quoting=csv.QUOTE_NONE)\n",
        "len(df_dev), len(df_test)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(37350, 75150)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jaCHzmGM4p5G",
        "outputId": "b205636e-de8f-4499-a719-6aaade829644"
      },
      "source": [
        "df_dev_tr= df_dev[df_dev['language']=='tr']\n",
        "df_test_tr= df_test[df_test['language']=='tr']\n",
        "len(df_dev_tr), len(df_test_tr)\n",
        "# Each language has 2490 dev samples and 5010 test samples (from readme)."
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(2490, 5010)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 289
        },
        "id": "QMk8wXgq4xVk",
        "outputId": "b1cd7c37-e87a-4cc4-c7c9-d6cc278e46a6"
      },
      "source": [
        "df_dev_tr = df_dev_tr[['gold_label','sentence1_tokenized', 'sentence2_tokenized']]\n",
        "df_test_tr = df_test_tr[['gold_label','sentence1_tokenized', 'sentence2_tokenized']]\n",
        "df_dev_tr['is_valid'] = False\n",
        "df_test_tr['is_valid'] = True\n",
        "df_test_tr.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>gold_label</th>\n",
              "      <th>sentence1_tokenized</th>\n",
              "      <th>sentence2_tokenized</th>\n",
              "      <th>is_valid</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>55110</th>\n",
              "      <td>contradiction</td>\n",
              "      <td>Pekala , bunu hiç düşünmemiştim ancak kafam ço...</td>\n",
              "      <td>Onunla bir daha konuşmadım .</td>\n",
              "      <td>True</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>55111</th>\n",
              "      <td>entailment</td>\n",
              "      <td>Pekala , bunu hiç düşünmemiştim ancak kafam ço...</td>\n",
              "      <td>O kadar çok üzgündüm ki onunla tekrar konuşmay...</td>\n",
              "      <td>True</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>55112</th>\n",
              "      <td>neutral</td>\n",
              "      <td>Pekala , bunu hiç düşünmemiştim ancak kafam ço...</td>\n",
              "      <td>Harika bir konuşma yaptık .</td>\n",
              "      <td>True</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>55113</th>\n",
              "      <td>neutral</td>\n",
              "      <td>Ve bunun bir ayrıcalık olduğunu sanıyordum , v...</td>\n",
              "      <td>O gün tarladaki tek kişi olmadığımın farkında ...</td>\n",
              "      <td>True</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>55114</th>\n",
              "      <td>entailment</td>\n",
              "      <td>Ve bunun bir ayrıcalık olduğunu sanıyordum , v...</td>\n",
              "      <td>AFFC Hava Kuvvetleri Kariyer alanında o numara...</td>\n",
              "      <td>True</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "          gold_label  ... is_valid\n",
              "55110  contradiction  ...     True\n",
              "55111     entailment  ...     True\n",
              "55112        neutral  ...     True\n",
              "55113        neutral  ...     True\n",
              "55114     entailment  ...     True\n",
              "\n",
              "[5 rows x 4 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "X4I3CZuR49Er",
        "outputId": "e1d07a1c-dd8a-48bb-a126-4ff4f2ec23fd"
      },
      "source": [
        "df_all = pd.concat([df_dev_tr,df_test_tr])\n",
        "df_all['gold_label'].value_counts()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "neutral          2500\n",
              "entailment       2500\n",
              "contradiction    2500\n",
              "Name: gold_label, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1RFrJTV85Djn"
      },
      "source": [
        "# create a data folder and copy train and test data into it \n",
        "Path('data/xnli_tr/').mkdir(parents=True, exist_ok=True)\n",
        "df_all.to_csv('data/xnli_tr/xnli_text.csv', index=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7koFddZ75I5J"
