lstm_and_gru/lstm_and_gru_pytorch.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "lstm_and_gru_pytorch.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "metadata": {
        "id": "ZIC0nEEjEjGG",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# LSTM\n",
        "\n",
        "In this notebook, we will go through basics of LSTM, GRU and Bi-LSTM using pretrained embeddings like glove and fasttext and  also custom embeddings on IMDB dataset. \n",
        "\n",
        "Here we will use [PyTorch](https://pytorch.org/tutorials/  \"PyTorch Tutorial\").\n",
        "\n",
        "\n",
        "Everything is explained in-detail in [blog post](https://dudeperf3ct.github.io/lstm/gru/nlp/2019/01/28/Force-of-LSTM-and-GRU/). This is notebook which replicates the result of blog and runs in colab. Enjoy!\n",
        "\n",
        "Let's see at [nlpprogess](http://nlpprogress.com/english/sentiment_analysis.html) what is the current state-of-the-art in sentiment analysis.\n",
        "\n",
        "\n",
        "Model | Accuracy | Paper |\n",
        "----- | -------- | ------|\n",
        "ULMFit| 95.4     | [Universal Language Model Fine-tuning for Text Classification](https://arxiv.org/abs/1801.06146)|\n",
        "Block-sparse LSTM| 94.99 | [GPU Kernels for Block-Sparse Weights](https://s3-us-west-2.amazonaws.com/openai-assets/blocksparse/blocksparsepaper.pdf)|\n",
        "oh-LSTM | 94.1 | [Supervised and Semi-Supervised Text Categorization using LSTM for Region Embeddings](https://arxiv.org/abs/1602.02373) |\n",
        "Virtual adversarial training  | 94.1 | [Adversarial Training Methods for Semi-Supervised Text Classification](https://arxiv.org/abs/1605.07725) |\n",
        "BCN+Char+CoVe | 91.8 | [Learned in Translation: Contextualized Word Vectors](https://arxiv.org/abs/1708.00107) |\n",
        "\n",
        "\n",
        "#### Run in Colab\n",
        "\n",
        "You can run this notebook in google colab.\n",
        "\n",
        "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/dudeperf3ct/DL_notebooks/blob/master/lstm_and_gru/lstm_and_gru_pytorch.ipynb)\n"
      ]
    },
    {
      "metadata": {
        "id": "CPX0rk2CUG-f",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## IMDB data\n",
        "\n",
        "Code Adapted from: [link](https://github.com/bentrevett/pytorch-sentiment-analysis/blob/master/1%20-%20Simple%20Sentiment%20Analysis.ipynb)"
      ]
    },
    {
      "metadata": {
        "id": "pNkINkerEmzY",
        "colab_type": "code",
        "outputId": "3bdad1f4-174f-49f2-bc5a-310716f12778",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "import time\n",
        "import copy\n",
        "import torch\n",
        "torch.manual_seed(11)\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "from torchtext import datasets\n",
        "from torchtext import data\n",
        "import torch.optim as optim\n",
        "import random\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "\n",
        "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
        "print ('Training on...', device)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Training on... cuda\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "2Z2FYcgwYaW3",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "TEXT = data.Field(tokenize='spacy')\n",
        "LABEL = data.LabelField(dtype=torch.float)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "QffNOcXb51zO",
        "colab_type": "code",
        "outputId": "e68db0d9-44bd-415f-8875-accaae7cbad3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 88
        }
      },
      "cell_type": "code",
      "source": [
        "train_data, test_data = datasets.IMDB.splits(TEXT, LABEL)\n",
        "print('Number of training examples:', len(train_data))\n",
        "print('Number of testing examples:', len(test_data))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "downloading aclImdb_v1.tar.gz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "aclImdb_v1.tar.gz: 100%|██████████| 84.1M/84.1M [00:07<00:00, 10.8MB/s]\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Number of training examples: 25000\n",
            "Number of testing examples: 25000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "gCUAAmu5516M",
        "colab_type": "code",
        "outputId": "ebc4f32c-5703-4950-923b-0e79f4f18ba2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 55
        }
      },
      "cell_type": "code",
      "source": [
        "print (vars(train_data.examples[2]))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "{'text': ['OK', ',', 'I', 'do', \"n't\", 'really', 'think', 'that', 'Trailer', 'Park', 'Boys', 'has', 'bad', 'story', 'lines', ',', 'because', 'they', 'kick', 'ass', '.', 'They', 'just', '...', 'conflict', 'with', 'each', 'other.<br', '/><br', '/>For', 'Example', ':', 'Near', 'the', 'end', 'of', 'the', 'movie', ',', 'it', 'shows', 'Ricky', 'and', 'Julian', 'telling', '\"', 'Patrick', 'Lewis', '\"', 'to', 'put', 'the', 'dog', 'down', 'and', 'walk', 'away', '.', 'Then', 'at', 'the', 'end', ',', 'it', 'shows', 'Ricky', 'and', 'Julian', 'saying', 'that', 'they', \"'ve\", 'been', 'in', 'jail', 'for', '2', 'years', '.', 'In', 'the', 'TV', 'series', 'pilot', ',', 'the', 'first', 'clip', 'they', 'show', 'is', 'the', 'same', 'clip', 'of', 'Ricky', 'and', 'Julian', 'yelling', 'at', '\"', 'Patrick', 'Lewis', '\"', '.', 'But', 'in', 'the', 'TV', 'series', ',', 'they', \"'ve\", 'supposedly', 'only', 'been', 'in', 'jail', 'for', '18', 'months.<br', '/><br', '/>Also', ',', 'they', 'give', 'us', 'the', 'impression', 'that', 'the', 'movie', \"'s\", 'story', 'line', 'and', 'the', 'TV', 'series', \"'\", 'story', 'line', 'are', 'connected', '(', 'because', 'of', 'the', 'yelling', 'scene', 'between', 'the', 'guys', ')', '.', 'But', 'some', 'actors', 'portray', 'totally', 'different', 'characters', '.', 'Of', 'course', ',', 'Patrick', 'Roach', 'plays', '\"', 'Patrick', 'Lewis', '\"', 'in', 'the', 'movie', ',', 'but', 'in', 'the', 'series', 'he', 'plays', 'Randy', '.', 'Sam', 'Tarasco', 'plays', 'one', 'of', 'the', 'guys', 'who', 'pays', 'Ricky', 'for', 'an', 'extermination', ',', 'and', 'then', 'he', 'plays', 'Sam', 'Losco', 'in', 'the', 'series.<br', '/><br', '/>Also', '(', 'again', '...', 'I', 'know', ',', 'I', 'have', 'a', 'lot', 'to', 'say', ')', ',', 'in', 'the', 'movie', ',', 'the', 'guys', 'snort', 'coke', 'instead', 'of', 'smoking', 'hash', '.', 'The', 'thing', 'is', ',', 'they', 'never', 'actually', 'confirm', 'that', 'the', 'two', 'story', 'lines', 'are', 'connected', 'in', 'anyway', ',', 'other', 'than', 'the', 'yelling', 'scene.<br', '/><br', '/>Sorry', 'to', 'keep', 'on', 'blabbing', '.'], 'label': 'pos'}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "av06h1VL51_b",
        "colab_type": "code",
        "outputId": "35787c50-6f7f-4afc-a13d-f6a95cbe288d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        }
      },
      "cell_type": "code",
      "source": [
        "train_data, val_data = train_data.split(split_ratio=0.8)\n",
        "print('Number of training examples:', len(train_data))\n",
        "print('Number of validation examples:', len(val_data))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of training examples: 20000\n",
            "Number of validation examples: 5000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "7jCn8RLXAVBm",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Glove"
      ]
    },
    {
      "metadata": {
        "id": "FDBBxWYzAW4m",
        "colab_type": "code",
        "outputId": "a28a2fc9-fdd8-4208-d725-0ff9fdbb0280",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        }
      },
      "cell_type": "code",
      "source": [
        "TEXT.build_vocab(train_data, max_size=25000, vectors=\"glove.6B.100d\")\n",
        "LABEL.build_vocab(train_data)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            ".vector_cache/glove.6B.zip: 862MB [02:11, 6.55MB/s]                           \n",
            "100%|█████████▉| 399966/400000 [00:20<00:00, 20356.93it/s]"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "id": "uPIrBa2dAW8p",
        "colab_type": "code",
        "outputId": "f6492d8b-eea9-4c6d-dde3-433e1eeeadca",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        }
      },
      "cell_type": "code",
      "source": [
        "print('Unique tokens in TEXT vocabulary:', len(TEXT.vocab))\n",
        "print('Unique tokens in LABEL vocabulary:',len(LABEL.vocab))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Unique tokens in TEXT vocabulary: 25002\n",
            "Unique tokens in LABEL vocabulary: 2\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "7PQEquoJBJFz",
        "colab_type": "code",
        "outputId": "0f07773b-b7d6-41ad-c3f0-f29bc7e0fbbe",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "cell_type": "code",
      "source": [
        "pretrained_embeddings = TEXT.vocab.vectors\n",
        "print(pretrained_embeddings.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([25002, 100])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "kea64FQNAXC-",
        "colab_type": "code",
        "outputId": "e739ffe5-b050-4345-b7a0-dfbae750428f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 55
        }
      },
      "cell_type": "code",
      "source": [
        "print(TEXT.vocab.freqs.most_common(20))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[('the', 232653), (',', 220816), ('.', 189387), ('and', 125574), ('a', 125107), ('of', 115377), ('to', 107754), ('is', 87429), ('in', 70454), ('I', 62331), ('it', 61420), ('that', 56650), ('\"', 50472), (\"'s\", 49694), ('this', 48616), ('-', 42468), ('/><br', 41014), ('was', 40184), ('as', 34867), ('with', 34370)]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "e8PI39viA7f9",
        "colab_type": "code",
        "outputId": "8e354ff1-4ef8-41ed-8fe8-d79a1b11d9f8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "cell_type": "code",
      "source": [
        "print(TEXT.vocab.itos[:10])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "['<unk>', '<pad>', 'the', ',', '.', 'and', 'a', 'of', 'to', 'is']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Yer56HFbA7dm",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "BATCH_SIZE = 64\n",
        "\n",
        "train_iterator, val_iterator, test_iterator = data.BucketIterator.splits(\n",
        "    (train_data, val_data, test_data),\n",
        "    batch_size=BATCH_SIZE,\n",
        "    device=device)\n",
        "\n",
        "embedding_dim=100\n",
        "hidden_dim=128"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "5sACQM2XA7ay",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "class LSTM_Model(nn.Module):\n",
        "    \n",
        "    def __init__(self, vocab_size, embedding_dim, hidden_dim, batch_size):\n",
        "      \n",
        "      super(LSTM_Model, self).__init__()\n",
        "      self.num_layers = 1\n",
        "      self.batch_size = batch_size\n",
        "      self.hidden_dim = hidden_dim\n",
        "      \n",
        "      self.word_embeddings = nn.Embedding(vocab_size, embedding_dim) \n",
        "      # The LSTM takes word embeddings as inputs, and outputs hidden states\n",
        "      # with dimensionality hidden_dim.\n",
        "      self.lstm = nn.LSTM(embedding_dim, hidden_dim, num_layers=self.num_layers) \n",
        "      self.fc = nn.Linear(hidden_dim, 1)\n",
        "      self.hidden = self.init_hidden()      \n",
        "      \n",
        "    def forward(self, sentence):\n",
        "      \n",
        "        embeds = self.word_embeddings(sentence)\n",
        "        # [sent_len, batch_size] --> [sent_len, batch_size, emb_dim]\n",
        "        lstm_out, self.hidden = self.lstm(embeds, self.hidden) \n",
        "        # [sent_len, batch_size, emb_dim] --> [seq_len, batch, num_directions*hidden_size]\n",
        "        preds = self.fc(lstm_out[-1].squeeze(0))\n",
        "        # [batch, num_directions * hidden_size] --> [batch_size, 1]\n",
        "        return preds\n",
        "      \n",
        "      \n",
        "    def init_hidden(self):\n",
        "        # Before we've done anything, we dont have any hidden state.\n",
        "        # The axes semantics are (num_layers, minibatch_size, hidden_dim)\n",
        "        return (torch.zeros(self.num_layers, self.batch_size, self.hidden_dim).to(device),\n",
        "                torch.zeros(self.num_layers, self.batch_size, self.hidden_dim).to(device))\n",
        "        "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "lYAzFBRCgWH_",
        "colab_type": "code",
        "outputId": "dc099ffe-f691-44ef-d20f-b23a3c8b1467",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 106
        }
      },
      "cell_type": "code",
      "source": [
        "model = LSTM_Model(vocab_size=len(TEXT.vocab), embedding_dim=100, hidden_dim=128, batch_size=BATCH_SIZE)\n",
        "model.to(device)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "LSTM_Model(\n",
              "  (word_embeddings): Embedding(25002, 100)\n",
              "  (lstm): LSTM(100, 128)\n",
              "  (fc): Linear(in_features=128, out_features=1, bias=True)\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "metadata": {
        "id": "4LqIfetoA7Xa",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model.word_embeddings.weight.data.copy_(pretrained_embeddings)\n",
        "model.word_embeddings.weight.requires_grad = False"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "as9gheTdgG9q",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "parameters = filter(lambda p: p.requires_grad, model.parameters())\n",
        "\n",
        "optimizer = optim.Adam(parameters, lr=1e-3, amsgrad=True)\n",
        "criterion = nn.BCEWithLogitsLoss()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Tdw2r12IgeYg",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Create training and validation dataloaders\n",
        "dataloaders_dict = {'train': train_iterator, \n",
        "                    'val': val_iterator}"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "oqMe184bgHDU",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def train_model(model, dataloaders, criterion, optimizer, num_epochs, batch_size=BATCH_SIZE):\n",
        "    since = time.time()\n",
        "\n",
        "    history = dict()\n",
        "\n",
        "    best_model_wts = copy.deepcopy(model.state_dict())\n",
        "    best_acc = 0.0\n",
        "    skip_count = 0\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        print('Epoch {}/{}'.format(epoch, num_epochs - 1))\n",
        "        print('-' * 10)\n",
        "\n",
        "        # Each epoch has a training and validation phase\n",
        "        for phase in ['train', 'val']:\n",
        "            if phase == 'train':\n",
        "                model.train()  # Set model to training mode\n",
        "            else:\n",
        "                model.eval()   # Set model to evaluate mode\n",
        "\n",
        "            running_loss = 0.0\n",
        "            running_corrects = 0\n",
        "            \n",
        "\n",
        "            # Iterate over data.\n",
        "            for data in dataloaders[phase]:\n",
        "                inputs, labels = data.text, data.label\n",
        "                inputs = inputs.to(device)\n",
        "                labels = labels.to(device)\n",
        "                # zero the parameter gradients\n",
        "                optimizer.zero_grad()\n",
        "                \n",
        "                # forward\n",
        "                # track history if only in train\n",
        "                with torch.set_grad_enabled(phase == 'train'):\n",
        "                    # Get model outputs and calculate loss\n",
        "\n",
        "                    # backward + optimize only if in training phase\n",
        "                    if phase == 'train':\n",
        "                        # we need to clear out the hidden state of the LSTM,\n",
        "                        # detaching it from its history on the last instance.\n",
        "                        model.batch_size = inputs.shape[1]\n",
        "                        model.hidden = model.init_hidden()\n",
        "                        \n",
        "                        outputs = model(inputs).squeeze(1)\n",
        "                        loss = criterion(outputs, labels)\n",
        "                        loss.backward()\n",
        "                        optimizer.step()\n",
        "                        \n",
        "                    else:\n",
        "                        model.batch_size = inputs.shape[1]\n",
        "                        model.hidden = model.init_hidden()\n",
        "                        \n",
        "                        outputs = model(inputs).squeeze(1)\n",
        "                        loss = criterion(outputs, labels)\n",
        "\n",
        "\n",
        "                # statistics\n",
        "                running_loss += loss.item()\n",
        "                outputs = torch.round(torch.sigmoid(outputs))\n",
        "                corrects = (outputs == labels).float()\n",
        "                acc = corrects.sum()/len(corrects)\n",
        "                running_corrects += acc.item()\n",
        "\n",
        "            epoch_loss = running_loss / len(dataloaders[phase])\n",
        "            epoch_acc = running_corrects / len(dataloaders[phase])\n",
        "\n",
        "            print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))\n",
        "\n",
        "            # deep copy the model\n",
        "            if phase == 'val' and epoch_acc > best_acc:\n",
        "                best_acc = epoch_acc\n",
        "                best_model_wts = copy.deepcopy(model.state_dict())\n",
        "            \n",
        "            if phase+'_acc' in history:\n",
        "                # append the new number to the existing array at this slot\n",
        "                history[phase+'_acc'].append(epoch_acc)\n",
        "            else:\n",
        "                # create a new array in this slot\n",
        "                history[phase+'_acc'] = [epoch_acc]\n",
        "            \n",
        "            if phase+'_loss' in history:\n",
        "                # append the new number to the existing array at this slot\n",
        "                history[phase+'_loss'].append(epoch_loss)\n",
        "            else:\n",
        "                # create a new array in this slot\n",
        "                history[phase+'_loss'] = [epoch_loss]            \n",
        "\n",
        "    time_elapsed = time.time() - since\n",
        "    print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))\n",
        "    print('Best val Acc: {:4f}'.format(best_acc))\n",
        "\n",
        "    # load best model weights\n",
        "    model.load_state_dict(best_model_wts)\n",
        "    return model, history\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "7Vx-7LsAgJbS",
        "colab_type": "code",
        "outputId": "d65803aa-798c-4fd6-aefb-1af19ba96dbb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 767
        }
      },
      "cell_type": "code",
      "source": [
        "model, history = train_model(model, dataloaders_dict, criterion, optimizer, num_epochs=10)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 0/9\n",
            "----------\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "\r100%|█████████▉| 399966/400000 [00:40<00:00, 20356.93it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "train Loss: 0.6934 Acc: 0.5037\n",
            "val Loss: 0.6932 Acc: 0.4990\n",
            "Epoch 1/9\n",
            "----------\n",
            "train Loss: 0.6931 Acc: 0.4985\n",
            "val Loss: 0.6921 Acc: 0.4956\n",
            "Epoch 2/9\n",
            "----------\n",
            "train Loss: 0.6929 Acc: 0.5042\n",
            "val Loss: 0.6913 Acc: 0.5257\n",
            "Epoch 3/9\n",
            "----------\n",
            "train Loss: 0.6928 Acc: 0.5014\n",
            "val Loss: 0.6919 Acc: 0.5047\n",
            "Epoch 4/9\n",
            "----------\n",
            "train Loss: 0.6925 Acc: 0.5001\n",
            "val Loss: 0.6912 Acc: 0.4921\n",
            "Epoch 5/9\n",
            "----------\n",
            "train Loss: 0.6921 Acc: 0.5014\n",
            "val Loss: 0.6927 Acc: 0.5222\n",
            "Epoch 6/9\n",
            "----------\n",
            "train Loss: 0.6920 Acc: 0.5073\n",
            "val Loss: 0.6901 Acc: 0.5109\n",
            "Epoch 7/9\n",
            "----------\n",
            "train Loss: 0.6918 Acc: 0.5032\n",
            "val Loss: 0.6893 Acc: 0.5190\n",
            "Epoch 8/9\n",
            "----------\n",
            "train Loss: 0.6915 Acc: 0.4978\n",
            "val Loss: 0.7057 Acc: 0.4929\n",
            "Epoch 9/9\n",
            "----------\n",
            "train Loss: 0.6906 Acc: 0.5013\n",
            "val Loss: 0.6934 Acc: 0.5443\n",
            "Training complete in 8m 46s\n",
            "Best val Acc: 0.544304\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "WkR7LTHxgJfu",
        "colab_type": "code",
        "outputId": "7bb2023c-6b48-4862-9de3-82d7b1502f80",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 571
        }
      },
      "cell_type": "code",
      "source": [
        "# summarize history for accuracy\n",
        "plt.plot(history['train_acc'])\n",
        "plt.plot(history['val_acc'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()\n",
        "# summarize history for loss\n",
        "plt.plot(history['train_loss'])\n",
        "plt.plot(history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEVCAYAAAAPRfkLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMcYYn/8H30zFl3gH2fkAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZEAAAEVCAYAAAAhANiZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n0jCLiCRpmEVEkjTMIiJJGvYFaRYo8FBTdoEAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "DlPrIesyhoH5",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "text",
        "id": "gWYj6yyzhoY_"
      },
      "cell_type": "markdown",
      "source": [
        "## fasttext"
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "eiuM3UV8hoZH",
        "outputId": "25d0b31f-804d-43c9-98fd-deb3e9f897e6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 4609
        }
      },
      "cell_type": "code",
      "source": [
        "TEXT.build_vocab(train_data, max_size=25000, vectors=\"fasttext.simple.300d\")\n",
        "LABEL.build_vocab(train_data)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            ".vector_cache/wiki.simple.vec: 0.00B [00:00, ?B/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   0%|          | 8.19k/293M [00:00<6:54:44, 11.8kB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   0%|          | 24.6k/293M [00:00<5:03:00, 16.1kB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   0%|          | 57.3k/293M [00:00<3:38:26, 22.4kB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   0%|          | 123k/293M [00:01<2:36:03, 31.3kB/s] \u001b[A\n",
            ".vector_cache/wiki.simple.vec:   0%|          | 246k/293M [00:01<1:50:53, 44.0kB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   0%|          | 459k/293M [00:01<1:18:32, 62.1kB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   0%|          | 893k/293M [00:01<55:22, 88.0kB/s]  \u001b[A\n",