      },
      "source": [
        "processor = [TokenizeProcessor(mark_fields=True), NumericalizeProcessor(max_vocab=60000)]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kl7kt_GJ5Npy",
        "outputId": "0e1e6494-7b01-422c-819d-1601db6b623f"
      },
      "source": [
        "path = Path('data/xnli_tr/')\n",
        "path.ls()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[PosixPath('data/xnli_tr/xnli_text.csv')]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 105
        },
        "id": "zSbwglri5RaB",
        "outputId": "d2b7eea0-be12-4d96-9dc0-5308f9c9c1c5"
      },
      "source": [
        "data_lm = (TextList.from_csv(path, 'xnli_text.csv', cols=['sentence1_tokenized', 'sentence2_tokenized'],processor = processor)\n",
        "                   .random_split_by_pct()\n",
        "                   .label_for_lm()\n",
        "                   .databunch())\n",
        "data_lm.save()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/fastai/data_block.py:217: UserWarning: `random_split_by_pct` is deprecated, please use `split_by_rand_pct`.\n",
            "  warn(\"`random_split_by_pct` is deprecated, please use `split_by_rand_pct`.\")\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/fastai/core.py:302: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return np.array(a, dtype=dtype, **kwargs)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 818
        },
        "id": "DWWJNDjn5U5c",
        "outputId": "8c3e2fc9-7309-4917-d848-b0411340e6ad"
      },
      "source": [
        "data_lm.show_batch(10)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "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>xxunk yeri bulmaya çalışırken xxunk . xxbos xxfld 1 xxmaj ne için gittiğimi falan bilmiyordum , xxmaj washington ' da belirtilen bir yere rapor vermem gerekiyordu . xxfld 2 xxmaj washington ' a xxunk ne yapmam gerektiğini tam olarak biliyordum . xxbos xxfld 1 xxmaj ne için gittiğimi falan bilmiyordum , xxmaj washington ' da belirtilen bir yere rapor vermem gerekiyordu . xxfld 2 xxmaj ne xxunk çok emin değildim</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>sundurmanın önünden aşağı doğru yürüdü ve oraya doğru yürüyordu ve o da orada durdu . xxfld 2 xxmaj büyükanne xxunk xxunk . xxbos xxfld 1 xxmaj büyükanne kalktı ve sundurmanın önünden aşağı doğru yürüdü ve oraya doğru yürüyordu ve o da orada durdu . xxfld 2 xxmaj büyükanne xxunk 100 adım yürüdü . xxbos xxfld 1 xxmaj büyükanne kalktı ve sundurmanın önünden aşağı doğru yürüdü ve oraya doğru yürüyordu ve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>şeyin etek ve etek ve bluz veya takım elbise veya elbise olduğunu biliyorsunuz , bu yüzden evde çalışmama rağmen güzel çünkü pantolon giyebiliyorum xxfld 2 xxmaj evde çalışırken hala elbise xxunk çünkü daha süslü hissediyorum . xxbos xxfld 1 evet , bilmem gereken bir yere sahip olmakla ilgili bir şey var xxfld 2 xxmaj yaşayacak bir yere sahip olma hayali gerçek oldu . xxbos xxfld 1 evet , bilmem gereken</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>anlatacak ne yapıyoruz ? xxbos xxfld 1 xxmaj tabii ki onlar , onlar için kesinlikle xxunk bildiğiniz yerlerde xxunk , ama geniş geniş xxunk geleceğini xxunk . xxfld 2 xxmaj bazı insanlar yaşlı akrabalarını önemsemiyor çünkü xxunk çok sayıda merdiven var . xxbos xxfld 1 xxmaj tabii ki onlar , onlar için kesinlikle xxunk bildiğiniz yerlerde xxunk , ama geniş geniş xxunk geleceğini xxunk . xxfld 2 xxmaj bazı insanlar</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>ulaşır ve tablonun sınırlarına kadar uzanır . xxfld 2 xxmaj kum her zaman düz ve xxunk . xxbos xxfld 1 xxmaj kum yığılırken , sonuç olarak kum için dinlenme açısına ulaşır ve tablonun sınırlarına kadar uzanır . xxfld 2 xxmaj kum bir buldozer yardımıyla xxunk . xxbos xxfld 1 xxmaj biyosfer , sürekli xxunk bitişik olana doğru xxunk , hatta , az ya da çok ısrarla patladı . xxfld 2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>kondisyonun doruklarına doğru çekmeye zorlayabilir . xxfld 2 xxmaj daha yüksek fitness seviyelerinde , xxunk xxunk hayatta kalma olasılığı daha yüksektir . xxbos xxfld 1 xxmaj fitness manzarasında , görüntü , mutasyon , rekombinasyon ve seçim , gelişen nüfusları yüksek kondisyonun doruklarına doğru çekmeye zorlayabilir . xxfld 2 xxmaj xxunk nüfusları daha yüksek xxunk xxunk çekmek için birçok etmen aynı anda xxunk . xxbos xxfld 1 xxmaj fitness manzarasında ,</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>xxunk . xxfld 2 xxmaj onların xxunk xxunk xxunk geldi . xxbos xxfld 1 xxmaj hangisi daha iyi bir yol ? talep etti . xxfld 2 xxmaj en iyi yolu xxunk bildiği için herhangi bir soru sormadı . xxbos xxfld 1 xxmaj hangisi daha iyi bir yol ? talep etti . xxfld 2 xxmaj daha iyi bir yol olup olmadığını bilmek istiyordu . xxbos xxfld 1 xxmaj hangisi daha iyi</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>gelmesi ile sürekli olarak artan sırasında belindeki parmağını erkeklere doğru aşağılayıcı bir biçimde salladı . xxfld 2 xxmaj herkes üst xxunk xxunk yol xxunk xxunk xxunk xxunk . xxbos xxfld 1 xxmaj bilakis bu onun heyecanını artırmıştı . xxfld 2 o heyecanını azalttı . xxbos xxfld 1 xxmaj bilakis bu onun heyecanını artırmıştı . xxfld 2 xxmaj bu , onu hiç olmadığı kadar heyecanlı biri yaptı . xxbos xxfld 1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>dır . xxfld 2 xxmaj brahma xxmaj hindu xxunk bir parçasıdır . xxbos xxfld 1 xxmaj xxunk xxunk aldı , ancak silahlı mücadele xxunk devam etti ve ondan sonra da uzun süre devam eden acı . xxfld 2 xxmaj savaş sadece bir günde sona xxunk . xxbos xxfld 1 xxmaj xxunk xxmaj xxunk ' ın altındaki birkaç yüz xxunk xxunk , xxmaj atinalılar ' ın xxmaj xxunk xxunk tahliye xxunk</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>xxfld 1 o gece en büyük kalabalığı çekmek amacıyla özel kuponlar gün boyu plajlarda yoğun şekilde dağıtılmakta . xxfld 2 xxmaj uzun vadeli müşteri umuduyla xxunk kuponlar xxunk . xxbos xxfld 1 o gece en büyük kalabalığı çekmek amacıyla özel kuponlar gün boyu plajlarda yoğun şekilde dağıtılmakta . xxfld 2 xxmaj kuponlar xxunk xxunk çekmek umuduyla sahilde xxunk . xxbos xxfld 1 . xxmaj passeig de xxmaj gracia ' ya</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 453
        },
        "id": "tj2P9F1152OV",
        "outputId": "df3711bb-f418-4dcc-8185-1896cac30f8c"
      },
      "source": [
        "learn = language_model_learner(data_lm, AWD_LSTM, drop_mult=0.1)\n",
        "#learn = language_model_learner(data_lm, pretrained_model=AWD_LSTM, drop_mult=0.1)\n",
        "learn.lr_find()\n",
        "learn.recorder.plot(skip_end=12)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading https://s3.amazonaws.com/fast-ai-modelzoo/wt103-fwd.tgz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "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='2' class='' max='3' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      66.67% [2/3 13:37<06:48]\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",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>8.918973</td>\n",
              "      <td>#na#</td>\n",
              "      <td>06:49</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>7.275417</td>\n",
              "      <td>#na#</td>\n",
              "      <td>06:48</td>\n",
              "    </tr>\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='7' class='' max='46' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      15.22% [7/46 03:20<18:38 9.2176]\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"
        },
        {
          "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": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 80
        },
        "id": "Dz0LuSX756z9",
        "outputId": "804fdd98-e124-433f-9965-4eb338be34cc"
      },
      "source": [
        "learn.fit_one_cycle(1, 4e-2) # was 10 "
      ],
      "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>5.336092</td>\n",