            ".vector_cache/wiki.simple.vec:   1%|          | 1.70M/293M [00:01<38:55, 125kB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   1%|          | 3.40M/293M [00:01<27:12, 178kB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   2%|▏         | 6.00M/293M [00:01<18:57, 253kB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   3%|▎         | 8.68M/293M [00:02<13:13, 359kB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   4%|▍         | 11.2M/293M [00:02<09:14, 508kB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   5%|▍         | 14.2M/293M [00:02<06:28, 718kB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   6%|▌         | 16.9M/293M [00:02<04:33, 1.01MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   7%|▋         | 19.4M/293M [00:02<03:14, 1.41MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   8%|▊         | 22.2M/293M [00:02<02:18, 1.95MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   8%|▊         | 24.3M/293M [00:02<01:41, 2.64MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:   9%|▉         | 27.5M/293M [00:03<01:14, 3.59MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  10%|█         | 29.6M/293M [00:03<00:56, 4.66MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  11%|█         | 32.6M/293M [00:03<00:42, 6.08MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  12%|█▏        | 35.3M/293M [00:03<00:33, 7.60MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  13%|█▎        | 37.9M/293M [00:03<00:27, 9.26MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  14%|█▎        | 39.9M/293M [00:03<00:24, 10.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  15%|█▍        | 43.0M/293M [00:03<00:20, 12.2MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  16%|█▌        | 45.9M/293M [00:04<00:17, 13.9MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  17%|█▋        | 48.8M/293M [00:04<00:15, 15.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  18%|█▊        | 51.7M/293M [00:04<00:14, 16.6MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  19%|█▊        | 54.6M/293M [00:04<00:13, 17.6MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  20%|█▉        | 57.5M/293M [00:04<00:12, 18.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  21%|██        | 60.5M/293M [00:04<00:12, 19.0MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  22%|██▏       | 63.2M/293M [00:04<00:12, 19.1MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  23%|██▎       | 66.2M/293M [00:05<00:11, 19.6MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  24%|██▎       | 69.1M/293M [00:05<00:11, 19.8MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  25%|██▍       | 72.1M/293M [00:05<00:10, 20.1MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  26%|██▌       | 75.0M/293M [00:05<00:10, 20.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  27%|██▋       | 78.0M/293M [00:05<00:10, 20.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  28%|██▊       | 81.0M/293M [00:05<00:10, 20.6MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  29%|██▊       | 84.0M/293M [00:05<00:10, 20.7MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  30%|██▉       | 86.9M/293M [00:06<00:09, 20.7MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  31%|███       | 89.9M/293M [00:06<00:09, 20.8MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  32%|███▏      | 92.9M/293M [00:06<00:09, 20.9MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  33%|███▎      | 95.9M/293M [00:06<00:09, 20.9MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  34%|███▍      | 99.0M/293M [00:06<00:09, 21.2MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  35%|███▍      | 102M/293M [00:06<00:08, 21.3MB/s] \u001b[A\n",
            ".vector_cache/wiki.simple.vec:  36%|███▌      | 105M/293M [00:06<00:08, 21.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  37%|███▋      | 108M/293M [00:07<00:08, 21.5MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  38%|███▊      | 111M/293M [00:07<00:08, 21.5MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  39%|███▉      | 114M/293M [00:07<00:08, 21.2MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  40%|███▉      | 117M/293M [00:07<00:08, 21.0MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  41%|████      | 120M/293M [00:07<00:08, 21.2MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  42%|████▏     | 123M/293M [00:07<00:07, 21.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  43%|████▎     | 126M/293M [00:07<00:07, 21.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  44%|████▍     | 129M/293M [00:08<00:07, 21.5MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  45%|████▌     | 132M/293M [00:08<00:07, 21.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  46%|████▌     | 136M/293M [00:08<00:07, 21.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  47%|████▋     | 138M/293M [00:08<00:07, 21.1MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  48%|████▊     | 142M/293M [00:08<00:07, 21.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  49%|████▉     | 145M/293M [00:08<00:06, 21.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  50%|█████     | 148M/293M [00:08<00:06, 21.5MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  51%|█████▏    | 151M/293M [00:09<00:06, 21.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  52%|█████▏    | 154M/293M [00:09<00:06, 21.2MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  53%|█████▎    | 157M/293M [00:09<00:06, 21.0MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  54%|█████▍    | 159M/293M [00:09<00:06, 20.6MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  55%|█████▌    | 162M/293M [00:09<00:06, 20.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  56%|█████▋    | 165M/293M [00:09<00:06, 20.6MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  57%|█████▋    | 168M/293M [00:09<00:06, 20.6MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  58%|█████▊    | 171M/293M [00:10<00:05, 20.7MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  59%|█████▉    | 174M/293M [00:10<00:05, 20.7MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  60%|██████    | 177M/293M [00:10<00:05, 20.8MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  61%|██████▏   | 180M/293M [00:10<00:05, 20.9MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  62%|██████▏   | 183M/293M [00:10<00:05, 21.0MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  63%|██████▎   | 186M/293M [00:10<00:05, 20.9MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  64%|██████▍   | 189M/293M [00:10<00:04, 20.9MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  65%|██████▌   | 192M/293M [00:11<00:04, 20.8MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  67%|██████▋   | 195M/293M [00:11<00:04, 20.9MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  68%|██████▊   | 198M/293M [00:11<00:04, 21.0MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  69%|██████▊   | 201M/293M [00:11<00:04, 21.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  70%|██████▉   | 204M/293M [00:11<00:04, 21.2MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  71%|███████   | 207M/293M [00:11<00:04, 21.2MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  72%|███████▏  | 210M/293M [00:11<00:03, 21.2MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  73%|███████▎  | 213M/293M [00:12<00:03, 21.2MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  74%|███████▍  | 216M/293M [00:12<00:03, 21.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  75%|███████▍  | 219M/293M [00:12<00:03, 21.1MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  76%|███████▌  | 222M/293M [00:12<00:03, 21.1MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  77%|███████▋  | 225M/293M [00:12<00:03, 21.2MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  78%|███████▊  | 228M/293M [00:12<00:03, 21.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  79%|███████▉  | 231M/293M [00:12<00:02, 21.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  80%|███████▉  | 234M/293M [00:13<00:02, 21.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  81%|████████  | 238M/293M [00:13<00:02, 21.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  82%|████████▏ | 241M/293M [00:13<00:02, 21.6MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  83%|████████▎ | 244M/293M [00:13<00:02, 21.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  84%|████████▍ | 246M/293M [00:13<00:02, 20.9MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  85%|████████▌ | 249M/293M [00:13<00:02, 20.8MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  86%|████████▌ | 252M/293M [00:13<00:01, 20.9MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  87%|████████▋ | 255M/293M [00:14<00:01, 20.9MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  88%|████████▊ | 258M/293M [00:14<00:01, 21.0MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  89%|████████▉ | 262M/293M [00:14<00:01, 21.1MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  90%|█████████ | 265M/293M [00:14<00:01, 21.2MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  91%|█████████▏| 268M/293M [00:14<00:01, 21.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  92%|█████████▏| 271M/293M [00:14<00:01, 21.5MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  93%|█████████▎| 274M/293M [00:14<00:00, 21.5MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  94%|█████████▍| 277M/293M [00:15<00:00, 21.5MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  95%|█████████▌| 280M/293M [00:15<00:00, 21.5MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  97%|█████████▋| 283M/293M [00:15<00:00, 21.5MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  98%|█████████▊| 286M/293M [00:15<00:00, 21.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec:  99%|█████████▊| 289M/293M [00:15<00:00, 21.4MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec: 100%|█████████▉| 292M/293M [00:15<00:00, 21.3MB/s]\u001b[A\n",
            ".vector_cache/wiki.simple.vec: 293MB [00:15, 18.5MB/s]                           \u001b[A\n",
            "  0%|          | 0/111051 [00:00<?, ?it/s]\u001b[ASkipping token b'111051' with 1-dimensional vector [b'300']; likely a header\n",
            "\n",
            "  1%|          | 788/111051 [00:00<00:13, 7878.45it/s]\u001b[A\n",
            "  1%|▏         | 1508/111051 [00:00<00:14, 7660.43it/s]\u001b[A\n",
            "  2%|▏         | 2270/111051 [00:00<00:14, 7646.71it/s]\u001b[A\n",
            "  3%|▎         | 3002/111051 [00:00<00:14, 7543.39it/s]\u001b[A\n",
            "  3%|▎         | 3673/111051 [00:00<00:14, 7272.34it/s]\u001b[A\n",
            "  4%|▍         | 4409/111051 [00:00<00:14, 7297.99it/s]\u001b[A\n",
            "  5%|▍         | 5153/111051 [00:00<00:14, 7338.98it/s]\u001b[A\n",
            "  5%|▌         | 5915/111051 [00:00<00:14, 7418.18it/s]\u001b[A\n",
            "  6%|▌         | 6679/111051 [00:00<00:13, 7481.32it/s]\u001b[A\n",
            "  7%|▋         | 7435/111051 [00:01<00:13, 7503.01it/s]\u001b[A\n",
            "  7%|▋         | 8187/111051 [00:01<00:13, 7508.05it/s]\u001b[A\n",
            "  8%|▊         | 8942/111051 [00:01<00:13, 7518.83it/s]\u001b[A\n",
            "  9%|▉         | 9753/111051 [00:01<00:13, 7686.16it/s]\u001b[A\n",
            " 10%|▉         | 10569/111051 [00:01<00:12, 7820.72it/s]\u001b[A\n",
            " 10%|█         | 11418/111051 [00:01<00:12, 8008.26it/s]\u001b[A\n",
            " 11%|█         | 12219/111051 [00:01<00:12, 8006.70it/s]\u001b[A\n",
            " 12%|█▏        | 13075/111051 [00:01<00:12, 8163.76it/s]\u001b[A\n",
            " 13%|█▎        | 13892/111051 [00:01<00:11, 8116.53it/s]\u001b[A\n",
            " 13%|█▎        | 14704/111051 [00:01<00:11, 8116.67it/s]\u001b[A\n",
            " 14%|█▍        | 15516/111051 [00:02<00:12, 7944.30it/s]\u001b[A\n",
            " 15%|█▍        | 16312/111051 [00:02<00:12, 7826.59it/s]\u001b[A\n",
            " 15%|█▌        | 17096/111051 [00:02<00:12, 7666.23it/s]\u001b[A\n",
            " 16%|█▌        | 17865/111051 [00:02<00:12, 7612.10it/s]\u001b[A\n",
            " 17%|█▋        | 18636/111051 [00:02<00:12, 7640.73it/s]\u001b[A\n",
            " 17%|█▋        | 19401/111051 [00:02<00:12, 7558.70it/s]\u001b[A\n",
            " 18%|█▊        | 20158/111051 [00:02<00:12, 7412.06it/s]\u001b[A\n",
            " 19%|█▉        | 20901/111051 [00:02<00:12, 7408.37it/s]\u001b[A\n",
            " 20%|█▉        | 21659/111051 [00:02<00:11, 7456.34it/s]\u001b[A\n",
            " 20%|██        | 22429/111051 [00:02<00:11, 7525.70it/s]\u001b[A\n",
            " 21%|██        | 23183/111051 [00:03<00:12, 7318.07it/s]\u001b[A\n",
            " 22%|██▏       | 23917/111051 [00:03<00:12, 7221.49it/s]\u001b[A\n",
            " 22%|██▏       | 24655/111051 [00:03<00:11, 7266.60it/s]\u001b[A\n",
            " 23%|██▎       | 25421/111051 [00:03<00:11, 7379.82it/s]\u001b[A\n",
            " 24%|██▎       | 26189/111051 [00:03<00:11, 7465.14it/s]\u001b[A\n",
            " 24%|██▍       | 26950/111051 [00:03<00:11, 7505.96it/s]\u001b[A\n",
            " 25%|██▍       | 27702/111051 [00:03<00:11, 7423.57it/s]\u001b[A\n",
            " 26%|██▌       | 28475/111051 [00:03<00:10, 7511.84it/s]\u001b[A\n",
            " 26%|██▋       | 29227/111051 [00:03<00:11, 7437.80it/s]\u001b[A\n",
            " 27%|██▋       | 29972/111051 [00:03<00:10, 7419.48it/s]\u001b[A\n",
            " 28%|██▊       | 30715/111051 [00:04<00:10, 7415.96it/s]\u001b[A\n",
            " 28%|██▊       | 31490/111051 [00:04<00:10, 7512.49it/s]\u001b[A\n",
            " 29%|██▉       | 32243/111051 [00:04<00:10, 7516.49it/s]\u001b[A\n",
            " 30%|██▉       | 33025/111051 [00:04<00:10, 7603.29it/s]\u001b[A\n",
            " 30%|███       | 33807/111051 [00:04<00:10, 7665.90it/s]\u001b[A\n",