              "      <td>3.936434</td>\n",
              "      <td>0.406771</td>\n",
              "      <td>07:41</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "eFvmDJ-I7XMo",
        "outputId": "335866b6-7952-4932-ff1e-f1c9eba2c1ef"
      },
      "source": [
        "learn.recorder.plot_losses()"
      ],
      "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": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 289
        },
        "id": "PS9rLTa1Rbh-",
        "outputId": "30b72294-8af9-415f-86f6-879dfea26011"
      },
      "source": [
        "# first amount of words (here 10), the next 10 target words (actual) and the ones predicted.\n",
        "learn.show_results(max_len = 10)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th>text</th>\n",
              "      <th>target</th>\n",
              "      <th>pred</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>xxbos xxfld 1 xxmaj onunla ilgili xxunk , bütün gece</td>\n",
              "      <td>çok huzurlu olan , xxunk erkeklerin çılgınca aktif xxunk vardı</td>\n",
              "      <td>xxunk xxunk xxunk , xxunk xxunk xxunk xxunk bir xxunk</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>xxunk , sosyal ve politik adaletin xxunk inşa edilmiştir .</td>\n",
              "      <td>xxbos xxfld 1 yaklaşık yüzde xxunk sülfür bitümlü kömür xxunk</td>\n",
              "      <td>xxbos xxfld 1 xxmaj xxunk on ve . . .</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>tüm bilgiyi xxunk için hayal kırıklığı yaşadığımı ona açıklamaya çalıştım</td>\n",
              "      <td>. xxfld 2 xxmaj ona başka bir şey duymak xxunk</td>\n",
              "      <td>. xxfld 2 xxmaj xxunk xxunk xxfld xxunk xxunk .</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>yok bu yüzden söylemek pek kolay değil xxfld 2 xxmaj</td>\n",
              "      <td>yedi çocuğum var yani xxunk xxunk biliyorum ! xxbos xxfld</td>\n",
              "      <td>xxunk xxfld xxfld . xxunk xxunk . . xxbos xxfld</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1 xxmaj sant xxmaj xxunk ' nun xxmaj romanesk xxunk</td>\n",
              "      <td>xxunk , xxmaj barselona ' nın xxunk ve xxmaj gotik</td>\n",
              "      <td>. . xxunk xxunk ' nın xxunk xxunk xxunk xxunk</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2EEBw8mORlPJ"
      },
      "source": [
        "learn.save('mini_train_lm')\n",
        "learn.save_encoder('mini_train_encoder')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 122
        },
        "id": "iSU0H6wPRnSk",
        "outputId": "89cc5344-fb68-48a6-d37d-10b59237bc5b"
      },
      "source": [
        "data_clas = (TextList.from_csv(path, 'xnli_text.csv', cols=['sentence1_tokenized', 'sentence2_tokenized'],processor = processor, vocab=data_lm.vocab)\n",
        "                   .split_from_df(col='is_valid')\n",
        "                   .label_from_df(cols='gold_label')\n",
        "                   .databunch(bs=42))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/fastai/core.py:302: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return np.array(a, dtype=dtype, **kwargs)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 479
        },
        "id": "NlAD8Tr9RudM",
        "outputId": "6af9fed9-0337-47bd-dcad-c5bbe8207701"
      },
      "source": [
        "data_clas.show_batch()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\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>text</th>\n",
              "      <th>target</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>xxbos xxfld 1 6 : 45 ve 7 : 40 arası , xxmaj atta ve xxmaj omari , xxmaj xxunk al xxmaj xxunk , xxmaj wail al xxmaj shehri ve xxmaj xxunk al xxmaj shehri ile birlikte , xxunk - in yaptı ve xxmaj los xxmaj angeles ' a giden xxmaj american xxmaj xxunk 11 xxunk xxunk bindi . xxfld 2 xxmaj atta ve xxmaj omari xxunk yanı xxunk ve</td>\n",
              "      <td>neutral</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>xxbos xxfld 1 6 : 45 ve 7 : 40 arası , xxmaj atta ve xxmaj omari , xxmaj xxunk al xxmaj xxunk , xxmaj wail al xxmaj shehri ve xxmaj xxunk al xxmaj shehri ile birlikte , xxunk - in yaptı ve xxmaj los xxmaj angeles ' a giden xxmaj american xxmaj xxunk 11 xxunk xxunk bindi . xxfld 2 xxmaj xxunk gitmeye xxunk ama xxmaj wail al xxmaj</td>\n",
              "      <td>contradiction</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>xxbos xxfld 1 6 : 45 ve 7 : 40 arası , xxmaj atta ve xxmaj omari , xxmaj xxunk al xxmaj xxunk , xxmaj wail al xxmaj shehri ve xxmaj xxunk al xxmaj shehri ile birlikte , xxunk - in yaptı ve xxmaj los xxmaj angeles ' a giden xxmaj american xxmaj xxunk 11 xxunk xxunk bindi . xxfld 2 xxmaj los xxmaj angeles ' a saat xxunk :</td>\n",
              "      <td>entailment</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>xxbos xxfld 1 xxmaj böylece , ortalama olarak xxmaj webster ' s xxmaj xxunk xxmaj new xxmaj xxunk ve xxmaj xxunk xxmaj house xxmaj webster xxmaj xxunk , xxmaj amerikan xxmaj xxunk ve xxmaj webster ' ın xxmaj yeni xxmaj dünya ' sından en az yüzde on beş daha fazla bilgi içeriyor . xxfld 2 xxmaj webster ' s xxmaj college ' ta diğer xxunk çok daha fazla eksik bilgi</td>\n",
              "      <td>contradiction</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>xxbos xxfld 1 xxmaj böylece , ortalama olarak xxmaj webster ' s xxmaj xxunk xxmaj new xxmaj xxunk ve xxmaj xxunk xxmaj house xxmaj webster xxmaj xxunk , xxmaj amerikan xxmaj xxunk ve xxmaj webster ' ın xxmaj yeni xxmaj dünya ' sından en az yüzde on beş daha fazla bilgi içeriyor . xxfld 2 xxmaj webster ' s xxmaj college ' ta xxmaj yeni xxmaj dünyadan daha fazla bilgi</td>\n",
              "      <td>entailment</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0QQrCYPWgRdM",
        "outputId": "bc7bd844-0017-460c-a034-2e3543f471fc"
      },
      "source": [
        "learn_clas = text_classifier_learner(data_clas,AWD_LSTM, drop_mult=0.5)\n",
        "# learn = language_model_learner(data_lm, AWD_LSTM, drop_mult=0.5)\n",
        "\n",
        "learn_clas.load_encoder('mini_train_encoder')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "RNNLearner(data=TextClasDataBunch;\n",
              "\n",
              "Train: LabelList (2490 items)\n",
              "x: TextList\n",
              "xxbos xxfld 1 xxmaj ve xxmaj anne , xxunk dedi . xxfld 2 xxmaj okul servisi onu xxunk xxunk xxunk xxunk .,xxbos xxfld 1 xxmaj ve xxmaj anne , xxunk dedi . xxfld 2 xxmaj bir kelime söylemedi .,xxbos xxfld 1 xxmaj ve xxmaj anne , xxunk dedi . xxfld 2 xxmaj annesine eve gittiğini söyledi .,xxbos xxfld 1 xxmaj ne için gittiğimi falan bilmiyordum , xxmaj washington ' da belirtilen bir yere rapor vermem gerekiyordu . xxfld 2 xxmaj washington ' a hiç gitmedim , bu yüzden oraya xxunk yeri bulmaya çalışırken xxunk .,xxbos xxfld 1 xxmaj ne için gittiğimi falan bilmiyordum , xxmaj washington ' da belirtilen bir yere rapor vermem gerekiyordu . xxfld 2 xxmaj washington ' a xxunk ne yapmam gerektiğini tam olarak biliyordum .\n",
              "y: CategoryList\n",
              "neutral,contradiction,entailment,neutral,contradiction\n",
              "Path: data/xnli_tr;\n",
              "\n",
              "Valid: LabelList (5010 items)\n",
              "x: TextList\n",
              "xxbos xxfld 1 xxmaj pekala , bunu hiç düşünmemiştim ancak kafam çok karıştı ve onunla tekrar konuşmadım . xxfld 2 xxmaj onunla bir daha konuşmadım .,xxbos xxfld 1 xxmaj pekala , bunu hiç düşünmemiştim ancak kafam çok karıştı ve onunla tekrar konuşmadım . xxfld 2 o kadar çok xxunk ki onunla tekrar konuşmaya başladım .,xxbos xxfld 1 xxmaj pekala , bunu hiç düşünmemiştim ancak kafam çok karıştı ve onunla tekrar konuşmadım . xxfld 2 xxmaj harika bir konuşma yaptık .,xxbos xxfld 1 xxmaj ve bunun bir ayrıcalık olduğunu sanıyordum , ve hala , hala benim , xxup affc xxmaj hava xxmaj kuvvetleri xxmaj kariyer alanım olan dokuz tane iki iki x - o ' ydu . xxfld 2 o gün xxunk tek kişi xxunk farkında değildim .,xxbos xxfld 1 xxmaj ve bunun bir ayrıcalık olduğunu sanıyordum , ve hala , hala benim , xxup affc xxmaj hava xxmaj kuvvetleri xxmaj kariyer alanım olan dokuz tane iki iki x - o ' ydu . xxfld 2 xxup affc xxmaj hava xxmaj kuvvetleri xxmaj kariyer alanında o xxunk sahip olan tek kişi olduğum xxunk .\n",