            " 31%|███       | 34575/111051 [00:04<00:09, 7662.06it/s]\u001b[A\n",
            " 32%|███▏      | 35342/111051 [00:04<00:10, 7412.05it/s]\u001b[A\n",
            " 33%|███▎      | 36093/111051 [00:04<00:10, 7441.01it/s]\u001b[A\n",
            " 33%|███▎      | 36866/111051 [00:04<00:09, 7525.19it/s]\u001b[A\n",
            " 34%|███▍      | 37629/111051 [00:04<00:09, 7554.18it/s]\u001b[A\n",
            " 35%|███▍      | 38392/111051 [00:05<00:09, 7574.51it/s]\u001b[A\n",
            " 35%|███▌      | 39162/111051 [00:05<00:09, 7611.72it/s]\u001b[A\n",
            " 36%|███▌      | 39948/111051 [00:05<00:09, 7684.18it/s]\u001b[A\n",
            " 37%|███▋      | 40720/111051 [00:05<00:09, 7693.20it/s]\u001b[A\n",
            " 37%|███▋      | 41498/111051 [00:05<00:09, 7717.73it/s]\u001b[A\n",
            " 38%|███▊      | 42271/111051 [00:05<00:08, 7708.36it/s]\u001b[A\n",
            " 39%|███▉      | 43043/111051 [00:05<00:09, 7445.33it/s]\u001b[A\n",
            " 39%|███▉      | 43790/111051 [00:05<00:09, 7349.38it/s]\u001b[A\n",
            " 40%|████      | 44544/111051 [00:05<00:08, 7405.48it/s]\u001b[A\n",
            " 41%|████      | 45294/111051 [00:05<00:08, 7433.24it/s]\u001b[A\n",
            " 41%|████▏     | 46045/111051 [00:06<00:08, 7455.26it/s]\u001b[A\n",
            " 42%|████▏     | 46799/111051 [00:06<00:08, 7479.12it/s]\u001b[A\n",
            " 43%|████▎     | 47557/111051 [00:06<00:08, 7508.04it/s]\u001b[A\n",
            " 44%|████▎     | 48309/111051 [00:06<00:08, 7500.95it/s]\u001b[A\n",
            " 44%|████▍     | 49088/111051 [00:06<00:08, 7584.30it/s]\u001b[A\n",
            " 45%|████▍     | 49847/111051 [00:06<00:08, 7583.45it/s]\u001b[A\n",
            " 46%|████▌     | 50606/111051 [00:06<00:08, 7366.88it/s]\u001b[A\n",
            " 46%|████▋     | 51370/111051 [00:06<00:08, 7445.58it/s]\u001b[A\n",
            " 47%|████▋     | 52116/111051 [00:06<00:07, 7438.57it/s]\u001b[A\n",
            " 48%|████▊     | 52878/111051 [00:07<00:07, 7488.58it/s]\u001b[A\n",
            " 48%|████▊     | 53644/111051 [00:07<00:07, 7537.24it/s]\u001b[A\n",
            " 49%|████▉     | 54406/111051 [00:07<00:07, 7561.75it/s]\u001b[A\n",
            " 50%|████▉     | 55176/111051 [00:07<00:07, 7601.20it/s]\u001b[A\n",
            " 50%|█████     | 55942/111051 [00:07<00:07, 7617.90it/s]\u001b[A\n",
            " 51%|█████     | 56705/111051 [00:07<00:07, 7607.32it/s]\u001b[A\n",
            " 52%|█████▏    | 57466/111051 [00:07<00:07, 7537.00it/s]\u001b[A\n",
            " 52%|█████▏    | 58220/111051 [00:07<00:07, 7320.95it/s]\u001b[A\n",
            " 53%|█████▎    | 58962/111051 [00:07<00:07, 7348.89it/s]\u001b[A\n",
            " 54%|█████▍    | 59723/111051 [00:07<00:06, 7423.68it/s]\u001b[A\n",
            " 54%|█████▍    | 60485/111051 [00:08<00:06, 7478.96it/s]\u001b[A\n",
            " 55%|█████▌    | 61243/111051 [00:08<00:06, 7507.46it/s]\u001b[A\n",
            " 56%|█████▌    | 62014/111051 [00:08<00:06, 7566.07it/s]\u001b[A\n",
            " 57%|█████▋    | 62792/111051 [00:08<00:06, 7628.26it/s]\u001b[A\n",
            " 57%|█████▋    | 63562/111051 [00:08<00:06, 7648.08it/s]\u001b[A\n",
            " 58%|█████▊    | 64328/111051 [00:08<00:06, 7633.74it/s]\u001b[A\n",
            " 59%|█████▊    | 65092/111051 [00:08<00:06, 7588.89it/s]\u001b[A\n",
            " 59%|█████▉    | 65852/111051 [00:08<00:06, 7259.65it/s]\u001b[A\n",
            " 60%|█████▉    | 66582/111051 [00:08<00:06, 7204.93it/s]\u001b[A\n",
            " 61%|██████    | 67340/111051 [00:08<00:05, 7310.52it/s]\u001b[A\n",
            " 61%|██████▏   | 68101/111051 [00:09<00:05, 7396.13it/s]\u001b[A\n",
            " 62%|██████▏   | 68851/111051 [00:09<00:05, 7425.08it/s]\u001b[A\n",
            " 63%|██████▎   | 69621/111051 [00:09<00:05, 7503.65it/s]\u001b[A\n",
            " 63%|██████▎   | 70400/111051 [00:09<00:05, 7587.24it/s]\u001b[A\n",
            " 64%|██████▍   | 71174/111051 [00:09<00:05, 7631.48it/s]\u001b[A\n",
            " 65%|██████▍   | 71952/111051 [00:09<00:05, 7674.13it/s]\u001b[A\n",
            " 65%|██████▌   | 72720/111051 [00:09<00:05, 7658.59it/s]\u001b[A\n",
            " 66%|██████▌   | 73487/111051 [00:09<00:05, 7336.13it/s]\u001b[A\n",
            " 67%|██████▋   | 74237/111051 [00:09<00:04, 7384.07it/s]\u001b[A\n",
            " 68%|██████▊   | 74978/111051 [00:09<00:04, 7259.15it/s]\u001b[A\n",
            " 68%|██████▊   | 75706/111051 [00:10<00:04, 7220.91it/s]\u001b[A\n",
            " 69%|██████▉   | 76451/111051 [00:10<00:04, 7287.76it/s]\u001b[A\n",
            " 70%|██████▉   | 77207/111051 [00:10<00:04, 7367.02it/s]\u001b[A\n",
            " 70%|███████   | 77966/111051 [00:10<00:04, 7430.62it/s]\u001b[A\n",
            " 71%|███████   | 78723/111051 [00:10<00:04, 7470.98it/s]\u001b[A\n",
            " 72%|███████▏  | 79479/111051 [00:10<00:04, 7495.14it/s]\u001b[A\n",
            " 72%|███████▏  | 80252/111051 [00:10<00:04, 7563.38it/s]\u001b[A\n",
            " 73%|███████▎  | 81009/111051 [00:10<00:04, 7353.64it/s]\u001b[A\n",
            " 74%|███████▎  | 81746/111051 [00:10<00:04, 7300.61it/s]\u001b[A\n",
            " 74%|███████▍  | 82500/111051 [00:10<00:03, 7366.87it/s]\u001b[A\n",
            " 75%|███████▍  | 83275/111051 [00:11<00:03, 7475.98it/s]\u001b[A\n",
            " 76%|███████▌  | 84063/111051 [00:11<00:03, 7592.66it/s]\u001b[A\n",
            " 76%|███████▋  | 84824/111051 [00:11<00:03, 7472.45it/s]\u001b[A\n",
            " 77%|███████▋  | 85584/111051 [00:11<00:03, 7508.37it/s]\u001b[A\n",
            " 78%|███████▊  | 86336/111051 [00:11<00:03, 7487.16it/s]\u001b[A\n",
            " 78%|███████▊  | 87105/111051 [00:11<00:03, 7544.84it/s]\u001b[A\n",
            " 79%|███████▉  | 87861/111051 [00:11<00:03, 7515.65it/s]\u001b[A\n",
            " 80%|███████▉  | 88613/111051 [00:11<00:03, 7219.83it/s]\u001b[A\n",
            " 80%|████████  | 89338/111051 [00:11<00:03, 7198.74it/s]\u001b[A\n",
            " 81%|████████  | 90095/111051 [00:11<00:02, 7303.29it/s]\u001b[A\n",
            " 82%|████████▏ | 90832/111051 [00:12<00:02, 7322.39it/s]\u001b[A\n",
            " 82%|████████▏ | 91586/111051 [00:12<00:02, 7385.90it/s]\u001b[A\n",
            " 83%|████████▎ | 92347/111051 [00:12<00:02, 7451.49it/s]\u001b[A\n",
            " 84%|████████▍ | 93117/111051 [00:12<00:02, 7523.50it/s]\u001b[A\n",
            " 85%|████████▍ | 93875/111051 [00:12<00:02, 7539.22it/s]\u001b[A\n",
            " 85%|████████▌ | 94630/111051 [00:12<00:02, 7429.61it/s]\u001b[A\n",
            " 86%|████████▌ | 95374/111051 [00:12<00:02, 7385.76it/s]\u001b[A\n",
            " 87%|████████▋ | 96114/111051 [00:12<00:02, 7223.67it/s]\u001b[A\n",
            " 87%|████████▋ | 96846/111051 [00:12<00:01, 7250.96it/s]\u001b[A\n",
            " 88%|████████▊ | 97617/111051 [00:13<00:01, 7381.77it/s]\u001b[A\n",
            " 89%|████████▊ | 98371/111051 [00:13<00:01, 7425.70it/s]\u001b[A\n",
            " 89%|████████▉ | 99138/111051 [00:13<00:01, 7495.11it/s]\u001b[A\n",
            " 90%|████████▉ | 99900/111051 [00:13<00:01, 7530.84it/s]\u001b[A\n",
            " 91%|█████████ | 100666/111051 [00:13<00:01, 7566.87it/s]\u001b[A\n",
            " 91%|█████████▏| 101424/111051 [00:13<00:01, 7570.22it/s]\u001b[A\n",
            " 92%|█████████▏| 102186/111051 [00:13<00:01, 7584.72it/s]\u001b[A\n",
            " 93%|█████████▎| 102945/111051 [00:13<00:01, 7576.80it/s]\u001b[A\n",
            " 93%|█████████▎| 103703/111051 [00:13<00:00, 7448.57it/s]\u001b[A\n",
            " 94%|█████████▍| 104451/111051 [00:13<00:00, 7455.34it/s]\u001b[A\n",
            " 95%|█████████▍| 105220/111051 [00:14<00:00, 7522.07it/s]\u001b[A\n",
            " 95%|█████████▌| 105985/111051 [00:14<00:00, 7559.42it/s]\u001b[A\n",
            " 96%|█████████▌| 106742/111051 [00:14<00:00, 7535.54it/s]\u001b[A\n",
            " 97%|█████████▋| 107520/111051 [00:14<00:00, 7605.58it/s]\u001b[A\n",
            " 98%|█████████▊| 108305/111051 [00:14<00:00, 7676.99it/s]\u001b[A\n",
            " 98%|█████████▊| 109083/111051 [00:14<00:00, 7707.40it/s]\u001b[A\n",
            " 99%|█████████▉| 109855/111051 [00:14<00:00, 7679.54it/s]\u001b[A\n",
            "100%|█████████▉| 110624/111051 [00:14<00:00, 7678.10it/s]\u001b[A"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "jvUxWGGqhoZU",
        "outputId": "ae5e9fd8-4c27-4992-9e3a-2aca38795736",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        }
      },
      "cell_type": "code",
      "source": [
        "print('Unique tokens in TEXT vocabulary:', len(TEXT.vocab))\n",
        "print('Unique tokens in LABEL vocabulary:',len(LABEL.vocab))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Unique tokens in TEXT vocabulary: 25002\n",
            "Unique tokens in LABEL vocabulary: 2\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "Z7_ZcNtLhoZf",
        "outputId": "8bb6e2d3-6e07-4c46-904c-3a7b8cdfd1ca",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "cell_type": "code",
      "source": [
        "pretrained_embeddings = TEXT.vocab.vectors\n",
        "print(pretrained_embeddings.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([25002, 300])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "2oS3-FfLhoZm",
        "outputId": "336d1727-dbd2-4b1a-ed29-dc8970b956c8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 55
        }
      },
      "cell_type": "code",
      "source": [
        "print(TEXT.vocab.freqs.most_common(20))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[('the', 232653), (',', 220816), ('.', 189387), ('and', 125574), ('a', 125107), ('of', 115377), ('to', 107754), ('is', 87429), ('in', 70454), ('I', 62331), ('it', 61420), ('that', 56650), ('\"', 50472), (\"'s\", 49694), ('this', 48616), ('-', 42468), ('/><br', 41014), ('was', 40184), ('as', 34867), ('with', 34370)]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "sHHDfHeohoZw",
        "outputId": "5db13728-b798-4991-ede5-92e2efc76329",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "cell_type": "code",
      "source": [
        "print(TEXT.vocab.itos[:10])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "['<unk>', '<pad>', 'the', ',', '.', 'and', 'a', 'of', 'to', 'is']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "SPz9MEb9hoZ4",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "BATCH_SIZE = 64\n",
        "\n",
        "train_iterator, val_iterator, test_iterator = data.BucketIterator.splits(\n",
        "    (train_data, val_data, test_data),\n",
        "    batch_size=BATCH_SIZE,\n",
        "    device=device)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "JC5TAxophoaM",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "class LSTM_Model(nn.Module):\n",
        "    \n",
        "    def __init__(self, vocab_size, embedding_dim, hidden_dim, batch_size):\n",
        "      \n",
        "      super(LSTM_Model, self).__init__()\n",
        "      self.num_layers = 1\n",
        "      self.batch_size = batch_size\n",
        "      self.hidden_dim = hidden_dim\n",
        "      \n",
        "      self.word_embeddings = nn.Embedding(vocab_size, embedding_dim) \n",
        "      # The LSTM takes word embeddings as inputs, and outputs hidden states\n",
        "      # with dimensionality hidden_dim.\n",
        "      self.lstm = nn.LSTM(embedding_dim, hidden_dim, num_layers=self.num_layers) \n",
        "      self.fc = nn.Linear(hidden_dim, 1)\n",
        "      self.hidden = self.init_hidden()      \n",
        "      \n",
        "    def forward(self, sentence):\n",
        "      \n",
        "        embeds = self.word_embeddings(sentence)\n",
        "        # [sent_len, batch_size] --> [sent_len, batch_size, emb_dim]\n",
        "        lstm_out, self.hidden = self.lstm(embeds, self.hidden) \n",
        "        # [sent_len, batch_size, emb_dim] --> [seq_len, batch, num_directions*hidden_size]\n",
        "        preds = self.fc(lstm_out[-1].squeeze(0))\n",
        "        # [batch, num_directions*hidden_size] --> [batch_size, 1]\n",
        "        return preds\n",
        "      \n",
        "      \n",
        "    def init_hidden(self):\n",
        "        # Before we've done anything, we dont have any hidden state.\n",
        "        # The axes semantics are (num_layers, minibatch_size, hidden_dim)\n",
        "        return (torch.zeros(self.num_layers, self.batch_size, self.hidden_dim).to(device),\n",
        "                torch.zeros(self.num_layers, self.batch_size, self.hidden_dim).to(device))\n",