              "y: CategoryList\n",
              "contradiction,entailment,neutral,neutral,entailment\n",
              "Path: data/xnli_tr;\n",
              "\n",
              "Test: None, model=SequentialRNN(\n",
              "  (0): MultiBatchEncoder(\n",
              "    (module): AWD_LSTM(\n",
              "      (encoder): Embedding(10720, 400, padding_idx=1)\n",
              "      (encoder_dp): EmbeddingDropout(\n",
              "        (emb): Embedding(10720, 400, padding_idx=1)\n",
              "      )\n",
              "      (rnns): ModuleList(\n",
              "        (0): WeightDropout(\n",
              "          (module): LSTM(400, 1152, batch_first=True)\n",
              "        )\n",
              "        (1): WeightDropout(\n",
              "          (module): LSTM(1152, 1152, batch_first=True)\n",
              "        )\n",
              "        (2): WeightDropout(\n",
              "          (module): LSTM(1152, 400, batch_first=True)\n",
              "        )\n",
              "      )\n",
              "      (input_dp): RNNDropout()\n",
              "      (hidden_dps): ModuleList(\n",
              "        (0): RNNDropout()\n",
              "        (1): RNNDropout()\n",
              "        (2): RNNDropout()\n",
              "      )\n",
              "    )\n",
              "  )\n",
              "  (1): PoolingLinearClassifier(\n",
              "    (layers): Sequential(\n",
              "      (0): BatchNorm1d(1200, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
              "      (1): Dropout(p=0.2, inplace=False)\n",
              "      (2): Linear(in_features=1200, out_features=50, bias=True)\n",
              "      (3): ReLU(inplace=True)\n",
              "      (4): BatchNorm1d(50, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
              "      (5): Dropout(p=0.1, inplace=False)\n",
              "      (6): Linear(in_features=50, out_features=3, bias=True)\n",
              "    )\n",
              "  )\n",
              "), opt_func=functools.partial(<class 'torch.optim.adam.Adam'>, betas=(0.9, 0.99)), loss_func=FlattenedLoss of CrossEntropyLoss(), metrics=[<function accuracy at 0x7f020d684a70>], true_wd=True, bn_wd=True, wd=0.01, train_bn=True, path=PosixPath('data/xnli_tr'), model_dir='models', callback_fns=[functools.partial(<class 'fastai.basic_train.Recorder'>, add_time=True, silent=False)], callbacks=[RNNTrainer\n",
              "learn: ...\n",
              "alpha: 2.0\n",
              "beta: 1.0], layer_groups=[Sequential(\n",
              "  (0): Embedding(10720, 400, padding_idx=1)\n",
              "  (1): EmbeddingDropout(\n",
              "    (emb): Embedding(10720, 400, padding_idx=1)\n",
              "  )\n",
              "), Sequential(\n",
              "  (0): WeightDropout(\n",
              "    (module): LSTM(400, 1152, batch_first=True)\n",
              "  )\n",
              "  (1): RNNDropout()\n",
              "), Sequential(\n",
              "  (0): WeightDropout(\n",
              "    (module): LSTM(1152, 1152, batch_first=True)\n",
              "  )\n",
              "  (1): RNNDropout()\n",
              "), Sequential(\n",
              "  (0): WeightDropout(\n",
              "    (module): LSTM(1152, 400, batch_first=True)\n",
              "  )\n",
              "  (1): RNNDropout()\n",
              "), Sequential(\n",
              "  (0): PoolingLinearClassifier(\n",
              "    (layers): Sequential(\n",
              "      (0): BatchNorm1d(1200, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
              "      (1): Dropout(p=0.2, inplace=False)\n",
              "      (2): Linear(in_features=1200, out_features=50, bias=True)\n",
              "      (3): ReLU(inplace=True)\n",
              "      (4): BatchNorm1d(50, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
              "      (5): Dropout(p=0.1, inplace=False)\n",
              "      (6): Linear(in_features=50, out_features=3, bias=True)\n",
              "    )\n",
              "  )\n",
              ")], add_time=True, silent=False)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 21
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 527
        },
        "id": "DIksZuxWgVTj",
        "outputId": "e21a3a7b-6c4b-48e6-f608-605625207c50"
      },
      "source": [
        "learn_clas.lr_find() # start_lr=1e-15, end_lr=1e2\n",
        "learn_clas.recorder.plot() # skip_end=4"
      ],
      "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='1' class='' max='2' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      50.00% [1/2 01:23<01:23]\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",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.207254</td>\n",
              "      <td>#na#</td>\n",
              "      <td>01:23</td>\n",
              "    </tr>\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='37' class='' max='59' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      62.71% [37/59 00:50<00:30 4.5786]\n",
              "    </div>\n",
              "    "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 397
        },
        "id": "Sj4EAI3ggl96",
        "outputId": "133b30eb-8cc5-44c7-d39b-876e544f25ce"
      },
      "source": [
        "learn_clas.fit_one_cycle(4, 4e-3)"
      ],
      "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>1.188049</td>\n",
              "      <td>1.125657</td>\n",
              "      <td>0.364870</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.132703</td>\n",
              "      <td>1.097534</td>\n",
              "      <td>0.359082</td>\n",
              "      <td>03:55</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.111475</td>\n",
              "      <td>1.090345</td>\n",
              "      <td>0.384032</td>\n",
              "      <td>03:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.095338</td>\n",
              "      <td>1.088027</td>\n",
              "      <td>0.390419</td>\n",
              "      <td>03:51</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 459
        },
        "id": "Bm_cfQpYhBc6",
        "outputId": "f978b60f-779c-4c96-aabf-feaa01b060af"
      },
      "source": [
        "learn_clas.show_results()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th>text</th>\n",
              "      <th>target</th>\n",
              "      <th>prediction</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>xxbos xxfld 1 1972 ' de xxmaj phillip xxmaj morris , xxmaj inc . ' in xxmaj miller xxmaj brewing xxmaj co. şirketi , xxmaj meister xxmaj brau xxmaj inc. şirketinin satınalmasında xxmaj lite bira etiketini satın aldı . xxfld 2 xxmaj phillip xxmaj morris , xxmaj inc . ' in xxunk xxmaj miller xxmaj brewing xxmaj co . xxmaj meister xxmaj brau xxmaj inc . , xxunk tümünü satın</td>\n",
              "      <td>contradiction</td>\n",
              "      <td>entailment</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>xxbos xxfld 1 1972 ' de xxmaj phillip xxmaj morris , xxmaj inc . ' in xxmaj miller xxmaj brewing xxmaj co. şirketi , xxmaj meister xxmaj brau xxmaj inc. şirketinin satınalmasında xxmaj lite bira etiketini satın aldı . xxfld 2 xxmaj phillip xxmaj morris , xxmaj inc . ' in xxmaj miller xxmaj brewing xxmaj co . , 1972 ' de xxmaj meister xxmaj brau xxmaj inc . '</td>\n",
              "      <td>entailment</td>\n",
              "      <td>entailment</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>xxbos xxfld 1 xxmaj daha fazla xxmaj anti - xxmaj defamation xxmaj league ' in araştırması , xxmaj amerika ' nın sert çekirdekli anti - semitlerinde , 1964 ' teki yüzde 29 ' dan 1992 ' de yüzde 20 ' ye , şu an yüzde 12 ' ye kadar devam eden bir düşüşü açıklıyor . xxfld 2 xxmaj xxunk ve xxmaj xxunk xxmaj mücadele xxmaj birliği anketi xxmaj amerika '</td>\n",