        "        "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "QzeSDCk8hoaV",
        "outputId": "94f77e38-1088-4ca6-f42c-83021b5993a7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 106
        }
      },
      "cell_type": "code",
      "source": [
        "model = LSTM_Model(vocab_size=len(TEXT.vocab), embedding_dim=300, hidden_dim=128, batch_size=BATCH_SIZE)\n",
        "model.to(device)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "LSTM_Model(\n",
              "  (word_embeddings): Embedding(25002, 300)\n",
              "  (lstm): LSTM(300, 128)\n",
              "  (fc): Linear(in_features=128, out_features=1, bias=True)\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 27
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "xbaggwufhoag",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model.word_embeddings.weight.data.copy_(pretrained_embeddings)\n",
        "model.word_embeddings.weight.requires_grad = False"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "7qxemdk0hoa1",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "parameters = filter(lambda p: p.requires_grad, model.parameters())\n",
        "\n",
        "optimizer = optim.Adam(parameters, lr=1e-3, amsgrad=True)\n",
        "criterion = nn.BCEWithLogitsLoss()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "Rs6KVJtvhoa_",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Create training and validation dataloaders\n",
        "dataloaders_dict = {'train': train_iterator, \n",
        "                    'val': val_iterator}"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "ngV_IdqthobO",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def train_model(model, dataloaders, criterion, optimizer, num_epochs, batch_size=BATCH_SIZE):\n",
        "    since = time.time()\n",
        "\n",
        "    history = dict()\n",
        "\n",
        "    best_model_wts = copy.deepcopy(model.state_dict())\n",
        "    best_acc = 0.0\n",
        "    skip_count = 0\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        print('Epoch {}/{}'.format(epoch, num_epochs - 1))\n",
        "        print('-' * 10)\n",
        "\n",
        "        # Each epoch has a training and validation phase\n",
        "        for phase in ['train', 'val']:\n",
        "            if phase == 'train':\n",
        "                model.train()  # Set model to training mode\n",
        "            else:\n",
        "                model.eval()   # Set model to evaluate mode\n",
        "\n",
        "            running_loss = 0.0\n",
        "            running_corrects = 0\n",
        "            \n",
        "\n",
        "            # Iterate over data.\n",
        "            for data in dataloaders[phase]:\n",
        "                inputs, labels = data.text, data.label\n",
        "                inputs = inputs.to(device)\n",
        "                labels = labels.to(device)\n",
        "                # zero the parameter gradients\n",
        "                optimizer.zero_grad()\n",
        "                \n",
        "                # forward\n",
        "                # track history if only in train\n",
        "                with torch.set_grad_enabled(phase == 'train'):\n",
        "                    # Get model outputs and calculate loss\n",
        "\n",
        "                    # backward + optimize only if in training phase\n",
        "                    if phase == 'train':\n",
        "                        # we need to clear out the hidden state of the LSTM,\n",
        "                        # detaching it from its history on the last instance.\n",
        "                        model.batch_size = inputs.shape[1]\n",
        "                        model.hidden = model.init_hidden()\n",
        "                        \n",
        "                        outputs = model(inputs).squeeze(1)\n",
        "                        loss = criterion(outputs, labels)\n",
        "                        loss.backward()\n",
        "                        optimizer.step()\n",
        "                        \n",
        "                    else:\n",
        "                        model.batch_size = inputs.shape[1]\n",
        "                        model.hidden = model.init_hidden()\n",
        "                        \n",
        "                        outputs = model(inputs).squeeze(1)\n",
        "                        loss = criterion(outputs, labels)\n",
        "\n",
        "\n",
        "                # statistics\n",
        "                running_loss += loss.item()\n",
        "                outputs = torch.round(torch.sigmoid(outputs))\n",
        "                corrects = (outputs == labels).float()\n",
        "                acc = corrects.sum()/len(corrects)\n",
        "                running_corrects += acc.item()\n",
        "\n",
        "            epoch_loss = running_loss / len(dataloaders[phase])\n",
        "            epoch_acc = running_corrects / len(dataloaders[phase])\n",
        "\n",
        "            print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))\n",
        "\n",
        "            # deep copy the model\n",
        "            if phase == 'val' and epoch_acc > best_acc:\n",
        "                best_acc = epoch_acc\n",
        "                best_model_wts = copy.deepcopy(model.state_dict())\n",
        "            \n",
        "            if phase+'_acc' in history:\n",
        "                # append the new number to the existing array at this slot\n",
        "                history[phase+'_acc'].append(epoch_acc)\n",
        "            else:\n",
        "                # create a new array in this slot\n",
        "                history[phase+'_acc'] = [epoch_acc]\n",
        "            \n",
        "            if phase+'_loss' in history:\n",
        "                # append the new number to the existing array at this slot\n",
        "                history[phase+'_loss'].append(epoch_loss)\n",
        "            else:\n",
        "                # create a new array in this slot\n",
        "                history[phase+'_loss'] = [epoch_loss]            \n",
        "\n",
        "    time_elapsed = time.time() - since\n",
        "    print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))\n",
        "    print('Best val Acc: {:4f}'.format(best_acc))\n",
        "\n",
        "    # load best model weights\n",
        "    model.load_state_dict(best_model_wts)\n",
        "    return model, history\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "I1tZhBPOhobY",
        "outputId": "4210ac80-9980-4ea3-9749-f6f7c28ff2b1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 785
        }
      },
      "cell_type": "code",
      "source": [
        "model, history = train_model(model, dataloaders_dict, criterion, optimizer, num_epochs=10)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 0/9\n",
            "----------\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "100%|█████████▉| 110624/111051 [00:28<00:00, 7678.10it/s]\u001b[A"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "train Loss: 0.6934 Acc: 0.5043\n",
            "val Loss: 0.6936 Acc: 0.4915\n",
            "Epoch 1/9\n",
            "----------\n",
            "train Loss: 0.6932 Acc: 0.5009\n",
            "val Loss: 0.6925 Acc: 0.4937\n",
            "Epoch 2/9\n",
            "----------\n",
            "train Loss: 0.6930 Acc: 0.5040\n",
            "val Loss: 0.6939 Acc: 0.5117\n",
            "Epoch 3/9\n",
            "----------\n",
            "train Loss: 0.6928 Acc: 0.5056\n",
            "val Loss: 0.6930 Acc: 0.4864\n",
            "Epoch 4/9\n",
            "----------\n",
            "train Loss: 0.6921 Acc: 0.5003\n",
            "val Loss: 0.8428 Acc: 0.4976\n",
            "Epoch 5/9\n",
            "----------\n",
            "train Loss: 0.6921 Acc: 0.5005\n",
            "val Loss: 0.7026 Acc: 0.4840\n",
            "Epoch 6/9\n",
            "----------\n",
            "train Loss: 0.6917 Acc: 0.5059\n",
            "val Loss: 0.7041 Acc: 0.4931\n",
            "Epoch 7/9\n",
            "----------\n",
            "train Loss: 0.6911 Acc: 0.5049\n",
            "val Loss: 0.7238 Acc: 0.5107\n",
            "Epoch 8/9\n",
            "----------\n",
            "train Loss: 0.6905 Acc: 0.5016\n",
            "val Loss: 0.7196 Acc: 0.5020\n",
            "Epoch 9/9\n",
            "----------\n",
            "train Loss: 0.6896 Acc: 0.5047\n",
            "val Loss: 0.7418 Acc: 0.5030\n",
            "Training complete in 9m 41s\n",
            "Best val Acc: 0.511669\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "ZHzjnrzdhobg",
        "outputId": "cf122631-ea98-4b98-b72e-94ff7efda4cb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 571
        }
      },
      "cell_type": "code",
      "source": [
        "# summarize history for accuracy\n",
        "plt.plot(history['train_acc'])\n",
        "plt.plot(history['val_acc'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()\n",
        "# summarize history for loss\n",
        "plt.plot(history['train_loss'])\n",
        "plt.plot(history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEVCAYAAAARjMm4AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEVCAYAAAAPRfkLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "94DnDDAln6TJ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "1thIJ65BASNb",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Custom Embedding Layer using LSTM"
      ]
    },
    {
      "metadata": {
        "id": "XAC0kTn2513w",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "TEXT.build_vocab(train_data, max_size=25000)\n",
        "LABEL.build_vocab(train_data)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "wfxqqDe08R8U",
        "colab_type": "code",
        "outputId": "3efc97b9-2d5d-4731-9052-ff7228e85fab",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        }
      },
      "cell_type": "code",
      "source": [
        "print('Unique tokens in TEXT vocabulary:', len(TEXT.vocab))\n",
        "print('Unique tokens in LABEL vocabulary:',len(LABEL.vocab))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Unique tokens in TEXT vocabulary: 25002\n",
            "Unique tokens in LABEL vocabulary: 2\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "RQNNaLqc8ffw",
        "colab_type": "code",
        "outputId": "60571dfd-94e1-4717-dff5-84ef23e99034",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 55
        }
      },
      "cell_type": "code",
      "source": [
        "print(TEXT.vocab.freqs.most_common(20))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[('the', 232653), (',', 220816), ('.', 189387), ('and', 125574), ('a', 125107), ('of', 115377), ('to', 107754), ('is', 87429), ('in', 70454), ('I', 62331), ('it', 61420), ('that', 56650), ('\"', 50472), (\"'s\", 49694), ('this', 48616), ('-', 42468), ('/><br', 41014), ('was', 40184), ('as', 34867), ('with', 34370)]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "PatGqW_s8kk1",
        "colab_type": "code",
        "outputId": "e22c9f8d-6ef9-4b53-881c-b8c04dbeec52",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "cell_type": "code",
      "source": [
        "print(TEXT.vocab.itos[:10])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "['<unk>', '<pad>', 'the', ',', '.', 'and', 'a', 'of', 'to', 'is']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "9fwOT2ff8yk9",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "BATCH_SIZE = 64\n",
        "\n",
        "train_iterator, val_iterator, test_iterator = data.BucketIterator.splits(\n",
        "    (train_data, val_data, test_data),\n",
        "    batch_size=BATCH_SIZE,\n",
        "    device=device)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ggocjCdV8-Qk",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "class LSTM_Model(nn.Module):\n",
        "    \n",
        "    def __init__(self, vocab_size, embedding_dim, hidden_dim, batch_size):\n",
        "      \n",
        "      super(LSTM_Model, self).__init__()\n",
        "      self.num_layers = 1\n",
        "      self.batch_size = batch_size\n",
        "      self.hidden_dim = hidden_dim\n",
        "      \n",
        "      self.word_embeddings = nn.Embedding(vocab_size, embedding_dim) \n",
        "      # The LSTM takes word embeddings as inputs, and outputs hidden states\n",