              "      <td>contradiction</td>\n",
              "      <td>contradiction</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>xxbos xxfld 1 xxmaj deborah xxmaj cameron ve xxmaj deborah xxmaj hills ( ' xxmaj radyo konuşması ' : radyo programlarına xxunk xxunk xxunk ve xxunk arasındaki ilişkileri xxunk ) xxmaj londra ' daki konuşma ve tartışma istasyonu olan ve ilgiyle xxunk xxup lbc xxmaj radyo ' nun xxunk xxunk . xxfld 2 xxmaj deborah xxmaj cameron ve xxmaj deborah xxmaj hills , xxup lbc xxmaj radyo ' da olanlarla</td>\n",
              "      <td>neutral</td>\n",
              "      <td>contradiction</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>xxbos xxfld 1 xxmaj daha fazla xxmaj anti - xxmaj defamation xxmaj league ' in araştırması , xxmaj amerika ' nın sert çekirdekli anti - semitlerinde , 1964 ' teki yüzde 29 ' dan 1992 ' de yüzde 20 ' ye , şu an yüzde 12 ' ye kadar devam eden bir düşüşü açıklıyor . xxfld 2 xxmaj xxunk xxmaj karşıtı xxmaj derneği ' nin anketi xxmaj amerika ' da</td>\n",
              "      <td>entailment</td>\n",
              "      <td>entailment</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "unjfumbqhE62",
        "outputId": "cfe5747c-dda3-4d34-a6b1-08b0701c3d11"
      },
      "source": [
        "learn_clas.recorder.plot_losses()"
      ],
      "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": "xGAckxKuhGs1"
      },
      "source": [
        "import os \n",
        "learn_clas.save(os.getcwd()+'mini_train_clas')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 889
        },
        "id": "puE-tWpVhKSF",
        "outputId": "0c76bdc3-1a43-4829-a678-cbe81f726164"
      },
      "source": [
        "#train for 10 epochs\n",
        "learn_clas.fit_one_cycle(10, 4e-3)"
      ],
      "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>1.084084</td>\n",
              "      <td>1.092469</td>\n",
              "      <td>0.380040</td>\n",
              "      <td>03:56</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.092571</td>\n",
              "      <td>1.087942</td>\n",
              "      <td>0.383433</td>\n",
              "      <td>03:53</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.102286</td>\n",
              "      <td>1.089049</td>\n",
              "      <td>0.373852</td>\n",
              "      <td>03:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.097809</td>\n",
              "      <td>1.103657</td>\n",
              "      <td>0.375050</td>\n",
              "      <td>03:51</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.093141</td>\n",
              "      <td>1.092478</td>\n",
              "      <td>0.379242</td>\n",
              "      <td>03:53</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.089430</td>\n",
              "      <td>1.085126</td>\n",
              "      <td>0.396607</td>\n",
              "      <td>03:52</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.087934</td>\n",
              "      <td>1.083125</td>\n",
              "      <td>0.395409</td>\n",
              "      <td>03:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.082444</td>\n",
              "      <td>1.081510</td>\n",
              "      <td>0.404192</td>\n",
              "      <td>03:52</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.079826</td>\n",
              "      <td>1.080719</td>\n",
              "      <td>0.408383</td>\n",
              "      <td>03:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.074489</td>\n",
              "      <td>1.081484</td>\n",
              "      <td>0.405589</td>\n",
              "      <td>03:53</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n"
          ],
          "name": "stderr"
        }
      ]
    }
  ]
}