        "      # with dimensionality hidden_dim.\n",
        "      self.lstm = nn.LSTM(embedding_dim, hidden_dim, num_layers=self.num_layers) \n",
        "      self.fc = nn.Linear(hidden_dim, 1)\n",
        "      self.hidden = self.init_hidden()      \n",
        "      \n",
        "    def forward(self, sentence):\n",
        "      \n",
        "        embeds = self.word_embeddings(sentence)\n",
        "        # [sent_len, batch_size] --> [sent_len, batch_size, emb_dim]\n",
        "        lstm_out, self.hidden = self.lstm(embeds, self.hidden) \n",
        "        # [sent_len, batch_size, emb_dim] --> [seq_len, batch, num_directions*hidden_size]\n",
        "        (hidden, cell) =  self.hidden\n",
        "        preds = self.fc(lstm_out[-1].squeeze(0))\n",
        "        # [batch, num_directions*hidden_size] --> [batch_size, 1]\n",
        "        return preds\n",
        "      \n",
        "      \n",
        "    def init_hidden(self):\n",
        "        # Before we've done anything, we dont have any hidden state.\n",
        "        # The axes semantics are (num_layers, minibatch_size, hidden_dim)\n",
        "        return (torch.zeros(self.num_layers, self.batch_size, self.hidden_dim).to(device),\n",
        "                torch.zeros(self.num_layers, self.batch_size, self.hidden_dim).to(device))\n",
        "        "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "H8ND7h-y9Neb",
        "colab_type": "code",
        "outputId": "788ac888-bfc6-44b4-83d2-94acc8457a61",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 106
        }
      },
      "cell_type": "code",
      "source": [
        "model = LSTM_Model(vocab_size=len(TEXT.vocab), embedding_dim=300, hidden_dim=128, batch_size=BATCH_SIZE)\n",
        "model.to(device)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "LSTM_Model(\n",
              "  (word_embeddings): Embedding(25002, 300)\n",
              "  (lstm): LSTM(300, 128)\n",
              "  (fc): Linear(in_features=128, out_features=1, bias=True)\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 40
        }
      ]
    },
    {
      "metadata": {
        "id": "s6lZpCaw9Nk7",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "optimizer = optim.Adam(model.parameters(), lr=1e-3, amsgrad=True)\n",
        "criterion = nn.BCEWithLogitsLoss()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "eJcncDeCF8CA",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Create training and validation dataloaders\n",
        "dataloaders_dict = {'train': train_iterator, \n",
        "                    'val': val_iterator}"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "babVkHilG5XR",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def train_model(model, dataloaders, criterion, optimizer, num_epochs, batch_size=BATCH_SIZE):\n",
        "    since = time.time()\n",
        "\n",
        "    history = dict()\n",
        "\n",
        "    best_model_wts = copy.deepcopy(model.state_dict())\n",
        "    best_acc = 0.0\n",
        "    skip_count = 0\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        print('Epoch {}/{}'.format(epoch, num_epochs - 1))\n",
        "        print('-' * 10)\n",
        "\n",
        "        # Each epoch has a training and validation phase\n",
        "        for phase in ['train', 'val']:\n",
        "            if phase == 'train':\n",
        "                model.train()  # Set model to training mode\n",
        "            else:\n",
        "                model.eval()   # Set model to evaluate mode\n",
        "\n",
        "            running_loss = 0.0\n",
        "            running_corrects = 0\n",
        "            \n",
        "\n",
        "            # Iterate over data.\n",
        "            for data in dataloaders[phase]:\n",
        "                inputs, labels = data.text, data.label\n",
        "                inputs = inputs.to(device)\n",
        "                labels = labels.to(device)\n",
        "                # zero the parameter gradients\n",
        "                optimizer.zero_grad()\n",
        "\n",
        "                # forward\n",
        "                # track history if only in train\n",
        "                with torch.set_grad_enabled(phase == 'train'):\n",
        "                    # Get model outputs and calculate loss\n",
        "\n",
        "                    # backward + optimize only if in training phase\n",
        "                    if phase == 'train':\n",
        "                        # we need to clear out the hidden state of the LSTM,\n",
        "                        # detaching it from its history on the last instance.\n",
        "                        model.batch_size = inputs.shape[1]\n",
        "                        model.hidden = model.init_hidden()\n",
        "                        \n",
        "                        outputs = model(inputs).squeeze(1)\n",
        "                        loss = criterion(outputs, labels)\n",
        "                        loss.backward()\n",
        "                        optimizer.step()\n",
        "                        \n",
        "                    else:\n",
        "                        model.batch_size = inputs.shape[1]\n",
        "                        model.hidden = model.init_hidden()\n",
        "                        \n",
        "                        outputs = model(inputs).squeeze(1)\n",
        "                        loss = criterion(outputs, labels)\n",
        "\n",
        "\n",
        "                # statistics\n",
        "                running_loss += loss.item()\n",
        "                outputs = torch.round(torch.sigmoid(outputs))\n",
        "                corrects = (outputs == labels).float()\n",
        "                acc = corrects.sum()/len(corrects)\n",
        "                running_corrects += acc.item()\n",
        "\n",
        "            epoch_loss = running_loss / len(dataloaders[phase])\n",
        "            epoch_acc = running_corrects / len(dataloaders[phase])\n",
        "\n",
        "            print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))\n",
        "\n",
        "            # deep copy the model\n",
        "            if phase == 'val' and epoch_acc > best_acc:\n",
        "                best_acc = epoch_acc\n",
        "                best_model_wts = copy.deepcopy(model.state_dict())\n",
        "            \n",
        "            if phase+'_acc' in history:\n",
        "                # append the new number to the existing array at this slot\n",
        "                history[phase+'_acc'].append(epoch_acc)\n",
        "            else:\n",
        "                # create a new array in this slot\n",
        "                history[phase+'_acc'] = [epoch_acc]\n",
        "            \n",
        "            if phase+'_loss' in history:\n",
        "                # append the new number to the existing array at this slot\n",
        "                history[phase+'_loss'].append(epoch_loss)\n",
        "            else:\n",
        "                # create a new array in this slot\n",
        "                history[phase+'_loss'] = [epoch_loss]            \n",
        "\n",
        "    time_elapsed = time.time() - since\n",
        "    print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))\n",
        "    print('Best val Acc: {:4f}'.format(best_acc))\n",
        "\n",
        "    # load best model weights\n",
        "    model.load_state_dict(best_model_wts)\n",
        "    return model, history\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "fsZ4FAgwH_OH",
        "colab_type": "code",
        "outputId": "bf8ce321-14d2-4cb9-88b6-ea3c5117b991",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 767
        }
      },
      "cell_type": "code",
      "source": [
        "model, history = train_model(model, dataloaders_dict, criterion, optimizer, num_epochs=10)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 0/9\n",
            "----------\n",
            "train Loss: 0.6944 Acc: 0.5022\n",
            "val Loss: 0.6921 Acc: 0.5063\n",
            "Epoch 1/9\n",
            "----------\n",
            "train Loss: 0.6928 Acc: 0.4967\n",
            "val Loss: 0.6925 Acc: 0.5192\n",
            "Epoch 2/9\n",
            "----------\n",
            "train Loss: 0.6912 Acc: 0.5061\n",
            "val Loss: 0.6949 Acc: 0.5299\n",
            "Epoch 3/9\n",
            "----------\n",
            "train Loss: 0.6896 Acc: 0.5057\n",
            "val Loss: 0.6969 Acc: 0.5113\n",
            "Epoch 4/9\n",
            "----------\n",
            "train Loss: 0.6882 Acc: 0.5063\n",
            "val Loss: 0.6963 Acc: 0.5548\n",
            "Epoch 5/9\n",
            "----------\n",
            "train Loss: 0.6872 Acc: 0.5032\n",
            "val Loss: 0.6986 Acc: 0.5318\n",
            "Epoch 6/9\n",
            "----------\n",
            "train Loss: 0.6806 Acc: 0.5523\n",
            "val Loss: 0.7395 Acc: 0.5659\n",
            "Epoch 7/9\n",
            "----------\n",
            "train Loss: 0.6429 Acc: 0.6541\n",
            "val Loss: 0.5850 Acc: 0.7312\n",
            "Epoch 8/9\n",
            "----------\n",
            "train Loss: 0.5076 Acc: 0.7909\n",
            "val Loss: 0.5730 Acc: 0.6948\n",
            "Epoch 9/9\n",
            "----------\n",
            "train Loss: 0.3957 Acc: 0.8481\n",
            "val Loss: 0.4420 Acc: 0.8131\n",
            "Training complete in 14m 20s\n",
            "Best val Acc: 0.813093\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Vc9qVM8OID3_",
        "colab_type": "code",
        "outputId": "3759b244-95fe-4af1-931a-fdc55b436553",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 571
        }
      },
      "cell_type": "code",
      "source": [
        "# summarize history for accuracy\n",
        "plt.plot(history['train_acc'])\n",
        "plt.plot(history['val_acc'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()\n",
        "# summarize history for loss\n",
        "plt.plot(history['train_loss'])\n",
        "plt.plot(history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEVCAYAAAAPRfkLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzs3Xd8VNeZ8PHfjKRRr0hCBRBF4iAE\nmGqQTTMYxzXudmwneZ04u3HWyTp9nd032fWbzWZ346yTbJKNsxs768Q17jY2pmMwvYMQBwRIqPfe\nRjNz3z9mJCShMipXo/J8Px8+mrlzz72PjuV57jnn3nMshmEghBBCAFh9HYAQQojRQ5KCEEKIDpIU\nhBBCdJCkIIQQooMkBSGEEB0kKQghhOggSUEIQCn1P0qpf+pnn0eVUltHKCQhfEKSghBCiA7+vg5A\niIFSSk0H9gHPAo8BFuCLwA+BhcDHWusve/a9H/hH3H/rRcBfaa0vKKUmAa8AacAZoAko8JSZC/wX\nkAi0Al/SWh/uJ6YfAp/3nCcb+LzWukYpFQw8B6wCWoCfaK3/3Mf2PwI5Wut/9hy3471SKhd4HngE\n2AAEA38AJgEBwA+11q94yt0M/Nyz/Zynfp4DDmitn/HsMw/YASRqrR3e1b4Y76SlIMaqWKBEa62A\nk8BrwP8BFgAPK6VmKaWmAf8N3KW1ngNsxP3FCPB3QLnWegbwBPAZAKWUFXgHeFFrPRt4HHhXKdXr\nBZRSagnwdWAZ7iQT6HkP8B3A5jnPBuDXSqmkPrb3Z4rWWmmtLwPPAB9ordOBLwN/UEoFKKVCgZeA\nBz2/Qw7wY9xJ8OFOx7obeFMSguhMkoIYq/yBv3henwIOaa0rtNaVQDGQhPvLdofWOsez3/8AN3i+\n4FcDrwNorXOBXZ595gDxuK/I0Vp/CpQD1/UWiNb6CDBVa12ntXYBe4GZno9vBV717FeA+0u9qI/t\n/fmg0+s7gZ95Xu8BgnC3bq4H8rXWpz2ffR/4FvAhMEsppTzb78adTIXoIN1HYqxyaq2b218DDZ0/\nA/yAOKC6faPWulYpZcHdyogBajuVad8vCggBsq98dxKBu4umR0qpEOBZpdRaz6YY3K0SPOeq6RRD\nQz/b+1PV6fVngP+rlIoDXLi70aw9HNveKda3cbek/oA7gexCiE4kKYjxrBTIbH+jlIrG/eVZgTsJ\nRHbaNw64iHvcoc7T3dSFUurRXs7zTdzdRku01g1KqZ8AyZ7PKnB/SbcfYwruL/betrcntHbRPZ1Q\nKRWAu6X0gNb6Q6VUINCeJLsfOwSI8bRIXsE9FlMLvOFp2QjRQbqPxHi2BVitlGrvynkc2OzpQ9+H\nu/sEpdQsYKVnnzygQCl1n+ezWKXUK55++t7EA2c9CSEFd9dQmOez94AvKqUsSqkE4BjuL+zethcD\n13jOPbNTXN2Fev61D4A/Cdg9590DJCillnk++yHwI8/rrbhbPX+LdB2JHkhSEOOW58r4K7gHis/i\nHkf4qufjnwIpSqlLwH8Cb3nKGMDngK97ynwCbNNaN/Zxqt8Ba5RSGvcdP98G1iulvon7qrwMd7LZ\nCXzXM0jc2/b/BqYrpc57Ynyjl9+tBvh34JhS6hhwAfcA+Qe4u5HuBf6slDqHe/D97z3lnLhbGH7A\np/3XophoLLKeghATi1Lq+0Cs1vr7vo5FjD4ypiDEBOIZlP5r4CZfxyJGJ+k+EmKCUEp9FfcYxL9p\nrS/6Oh4xOkn3kRBCiA7SUhBCCNFhzI8plJfXD7qpEx0dQnV103CGM6ZJfVwhddGV1EdX46E+4uLC\nLT1tn9AtBX9/v/53mkCkPq6QuuhK6qOr8VwfEzopCCGE6EqSghBCiA6SFIQQQnSQpCCEEKKDqXcf\nKaWeBVYABvCk1vpQp8+ewL1SlRM4rLX+pmcWyh/jnscFYIvW+idmxiiEEOIK05KCUmoNkKa1zlRK\npeNetCTT81kE8D0gVWvtUEptVkqt8BR9TWv9XbPiEkII0Tszu4/W4561Ea11NhDtSQbgnuLXDoR5\nVsEKoeviIUIIIXzAzO6jBOBIp/flnm11WusWpdTTuBc1aQZe1VqfU0pdh3sK4k24Fxz/rtb6WF8n\niY4OGdI9w3Fx4YMuOx5JfVwhddGV1EdXvqyPvJI6dh4p4MENswmyDe/X+Eg+0dzx9JynxfD3wGyg\nDtiulLoG2I97MfWNSqlM4EVgfl8HHcpThXFx4ZSX1w+6fF927tzG2rXr+93vl7/8Offf/zmSkpL7\n3ddsZtbHWCN10ZXUR1e+rI+LRXU8+/pxGlscLJoVQ+KkvtZ/6l1vSc3M7qMi3C2Ddkm4V5UCSAcu\nehZatwO7cS9leFZrvRFAa70PiFNKjblHB4uLi9i69WOv9n3yye+MioQghBj99OVqfvbqMZpaHTx2\nW/qgE0JfzGwpbAaeBp5TSi0GirTW7ak1F0hXSgV7Fl9fCnzoWfwjX2v9ilJqHu5Wg9PEGE3xH//x\nb2RnZ7Fq1TJuuukWiouL+MUvfstPf/r/KC8vo7m5mS9/+a+5/vpVfP3rf823v/19duzYRmNjA5cv\n51FYWMDf/u13yMy83te/ihBilDh9sZJfv3UKp8vg8TvnsWxOvCnnMS0paK33KqWOKKX24l4s/QnP\nLae1Wuu3lVI/A3YopRzAXq31bs/SiH9SSj3uie2xocbx+vYcDp0t6/EzPz8LTufA59NbNieeB9al\n9vr5Qw99gbfeep0ZM2Zx+XIuv/3t/1BdXcW1167glltup7CwgB/+8Cmuv35Vl3JlZaU888yv2L9/\nL++++6YkBSEEAEd0Ob979zQWi4Wv3zOfa1JjTTuXqWMKWuunum060emz54Dnuu1fANxgZkwjLT09\nA4Dw8Aiys7N47723sFis1NXVXrXvggULAYiPj6ehoWFE4xRCjE77Tpfwh43ZBPhb+dv7FpCeEm3q\n+cb81Nn9eWBdaq9X9SMxWBQQEADAli2bqKur4ze/+R/q6ur4yle+cNW+fn5Xhk9k8SMhxM7jhfxp\nkyYo0J9vPXANqcmRpp9z3CcFX7BarTidXYdCampqSExMwmq1smvXdtra2nwUnRBiLPj44GVe255D\neEgA33lwIdMmj8wtsDL3kQlSUmag9VkaG690Aa1du469e3fz5JNfIzg4mPj4eF544b99GKUQYjQy\nDIP39lzite05RIXZeOqRxSOWEGAcrNE8lJXX5N7rrqQ+rpC66Erqoyuz6sMwDP6y4wKbDl4mNjKI\n7z60iPio4GE/D/S+8pp0HwkhxCjgMgz+vPkcO48VkhATwnc/t5CYiKARj0OSghBC+JjT5eL5jWfZ\nl1XC1PgwvvPgQiJCbT6JRZKCEEL4kMPp4rn3sjiiy5mZFMG3HriG0KAAn8UjSUEIIXzE3ubkN2+f\n5tTFSuZMi+Ib9y4gONC3X8uSFIQQwgeaWx386o2T6Pwa5s+cxBN3z8MW4Pup3iQpCCHECGtsaePZ\n109wsaiOJSqOr342A3+/0fGEwOiIYoK67747aGoa/NTfQoixp67Rzr+/fIyLRXVcNy+Bx+8cPQkB\npKUghBAjpqquhWdePU5JVRM3LErmkZtmY7X0+LiAz0hSMMGXv/wI//IvPychIYGSkmJ+8IPvEBcX\nT3NzMy0tLXzrW99j7tx5vg5TCDGCymqaeeaVY1TUtnDz8mncv3YWllGWEGACJIW3cj7gWNmpHj/z\ns1pwugb+QPSi+Pnck3p7r5+vXn0Dn376Cffe+wC7d+9i9eobmDUrjdWr13LkyCFeeul/+clPfjbg\n8wohxqaiikaeefUYNQ127lo1gzuumz4qEwLImIIp3ElhNwB79uxi5co17Nq1ja997TH+67/+k9ra\nq6fNFkKMT3kl9fzrS0epabDz4LpUPnv9jCEnhDOVmn85+Cw1rcP/XTLuWwr3pN7e61W9WfOXzJw5\ni8rKckpLS6ivr2f37p3Exsbzwx/+mLNnz/DrX/9i2M8phBh9cgprefb1E7S0OvjizYq1C4e+9K7T\n5eSN8+9R3lw5DBFebdwnBV/JzFzJ73//W1atWkNNTTWzZqUBsGvXDhwOh4+jE0KYLTu3il+9eYo2\nh4u/umMuKzIS+i/khcOlxyltKue6xGuJChz+9RWk+8gka9bcwNatH7N27Xpuvvk2XnvtJb71rSfI\nyJhHZWUlGze+5+sQhRAmOZFTwbN/OYnT5eJv7p43bAnB6XLyYe5W/Cx+3Dx9/bAcsztpKZgkPT2D\nXbsOdLx/6aU3Ol6vXLkGgNtu++yIxyWEMNehs2X8/r0s/KwWvn7vAubNmDRsxz5YcpSK5kpWJWcy\nKdicZTlNTQpKqWeBFYABPKm1PtTpsyeAzwNO4LDW+ptKqQDgj0CKZ/uXtNYXzYxRCCGGy56Txbzw\nUTaBAX588/5rmD01atiO7XQ5+Sh3K/4WPz6TYt5S9qZ1Hyml1gBpWutM4DHgV50+iwC+B6zSWq8E\n5iqlVgAPAzWebT8BfmpWfEIIMZy2HSng+Q+zCQn053sPLRrWhACwv/gwlS3VXJ+8guig4T12Z2aO\nKawH3gHQWmcD0Z5kAGD3/AtTSvkDIUCVp8zbnn22AtebGJ8QQgyLD/fn8dKWc0SE2vi7hxczIzGi\n/0ID0OZy8FHuNgKs/tyUsnZYj92dmd1HCcCRTu/LPdvqtNYtSqmngYtAM/Cq1vqcUirBsx9aa5dS\nylBK2bTW9t5OEh0dgr//4GcWjIsbubVPxwKpjyukLrqS+ugqLi4cwzB4adNZ3th5gdioYP758etI\njgsb9nN9fH4X1a013DZ7PWlTpgz78TsbyYHmjqc1PC2GvwdmA3XAdqXUNX2V6U119eAnlJN1Z7uS\n+rhC6qIrqY+u4uLCKSur49VtOWw5nE98VDDffWghNoxhr6c2ZxtvnP4QmzWAlfHXDdvxe0vyZnYf\nFeFuGbRLAoo9r9OBi1rrCk8rYDewpHMZz6Czpa9WghBC+ILTZfC/m86y5XA+SbGhPPX5xcRGBpty\nrj1FB6i117FmyvVE2MxvrZmZFDYD9wEopRYDRVrr9hSXC6QrpdprcSlw3lPmfs+2O4AdJsYnhBAD\n5nC6+I+Xj/DJiWJSJofzdw8vIios0JRz2Z12NuftINDPxo3T1phyju5M6z7SWu9VSh1RSu0FXMAT\nSqlHgVqt9dtKqZ8BO5RSDmCv1nq3UsoP2KCU2gO0Ao+aFZ8QQgyUw+niv945zbHzFaQmR/LN+68h\nJMi8Xvjdhfups9fzmZR1hNlCTTtPZ6aOKWitn+q26USnz54Dnuu2vxP4kpkxCSHEYO05Wcyx8xVc\nkxbL43dkEGgzb/nMVk8rIcgvkPXTVpt2nu5kmgshhPCCyzDYcjgfP6uFbz+8xNSEAPBJwV4a2hq5\nYeoqQgNCTD1XZ5IUhBDCC1mXqiiubOLa9MnERASZeq4WRwtbLu8k2D+IdVNXmXqu7iQpCCGEF7Yc\nygfgpmVTTT/XzoJPaWxrYv3U1YQEmHNXU28kKQghRD8KKxo5famK2VOjSEkw97bQZkczWy9/Qqh/\nCGunrjT1XD2RpCCEEP3YenjkWgnb8/fQ7GjmxmlrCPY3t5uqJ5IUhBCiD/VNdvaeLiEuKoiFqbGm\nnquprYntl3cTFhDK6inXmXqu3khSEEKIPuw6XkSbw8WNS6ZitQ5tbeX+bMvfTYuzhQ0pawnyN+eB\nuP5IUhBCiF44nC62HS0gyObHygWJpp6roa2RHfm7CbeFsTo509Rz9UWSghBC9OLQ2TJqG+ysviaJ\n4EBz5w/ddvkTWp12bkq5AZufzdRz9UWSghBC9MAwDDYfysdigfVLzJ2uut7ewM78PUTaIliZtMLU\nc/VHkoIQQvTgfEEteSX1LEqLIy7K3GcFtuTtxO5q4zPT12HzCzD1XP2RpCCEED3YMkK3oda21vFJ\n4V6iAiO5LulaU8/lDUkKQgjRTXlNM0fPlZMyOZy0KZGmnmtL3k7aXA5unr6eAOtIrnvWM0kKQgjR\nzbYjBRiGu5VgsZh3G2pNay27i/YTExRNZuJS084zEJIUhBCik+ZWB7tPFhEZZmNZeryp5/o4dwcO\nl4Nbpq/HfxS0EkCSghBCdLHnVDHNrU7WLZ6Cv595X5FVLdXsLTpAbFAMyxOWmHaegZKkIIQQHi6X\nwdbD+QT4W1m7MMnUc23K3Y7DcHLrjA34Wc1dm2EgJCkIIYTHiZwKymtayMxIIDzEvAfIKpqr2Fd8\niPiQWJZOXmjaeQZDkoIQQnhs9qyZsGGpuQ+rbcrdhstwcev00dVKAJPXaFZKPQusAAzgSa31Ic/2\nZOClTrvOBJ4CbMCPgQue7Vu01j8xM0YhhADIK6lH59eQMSOG5Lgw085T1lTBgZIjJITEs2TyNaad\nZ7BMSwpKqTVAmtY6UymVDjwPZAJorQuBtZ79/IGdwHvAfcBrWuvvmhWXEEL0pH3NhA1LzX1YraOV\nMGMDVsvo66wxM6L1wDsAWutsIFopFdHDfo8Cb2qtG0yMRQghelXb0MqB7FISYkKYNzPGtPOUNpZx\nsOQoSaEJLIqfb9p5hsLM7qME4Ein9+WebXXd9vsKcFOn92uUUpuAAOC7WutjfZ0kOjoEf//B98nF\nxZm7tN5YI/VxhdRFV+O5PjYfKcThNLjnhlQmx/d07Xq1wdTHKzl/wcDgoYWfZXK8uU9KD9ZIPi1x\n1WOBSqlM4KzWuj1R7AfKtdYbPZ+9CPSZTqurmwYdUFxcOOXl9YMuP95IfVwhddHVeK6PNoeTjZ9e\nJDTIn/kp0V79noOpj6KGEj69fJipYUnMsM3yeX32ltTM7D4qwt0yaJcEFHfb53Zga/sbrfVZrfVG\nz+t9QJxSanQNzQshxpX9WaXUN7WxZmEygTbzvm4+zN2KgcFtM28ydeqMoTIzKWzGPXCMUmoxUKS1\n7p4alwEn2t8opb6vlHrI83oe7laD08QYhRATmGEYbDmcj5/VwrrFyaadp7ChmGNlJ5kWPoV5k9JN\nO89wMK37SGu9Vyl1RCm1F3ABTyilHgVqtdZve3ZLBMo6FXsZ+JNS6nFPbI+ZFZ8QQmTnVVNQ3sjy\nuZOJiQgy7TwbL20B4PZR3koAk8cUtNZPddt0otvn87u9LwBuMDMmIYRod+VhNfNuQ71cX8CJ8tPM\niJjG3Bhl2nmGy+i7SVYIIUZASVUTJy9Ukpocycwk7+44GowPPa2E0T6W0E6SghBiQmpfWW2DiSur\n5dXlc6oim1mRM5gTnWbaeYaTJAUhxITT2NLGp6eKmRQRyOLZsaad54OLm4GxMZbQTpKCEGLC+eRE\nEfY2F+uXTMXPas7X4MXaXM5UaWZHzWJ29CxTzmEGSQpCiFGvzdnGcyf/lzfOv0ezo3lIx3K6XGw7\nUkBggB+rr0kcpgivtvHilbGEsWR0rP8mhBB9OFmRxcmKLAAOlx7nntTbWTZ50aC6ZI7ocqrqWlm/\neAohQQHDHSoA56svcrb6PHOi00iNmmHKOcwiLQUhxKh3sOQoAGumXE+Lo5X/PfMqvzz2HEUNJQM+\n1pZD+ViAG01cM2HjJfdYwlhrJYAkBSHEKFdvb+BM1TmmhifzwOw7+eHy7zA/di7nay7y00O/4K2c\nD2hxtHp1rAuFtVwoquOa1Fgmx4SYEu+56hzO11xk7iTFzMgUU85hJkkKQohR7XDpcVyGi2sTFgMw\nKTiGxxc8yuMLHiU6MJJtlz/hxwee4WjZSQzD6PNYHbehmtRKMAzjyh1HM8ZeKwFkTEEIMcodLDmK\n1WK9ai3j+bFzUdFpbM7bzpa8nfzh9J+ZE53GA+ouJofEXXWcqroWDp8tZ0pcGHNSok2J9WzVeS7U\n5jI/di4pEeYu1mMWaSkIIUatksZSLtcXkB4zmwjb1VM92/wCuH3mZ/iH5d8mPWY2Z6vP8y8H/oP3\nL2zC7rR32XfbkQJchsGGZVNMeWbAMAw+aB9LmLFh2I8/UiQpCCFGrQOeAeb2rqPexIfE8cQ1j/GV\neV8gzBbGprzt/PjAzzlZ7r5jqdXuZNfxIiJCAlgxd7IpsWZVniW37jIL4+YxNdy8GVfNJt1HQohR\nyWW4OFRyjCC/QBbEZvS7v8ViYVH8fNJjZrMpdxvb8j/huVP/y7xJ6UyxX0tTq4PPXj+dgCGs1Ngb\nwzA67ji6dQy3EkCSghBilMqpuUR1aw0rEpdi8/P+eYIg/0DuSr2V5YlLeE2/zenKbE67NLYpM1m5\ncLkpsZ6qOMPl+kIWxy8gOcy8B+JGgnQfCSFGpfZnE5b303XUm8TQyTy56Kusi70dwxGAX1IOvz79\na7Iq9XCGictw8cGlzViwjPlWAkhSEEKMQnZnG8fKThEdGEVq1MxBH8disXDpTDgtJ1exNGY5lS3V\n/PbEH/jvUy9S3VIzLLGeLM+isKGYpZMXkhhqznjFSJLuIyHEqHOqIosWZwurp2RitQz+2rWgvIEz\nudXMmRbLlxbexIb6TF479w7Hy09zplJzy4wbWTd1Ff7WwX0VugwXGy9twYKFW2bcOOg4RxNpKQgh\nRp2DXt511J8th7qumTAlPIlvLX6cz6c/gM3PxrsXPuKnB3/BueqcQR3/WNlJihpLuDZhcY/PRoxF\nXiUFpdTYmAhcCDHmdZ7WYijdMXVNdvZllRIfFcw1qVfWTLBarGQmLuUfV3yPVcmZlDaV88tjv+eF\nrJepba3z+vjuVsJWrBYrt0wfH60E8L77KE8p9SLwvNb6orcHV0o9C6wADOBJrfUhz/Zk4KVOu84E\nngL+AvwRSAGcwJcGcj4hxNjXfVqLwdp5rBCH08WNS6dg7eFhtZCAED6n7iYzcSmv6Xc4XHqc0xXZ\n3DbzJtYkX4efte9bVw+XHqe0qYzrEpcRFzJpSLGOJt52H10LlADPK6W2KKUeVkrZ+iqglFoDpGmt\nM4HHgF+1f6a1LtRar9VarwVuBC4D7wEPAzVa65XAT4CfDvQXEkKMbb1NazEQbQ4XO44WEhzoz8oF\nfd8imhIxle8ufYLPqXuwWqy8ef59/u3wr7hQk9trGafLyUeXtuJn8ePm6esHHedo5FVS0FqXaK1/\n7fkS/5rnX7FS6p+VUkG9FFsPvOMpnw1EK6V6Wh37UeBNrXWDp8zbnu1bgeu9/UWEEGNff9NaeOtg\ndim1jXZWX5NIkK3/DhGrxcqq5BX8aMX3yExcRmFDMf9x9Lf86czr1Nsbrtp/d95ByporyExaxqTg\nmEHHORp5PdCslFqtlHoe+Aj4FFgJ1ODu8ulJAlDe6X25Z1t3XwH+0L2M1toFGP21SIQQ44e301r0\nxTAMthzOx2KB9UsGNhtquC2Mz6ffz3eW/A3JYYnsLznM0/t/xicF+3AZLsDdSngz60P8LX7cnLJu\n0HGOVl6NKSilcoBc4PfAV7XWbZ6PspVSd3l5rqs69ZRSmcBZrXVvozv9DnBHR4fgP4TH1uPiBn81\nMh5JfVwhddGV2fXhMlwc3X+CYP8g1s9Zjs1/cNeDpy9UcLm0gesXJJGeGj+oY8TFzWfpzLl8nLOL\n106/z2vn3uZw+VEeW/I5LtcXUtpYwc2pa5k9dWzOhNoXbweabwYsWuvzAEqpRVrrY57PVvVSpoiu\nLYMkoLjbPrfj7ibqXuaEUirAc047faiubvLuN+hBXFw45eX1gy4/3kh9XCF10dVI1Me56gtUNFWx\nInEptdWtgHcL53T3+hb3E8trFiQOOeZl0cuYfe0c3s75gEOlx/iHrf9OoJ+NAKs/qyZfP6b/RnpL\n8t52Hz0K/KDT+6eUUv8KoLXubVWLzcB9AEqpxUCR1rp7DS4DTnQrc7/n9R3ADi/jE0KMcUOd1gKg\nrKaZ4+crmJEYzqzknoYwBy4yMJxHMx7iyUVfZXJoPC3OVm5KXUNUYOSwHH+08balcIPWumPQV2v9\noFJqT18FtNZ7lVJHlFJ7ARfwhFLqUaBWa90+mJwIlHUq9hqwwXPsVtzJSAgxzg3XtBZbD+djABuW\nTh32NRNmR8/i75d9k/M1F8lMXUB1VfOwHn+08DYp2JRStvauHKVUGNDvtIVa66e6bTrR7fP53d47\ngS95GZMQYpwYjmktmlsd7DlZTFSYjaVzBjeW0B8/qx9zYtLw9xu/MwR5+5v9Dveg8mHAD3e3zz+Z\nFZQQYmIZjmktdp8oosXu5LbMFPz9ZAafwfIqKWit/6CU2oI7GRjAtwDvnwcXQoheDMe0Fi6XwdYj\nBdj8raxZOHZXPRsNBpJOw3A/Q1ABzAH2mxKREGJCGY5pLY6dL6eitoXr5iUQFuz9gjziat4+p/BL\n4Cbct4vmALOAZ0yMSwgxQQzHtBbts6HeuHT8PTcw0rye+0hrnQ4c11ovAzYAIeaFJYSYCEoay4Y8\nrUVuSR3nCmqZNzOGpNjQYY5w4vE2KbQ/RRKolLJorY8g8xIJIYZoOAaY21sJNy2TVsJw8PbuI62U\n+hvgE2CLUkoDUeaFJYQY71yGi4MlRwnyC2RBbMagjlFd38rB7DKSYkPJmD6+JqbzFW+TwuNANO4J\n8D4HTEamtRZCDEFOzSWqW2tYkbgUm9/gBod3HCvA6TLYsHTKsD+sNlF5mxSe1Vp/0/P6ZbOCEUJM\nHEOd1sLe5mTnsSLCggPIzOhpAmYxGN4mBadSah2wF+iYoM4zvbUQQgzIcExrsS+rhIbmNm7LTMEW\nMPiZkkVX3g40fwXYAjQBDs+/tj5LCCFEL9qntViWsGhQ01q410wowM9qYd3iga2ZIPrm7RPN43M6\nQCGETwz1rqOs3CqKKhpZkTGZ6PDA4QxtwvP24bX/19N2rfWPhjccIcR4NxzTWmw5VADIbahm8Lbd\n5uz0zw+4AZDWgxBiwNqntViesGRQ5YsrGzl1sZK0KZFMTxieNRPEFd52Hz3d+b1Syg9405SIhBDj\n2lCntdhyWFoJZhrs/LIBQOpwBiKEGP86T2sRbgsbcPmG5jb2niomNjKIRWlxJkQovB1TyMc9ZXa7\nGOCPZgQkhBi/hjrAvOt4IXaHixuXTMFqlYfVzODtcworO702gDqtdY0J8QghxqmhTmvhcLrYfrSQ\nQJsfKxckmRChAO+7j0KBx7XWeVrry8CzSqnBTVYihJiQ2qe1WBg/f1DTWhzWZVTXt7JqfiIhQeN3\nOUxf8zYp/Ab4sNP7P3i2CSFjppbOAAAgAElEQVSEV4YyrYVhGGw5lI8FuHGpPKxmJm/Trb/Wenf7\nG631HqVUvx16SqlngRW4u5ye1Fof6vTZVOAVwAYc1Vo/rpRaC/wFyPLsdkpr/Q0vYxRCjFJDndbi\nRE4ll4rrWZQWS3y0LOViJm+TQq1S6mvATtyti5uB+r4KKKXWAGla60ylVDrwPJDZaZefAz/XWr+t\nlPqNUmqaZ/surfV9A/klhBCjW/u0FqunZA54Wos2h5NXtp3Dz2rhnjWzTIpQtPP2v86XgCXA67iv\n7lM92/qyHngHQGudDUQrpSIAlFJWYBXwnufzJzxjFUKIcWgodx1tOphPeU0L65dMIVlWVjOdtw+v\nlSul/k1rfR5AKbVIa13eT7EE4Ein9+WebXVAHO6WxrNKqcXAbq31Dzz7zVVKvYf7ttentdZb+jpJ\ndHQI/v6DnyExLm5wSwCOV1IfV0hddDXY+qhtqeNM1TlmRk9jwfSBPd5UVt3Exn15RIUH8uU75xMa\nPLh1F8wwXv8+vH1O4SdAIvBlz6anlFKXtNZPDeBclm6vk4FfArnARqXUbcBx4GncLZKZwA6lVKrW\n2k4vqqubBhBCV3Fx4ZSX99kLNqFIfVwhddHVUOpjR/6nuAwXi2MXDvgY//XOaextTr5w02yaGlpo\namgZVAzDbTz8ffSW1LwdU1irte5Yk1lr/aBSak8/ZYpwtwzaJQHFntcVQJ7W+gKAUmobkKG13gi8\n5tnnglKqBHfyuORlnEKIUWaw01pk51Zx+GwZs5IiyJwni+iMFG/HFGxKKVv7G6VUGO6pLvqyGbjP\ns/9ioEhrXQ+gtXYAF5VSaZ59l+BeB/oRpdR3PWUScC/7WejtLyOEGF0GO62Fw+nipa3nsQCP3DQb\nqyy1OWK8bSn8DshWSh3GPUvqMuAXfRXQWu9VSh1RSu0FXMATSqlHgVqt9dvAN4E/egadTwHv435I\n7mWl1J24b1X9Wl9dR0KI0W2wA8zbjxZSVNHImoVJMhPqCPN2oPkPSqnzQCzuZw7eA34APNtPue5j\nDic6fZZD1+kzwD34fIc3MQkhRrfBTmtR22jn3T0XCQ3y557Vg1uqUwyetwPNvwA+g3uMIAeYBTxj\nYlxCiDGufVqLFYlLBzStxZs7L9Dc6uSRDbMJD7H1X0AMK2/HFJZrrdOB41rrZcAGQB4rFEL0ajDT\nWlworGXPqWKmxoexdpFMeucL3iaFVs/PQKWURWt9BLi+rwJCiIlrMNNauAyDl7acA+CRDbPxsw52\nuRcxFN4ONGul1N8AnwBblFIaiDIvLCHEWDaYaS32nCwmt6SeFXMnM3uqfL34irdJ4XEgGqgBPof7\nVtGfmhWUEGJsG+hdR40tbbyx8wKBAX7cf4Ms6uhL3t59ZABVnrcvmxeOEGKsq7c3cKbqHNPCk0kM\nnexVmXc+uURDcxv3r51FdHigyRGKvkinnRBiWB0uPY7LcHFtwhKv9s8va2D7sQImx4SwYdlUk6MT\n/ZGkIIQYVgOZ1sLwDC4bBjx8Yxr+fvKV5GvyX0AIMWwGOq3FgexSzuXXsCgtlvkzJ41AhKI/khSE\nEMNmIAPMLXYHr2/Pwd/PyoPr0/rdX4wMSQpCiGEx0GktPtibR02DnVuWTyM+KngEIhTekKQghBgW\nFzzTWiyMn9/vtBYlVU18fPAykyICuTUzZYQiFN6QpCCEGBYHvJzWwjAMXtl6HqfL4MF1aQQGDH7l\nRDH8JCkIIYZsINNanMip5NTFStJTolmi4kYoQuEtSQpCiCFrn9ZiWcKiPqe1aHM4eWXbOfysFh7e\nMBuLLJ4z6khSEEIMmbd3HW06cJnymhbWL5lCcmzoSIQmBkiSghBiSLyd1qKytoWN+/KICLVx58oZ\nIxihGAhJCkKIIfF2WovXduRgd7i4f+0sggO9nYtTjDRJCkKIIfFmWoszuVUcPlvGrOQIMucljGB0\nYqBMTddKqWeBFbjXdX5Sa32o02dTgVcAG3BUa/14f2WEEKNL+7QWGZPm9DqthcPp4uWt57EAn9+g\nsMrg8qhmWktBKbUGSNNaZwKPAb/qtsvPgZ9rra8FnEqpaV6UEUKMIt4MMG8/WkhRRSOrFyaRkhA+\nUqGJQTKz+2g98A6A1jobiFZKRQAopazAKuA9z+dPaK0v91VGCDG6eDOtRW2jnXf3XCQ0yJ97Vnu3\nLKfwLTO7jxKAI53el3u21QFxQD3wrFJqMbBba/2Dfsr0KDo6BH//wT8RGRcnVy6dSX1cIXXRVff6\nOFN2jurWGtbOyCQ5IabHMi9tO0pzq5PH71nAzJTxNQvqeP37GMlbACzdXicDvwRygY1Kqdv6KdOj\n6uqmQQcUFxdOeXn9oMuPN1IfV0hddNVTfWw+uweAa6Lm91hXFwpr2XYon6nxYSxNnTSu6nM8/H30\nltTMTApFuK/y2yUBxZ7XFUCe1voCgFJqG5DRTxkhxChhd7ZxtI9pLVwugz9vOQfAIxtmY7XK4PJY\nYeaYwmbgPgBPF1GR1roeQGvtAC4qpdonUV8C6L7KCCFGj/6mtdh9soi8knpWZExm9tQoH0QoBsu0\nloLWeq9S6ohSai/gAp5QSj0K1Gqt3wa+CfzRM+h8Cnhfa+3qXsas+IQQg3ewjxlRG1vaeHPXRQJt\nfty/NnWkQxNDZOqYgtb6qW6bTnT6LAdY6UUZIcQo0nlai4QeprV455NLNDS3cf/aWUSHB/ogQjEU\n8kSzEGJA+prW4nJpPduPFTA5JoQNy6b6IDoxVJIUhBAD0tu0FoZh8PKWcxgGPHJjGv5+8vUyFsl/\nNSGE19qntUiPmX3VtBYHsks5V1DLorRY5s0cX88kTCSSFIQQXuttWosWu4PXt+fg72flwfVpPRUV\nY4QkBSGEV/qa1uL9vbnUNNi5dcU04qOCfRShGA6SFIQQXrlQc4nq1hoWxs/H5hfQsb2kqonNB/OZ\nFBHILStSfBihGA6SFIQQXunp2QTDMHh56zmcLoMH16URGDD4ecjE6CBJQQjRL7vD3uO0FsdzKjh9\nsYr0lGiWqDgfRiiGiyQFIUS/DhedvGpaizaHk1e3ncfPauHhDbOxyOI544IkBSFEvz7JPQB07Tra\ndOAy5TUtrF8yheTYUF+FJoaZJAUhRJ/q7Q0cLznTZVqLitpmNu7LIyLUxp0rZ/g4QjGcJCkIIfrU\n07QWr2/Pwe5wcf/aWQQHjuSyLMJskhSEEL0qqC9iV8GnXaa1OJNbxWFdzqzkCDLnJfRzBDHWSIoX\nQlyltrWeDy5+zL7iQxgY3Jy2lnBbGA6ni5e2nMMCfH6DwiqDy+OOJAUhRAe7s40d+bv5OG87rU47\niaGTuSf1dtbMWUp5eT3bjxRQXNnEmoVJpCSMzzWKJzpJCkIIDMPgSNkJ3r3wEVUt1YQFhHJ36m1c\nl3gtflb3A2m1jXbe/fQSoUH+3LP66iU4xfggSUGICe5S7WXePP8+l+ry8Lf4ceO0Ndw8fR3B/l3n\nMHpjZw7NrU4+f9NswkNsPopWmE2SghATVFVLNe9e+IjDpccBWBg3n7tm3UpcyNXTXp/NreLTUyVM\njQ9j7cLkkQ5VjCBJCkJ0c776IluLc5kePINZkdPH3ZO6LY5WtlzeybbLu2hzOZgWnsw9qXeQFt1z\nl5DLZfDc2ycBeGTDbKzW8VUfoitTk4JS6llgBWAAT2qtD3X6LBfIB5yeTY8AacBfgCzPtlNa62+Y\nGaMQ7ZwuJxsvbWFz3g4MDADigiexInEZyxMWEx0U5eMIh8ZluDhQfIT3L26i1l5PpC2Cz866mWsT\nFndMXdGT3SeLyCmoZUXGZGZPHdt1IPpnWlJQSq0B0rTWmUqpdOB5ILPbbrdorRs6lUkDdmmt7zMr\nLiF6UtFcyQtZr5Bbd5lJQTF8bsEdHMw7xfHyk7x/cRMfXPyY9JjZZCYtY37sXAKsY6uRfa76Am+e\nf5+ChiICrAHcOv1GbkxZS6Bf32MDR3Q5r27LITjQj/vXpo5QtMKXzPzLXg+8A6C1zlZKRSulIrTW\ndSaeU4gBO1RyjFf1W7Q4W1k2eREPqruZlhjH3LAMmh13cqT0BPuKD3OmSnOmShPqH8LShEVkJi5j\naniSr8PvU1lTBe/kbOREhbvxfW3CYj478+Z+Wz0uw+D9T3N5d88lbAFWvv3wUqLDA0ciZOFjFsMw\nTDmwUur3wEat9bue97uBx7TW5zzvc4E9wHTPzx8Aa4DfAjlADPC01npLX+dxOJyGv7/M4S4Grrmt\nheePvsau3P0E+QfylSUPsXr68l73L6gtZselvXySe4Da1noAZkRNZe2MTFamLCM8MKzXsiOt0d7E\nm1kf8lHOTpwuJ3NiZ/HFhfeROml6v2WbWx384tWj7D1ZTHxMCP/3S9cyIynS/KDFSOtxcGgk28Dd\nA/gRsAmowt2iuBfYBzwNvA7MBHYopVK11vbeDlpd3TTogOLiwikvrx90+fFmItVHXl0+L2S9THlz\nJSnhU3k04yHiQ2I7fv+e6iKQMG5OvokNies5XXmW/cWHOV2ZzQvHXudPx99kQVwGKxKXkR6T1mcf\nvZmcLie7i/bz4aUtNLY1MSkohrtSb2VR3HwsLku//33La5r5zzdPUVDegJoaxd/cPY+wAPfvMlH+\nNrwxHv5fiYvr+eFDM5NCEdB5YpQkoLj9jdb6xfbXSqkPgfla6zeA1zybLyilSoBk4JKJcYoJxGW4\n2Hb5E96/+DFOw8mGaWu5feZN+A9gjMDP6sc1cRlcE5dBbWs9h0qPsq/oEEfLTnK07CRRgZGsSFjC\n8sSlxIfEmvjbXGEYBlmVZ3krZyOlTWUE+QVy16xbWTvlegI6LZ3Zl7N51fz2ndM0NLdxw+JkHlqf\nhr+fTI820ZiZFDbjvup/Tim1GCjSWtcDKKUicbcG7vC0AtYAbyilHgEStdbPKKUSgMlAoYkxigmk\ntrWOF8+8xtnq80TYwvni3AdJj5k9pGNGBoZz47Q1rJ+6mty6fPYVH+JI6Qk25W1nU952UqNmkJm4\njEXxC/od1B2sooYS3sr5gOyqc1iwsDJpObfP/AzhNu+6swzDYMexQl7Zeh6AL96s5FmECcy0MQUA\npdS/AqsBF/AEsAio1Vq/rZR6Evg/QDNwDPgGEAa8DEQBNtxjCh/2dY7y8vpB/wLjoQk4nMZzfZyu\nyOZP2a/T0NbIvElz+Hz6A31+aQ6lLuxOO8fLT7Ov6BDnai4AEOhnY0n8NWQmLWNGRMqwPPtQb2/g\ng0ub+bTwAAYGc6LTuCftdpLDEr0+RvsEd7uOFxEeEsATd8/v8bbT8fy3MRjjoT7i4sJ7/CM0NSmM\nBEkKw2c81keby8G7Fz5kR/4e/C1+3JV6G2unXN/vl/Jw1UVFcxX7iw+zv/gw1a01AEwOiWNF4lKW\nJywhMjBiwMdscznYmb+HTbnbaXG2MDkkjntSbydj0pwBJZu6Rju/efsU5wtqmRYfxtfvnU9sZHCP\n+47Hv42hGA/1IUmhB+PhP+xwGm/1UdJYxgtZL1PQUMTkkHi+nPEwU7y8hXS468JluDhXfYF9xYc4\nXn4ah8uB1WJlbowiM3Ep82LT+x3XMAyDY+WneCfnQypbqgj1D+HWmRtYlbSiY9I6b+WV1PPrt05S\nWdfKsjnxfPnWdAJtvR9jvP1tDNV4qI/eksLYegJHCC8YhsHe4oO8ce497K42rk+6lnvTPmtan743\nrBYrc2LSmBOTRlNbE4dLT7Cv+BCnK7M5XZlNWEAo1yYsZkXi0h67f/Lq8nnz/AdcqL2E1WJl3dRV\n3DJ9PSEBIQOO5WB2Kc9vzMbucHHP6pncljk83VlifJCkIMaVprZmXtZvcqzsJMH+wTw290EWxy/w\ndVhdhASEsHpKJqunZFLYUMz+4sMcLDnK9vzdbM/fzbTwKWQmLmPp5IW0Olt5/+LHHCg5AsCC2Azu\nTr2V+JC4AZ/XZRi8s/siH+zNI9Dmxzfunc+itIEfR4xv0n00xpuAw2ms18eFmlxeyHqZ6tYaZkXO\n4NGMzxETFD2oY410XThcDk5XZLOv+DBZlWcxMPC3+mPBQpurjeSwRO5Lu4PZ0YObaqK51cF/v3+G\n4zkVxEcF841755Mc5/3DdmP9b2O4jYf6kO4jcZWmtiYKG0ooaiyhrKmc2XUpTA+cOajBT19yupxs\nytvOR5e2AnDrjA3cnLJuwP3svuRv9Wdh/HwWxs+nprWWgyVH2V98hDZXG7dMX8+KxKWDfiCutLqJ\n/3zzFEUVjcydHs3jd84jLNi7ZxfExCNJYQJwuByUNpVT1FBCYUMxRY3unzWttV3221nwKQDTI6ax\nIHYuC+IySAiJH9X9zVUt1fwx6xUu1OYSHRjFoxkPkRo1w9dhDUlUYCQ3pdzATSk3DPlYWblV/O6d\n0zS2ONiwdCoPrJuFn1UeSBO9k6QwjhiGQU1rbceXfnsSKG0qx2k4u+wbaYsgPWY2yWGJJIUmEBcS\nS6WrjL2XjpJTe4ncusu8d3ETccGTWBCXwYLYDGZGpvhs+oaeHCs7xUtn36DZ0cyiuPk8POfeQQ28\njkeGYbDlcAGvbT+Pn9XCl26dw6oFo3vyPjE6SFIYo1ocrRQ3dr3yL2ooocnR3GU/mzWAKeFJJIcm\nkhSW4E4CYQmEBYRedczlcfNYFr2MxrYmTldkc7LiDGeqNNsuf8K2y58QFhDKvNh0FsRmkB6Ths1H\nd/PYnXbeOP8enxYdxGYN4OE593Jd4rWjukUzktocLl78+CyfniohMtTGE/fMJzVZJrQT3pGkMMq5\nDBflTRUUNpZQ1FDsHgNoKKaiparLfhYsxAVPYnZ06pUv/9AEYoNjBnx1HxoQwvLEJSxPXEKbsw1d\nncPJiixOVpzpeBArwBpAesxsFsTOZV5sutdTKgxVQX0RL2S9TElTGclhiXw54xESQuNH5NxjQU1D\nK7956xQXiuqYnhDON+5dIFNeiwGRpDCK1Nnrr/T7N5RQ2FhMSWMpbS5Hl/3CAkKZHZ1KcmgCSWGJ\nJIclkBg62ZQr9wC/AObFpjMvNp3PGS7y6vI5WXGGk+VZnkSRhQULMyNTPN1Mcwd1u2R/DMNgV8Fe\n3s75AIfh5IYpK7lz1i1eT/Y2EVwqruPXb52iur6VFRmTefTmOdgCxs5guxgdJCn4gGEYVLXUkFt3\nmby6fAoaiihqKKG+raHLfv5WfxJD4knydPm4u4ASibCF+aSrxGqxMiMyhRmRKdw56xbKmso5WXGG\nE+VZXKzN40JtLm/nbCQhdLJ7oDo2g5SIKUMeh6i3N/Dn7L90POT1hfQHmBebPky/1fiwL6uEFz48\ni9Pl4oEbUvnMtVOlO00MijynMAL3Gjc7msmrKyC37rLnXz719q4JYFJQtPuqv9PVf1xw7IjeVjmU\n+qi3N3CqIpuTFVmcrTrX0bqJtIUzL3YuC2LnoqJTB3xlf7bqPC+eeZVaez1zotP44twHR+SW2bFy\nH7rLZfDGrgtsOnCZ4EB/vvrZDBbMmjTs5xkr9TFSxkN9yHMKI8TpclLYWExubX5HAihtKuuyT1Rg\nJAvj5jM9YirTI6YyJTyZYP8gH0U8PMJtYVyXtIzrkpZhd9rJrjrPyfIsTldm82nRAT4tOkCgn425\nMYoFcRnMmzSnzzuFHC4HH1zczNbLu7BYLNydehvrpq4aVXc/+VpTSxvPvXeGUxcrmRwTwt/eO5/E\nSVffQCDEQEhSGALDMKhsqSbP8+WfW3eZ/PrCLmMAgX42ZkfNYnrkNKZHTCUlYipRgeP7ThCbn61j\nERqX4eJibR4ny7M4UZHFsfJTHCs/hdViJTVyRsc4xKTgmI7y5U2VvJD1Mnn1+cQFT+JLGQ+TEjHV\nh7/R6FNc2civ3jxFaVUT82dO4qufnUtIkIyviKGT7qMBNAGb2prJq8vvSAC5dZdpaGvs+NxqsZIY\nOpnpEdM8/6aSEBo/Zq5uzW4SG4ZBcWOpe6C6Iou8uvyOz5LDElkQm0GYLZT3LnxEq9PO8oQlPDD7\nToJ80Ioazd0DJy9U8tx7WTS3Orhl+TTuXTMLq9Xc8YPRXB++MB7qQ7qPBsjhclDYUExuXb4nEVym\ntKm8yz7RgVEsil/g6QaaxtTwZJ/OxDnaWSwWksISSApL4Obp66hpre0YhzhXlcNHDe7VWoP8Anl0\n7kMsS1jk44hHF8Mw2HTwMm/suIC/v5W/umMumRkJ/RcUYgAkKdDeDVRFbm17N1A++Q2FODp1AwX5\nBaGiU5keMY0UTxKIDOx54WvhnajASFYlr2BV8gpaHC2cqTpHUUMJKxKXEBs8/IOlY5m9zckfN51l\nf1Yp0eGBfP2e+cxIHFtzVImxYcImhbKmcnZlfUJWcQ55dflXdQMlhyaQEnmlG2hySNyY6QYaqMaW\nNkqrmimrt1Nb24TFYsGC+8reYsH9D/drAKt7Q6f9Ou0LPZS3eMq5d+6yX8ex/ZkbNZeMqAxchkFT\nSxsuwz3ds+EycBnu5O0yPK9dnV4bBi6XgeHZ32V4XrsMT5krZdu3d9mHzvsauFwQEVGNvaWN4EB/\ngmx+nn/+BAW6X4/k/EHV9a3855snyS2pZ1ZSBE/cM5+oMHkgTZhjwiaFV/TbnKvOAdy3g86OntUx\nFjA1PMlnUziYxeUyqKxrobiyiZLKRoqrmjpe1zW1+Tq8Mcfmb72SKGx+BHVPHja/qxOK5/WV7e6f\nAf7WXp8puFBYy6/fOkVto52V8xP5wmcUAf7j8+JEjA6mJgWl1LPACsAAntRaH+r0WS6QD7TP1PaI\n1rqwrzLD6aHZd9Pk30CMJZYI2/jpBmq1OympaqK4qpGSSvcXf3FlE6XVTbQ5XF32tVggLjKY6YkR\nJMSEEBUZTGNjK4YBBu4raTxX3+D5aeC+wqb9tftK2/BceXcu2/666zGA9n27lGs/jrtFYbFYsFot\nXV5bLO5WSvtP97ZO+7R/ZrV02a99W/u+VqsFC12PabWApVO5sLAgKiobabE7aLE7abE7abY7aGl1\nXrWttslOq73rhIMDYbVYCA68OnnYAvw4dbESlwseWp/GjUunyANpwnSmJQWl1BogTWudqZRKB54H\nMrvtdovWumGAZYbFO9vL2J9VSlhwAZGhNiJCbR0/u7wOsREZZiM8JGDUTDlsGAY1DfarrvhLqpqo\nrGu9av/AAD+SYkNJnBRCYkwIiZNCSZgUwuToYAL8rzwcNx7uqBguA60Ll2HQ6kkUHUmjtVtCaf+s\ntYdtdictrU5qGlppsTtxutxJNCw4gK/emUHG9Jh+IhBieJjZUlgPvAOgtc5WSkUrpSK01nXDXGZQ\nMqbH0NjipKKmiZqGVgorGvvc3wKEhQR0SRQRIbarEkpkqI3wENuw3CLocLoorW52f/l7rvhLqtyv\nW3q4Mo0ODyQ9Jdr95e/54k+MCSE6PFCuME3mvtr3JzjQHxhaf79hGDicLprtToJtfl0StxBmMzMp\nJABHOr0v92zr/AX/O6XUdGAP8AMvy3QRHR2C/yD+p7lrXTh3rZvd8b7N4aSm3k5NQwvV9a3U1LdS\nXd9CTcdr98+ahlYKy/tJIBaIDA0kKjyQqLBAoiLcP6PDg9zbwgOJ9vyMCA2kqaWNgtIGCsrqKShr\n8Pyrp6SqCZer62MY/n5WkuJCmRIfxpT4cM/PMJLjwobl4aW4uPHTlTZUUhddSX10NV7rYyQHmrtf\nqv4I2ARU4W4d3OtFmatUVzcNOqCeugiigvyJCvKHuN6nC2hzOKlrbKO20U5do526Jju1Da2eba3U\nNdqpbWqjtKqR3OK+GzkW3H3r3YUFBzArKYLESSEkxHiu+ieFEBsZ1GM3VmN9C431Ld782r2S7qMr\npC66kvroajzUR29JzcykUIT7Kr9dElDc/kZr/WL7a6XUh8D8/sqMFgH+fkyK9GNSZP9P2trbnJ4k\n4U4gHYmk0+uQQP8r3T2TQkiICSE8ZHzd/SSEGBvMTAqbgaeB55RSi4EirXU9gFIqEngduENrbQfW\nAG8Ahb2VGatsAX7ERgUTGxXs61CEEKJfpiUFrfVepdQRpdRewAU8oZR6FKjVWr/taR3sV0o1A8eA\nN7TWRvcyZsUnhBDiajIh3hjvFxxOUh9XSF10JfXR1Xioj94mxBsdN94LIYQYFSQpCCGE6CBJQQgh\nRAdJCkIIITpIUhBCCNFBkoIQQogOY/6WVCGEEMNHWgpCCCE6SFIQQgjRQZKCEEKIDpIUhBBCdJCk\nIIQQooMkBSGEEB0kKQghhOgwkstxjipKqWeBFbhXw3xSa33IxyH5jFLq34FVuP8efqq1fsvHIfmc\nUioYOA38WGv9Rx+H41NKqUeA7wMO4Eda640+DsknlFJhwItANBAIPK21/ti3UQ2/CdlSUEqtAdK0\n1pnAY8CvfBySzyilbgDmeeriZuAXPg5ptPi/uNcPn9CUUpOAfwRWArcDd/o2Ip96FNBa6xuA+4Bf\n+jYcc0zIpACsB94B0FpnA9FKqQjfhuQznwD3e17XAKFKKT8fxuNzSqk5wFxgQl4Rd3MjsFVrXa+1\nLtZa/7WvA/KhCmCS53W05/24M1GTQgJQ3ul9uWfbhKO1dmqtGz1vHwM+1Fo7fRnTKPBz4Nu+DmKU\nmA6EKKXeU0rtVkqt93VAvqK1fhWYppTKwX0x9V0fh2SKiZoUuutxWbqJRCl1J+6k8HVfx+JLSqkv\nAvu01pd8HcsoYcF9dXwP7u6TF5RSE/L/F6XU54HLWutUYB3wax+HZIqJmhSK6NoySAKKfRSLzyml\nPgP8A3CL1rrW1/H42G3AnUqp/cBXgB8qpW70cUy+VArs1Vo7tNYXgHogzscx+cr1wMcAWusTQNJ4\n7GqdqHcfbQaeBp5TSi0GirTWY3sV7kFSSkUCPwNu1FpP+IFVrfWD7a+VUv8E5Gqtt/ouIp/bDPxR\nKfVvuPvRwxinfeleyAGWA28qpVKAhvHY1Tohk4LWeq9S6ohSai/gAp7wdUw+9CAQC7yulGrf9kWt\n9WXfhSRGC611oVLqDdrdq+sAAAItSURBVGC/Z9M3tNYuX8bkQ88BzyulduH+7nzcx/GYQtZTEEII\n0WGijikIIYTogSQFIYQQHSQpCCHE/2/v7kGjCqIwDL+22qQQQRQREU9hQFTURqOFjVjFQLBR0iSI\nIAhpxcrKRi3sBJs0tiKkiyskEPCnEAuPnWAKwUC0iCgWFjMOQVZWRb0B3we22GF2uLe4HGbuzjdq\nLAqSpMaiIElqLApShyJiIiJmur4O6RuLgiSpcZ+C9BMi4hIwTtm09BK4DjwAZoF9tdvZutnrNHAV\nWK2fqdp+hBJN/pkSy30eGKPkCn2gJLO+Bs5kpg+mOuFMQRogIg4Do8BIPXdihRIpvQu4m5nHgB4w\nHREbgTvAWM3dnwWu1aFmgMnMPA48ouQsAewFpoCDwDBw4F/cl9TPfxlzIf2iE8Bu4GGNAtkEbAOW\nM/Np7bMAXAb2AG8z801t7wEXImIzMJSZLwAy8yaUdwrA48xcrd+XgKG/f0tSfxYFabBPwP3MbLHi\nEbETeLamzwbK0a7fL/usbf/RzPxLn99InXD5SBpsAThVz+glIi4CWykn9u2vfY4Cz4FXwJaI2FHb\nTwKLmbkMvIuIQ3WM6TqOtK5YFKQBMvMJcBvoRcQ8ZTnpPbAETETEHCVr/0ZmfqQcVnQvInqUo1+v\n1KHOAbdqyuYI5R2DtK747yPpN9Tlo/nM3N71tUh/kjMFSVLjTEGS1DhTkCQ1FgVJUmNRkCQ1FgVJ\nUmNRkCQ1XwG9V0vZoNwfgAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEVCAYAAAAPRfkLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "Zcd66NyVIXZN",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "text",
        "id": "YomNAD_Ve3dx"
      },
      "cell_type": "markdown",
      "source": [
        "## Custom Embedding Layer using Bidirectional LSTM"
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "9GOPy_Zhe3d0",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "TEXT.build_vocab(train_data, max_size=25000)\n",
        "LABEL.build_vocab(train_data)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "f1496559-c34f-43d1-91c6-fc95dd3368ee",
        "id": "4aGpOT8Ue3eA",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        }
      },
      "cell_type": "code",
      "source": [
        "print('Unique tokens in TEXT vocabulary:', len(TEXT.vocab))\n",
        "print('Unique tokens in LABEL vocabulary:',len(LABEL.vocab))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Unique tokens in TEXT vocabulary: 25002\n",
            "Unique tokens in LABEL vocabulary: 2\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "9d4e88e0-e65f-4222-f44a-813d69d1285a",
        "id": "sLrJEwB6e3eG",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 55
        }
      },
      "cell_type": "code",
      "source": [
        "print(TEXT.vocab.freqs.most_common(20))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[('the', 232653), (',', 220816), ('.', 189387), ('and', 125574), ('a', 125107), ('of', 115377), ('to', 107754), ('is', 87429), ('in', 70454), ('I', 62331), ('it', 61420), ('that', 56650), ('\"', 50472), (\"'s\", 49694), ('this', 48616), ('-', 42468), ('/><br', 41014), ('was', 40184), ('as', 34867), ('with', 34370)]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "4fea0235-ad4a-45fa-e372-04b7f65557a8",
        "id": "ei6EvHXye3eM",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "cell_type": "code",
      "source": [
        "print(TEXT.vocab.itos[:10])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "['<unk>', '<pad>', 'the', ',', '.', 'and', 'a', 'of', 'to', 'is']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "AqKMS3wQe3eU",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "BATCH_SIZE = 64\n",
        "\n",
        "train_iterator, val_iterator, test_iterator = data.BucketIterator.splits(\n",
        "    (train_data, val_data, test_data),\n",
        "    batch_size=BATCH_SIZE,\n",
        "    device=device)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "SvC4OOKge3ee",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "class BiLSTM_Model(nn.Module):\n",
        "    \n",
        "    def __init__(self, vocab_size, embedding_dim, hidden_dim, batch_size):\n",
        "      \n",
        "      super(BiLSTM_Model, self).__init__()\n",
        "      self.num_layers = 1\n",
        "      self.batch_size = batch_size\n",
        "      self.hidden_dim = hidden_dim\n",
        "      \n",
        "      self.word_embeddings = nn.Embedding(vocab_size, embedding_dim) \n",
        "      # The LSTM takes word embeddings as inputs, and outputs hidden states\n",
        "      # with dimensionality hidden_dim.\n",
        "      self.bilstm = nn.LSTM(embedding_dim, hidden_dim, num_layers=self.num_layers, bidirectional=True) \n",
        "      self.fc = nn.Linear(hidden_dim*2, 1)\n",
        "      self.hidden = self.init_hidden()      \n",
        "      \n",
        "    def forward(self, sentence):\n",
        "      \n",
        "        embeds = self.word_embeddings(sentence)\n",
        "        # [sent_len, batch_size] --> [sent_len, batch_size, emb_dim]\n",
        "        bilstm_out, self.hidden = self.bilstm(embeds, self.hidden) \n",
        "        # [sent_len, batch_size, emb_dim] --> [seq_len, batch, num_directions*hidden_size]\n",
        "        preds = self.fc(bilstm_out[-1])\n",
        "        # [batch, num_directions*hidden_size] --> [batch_size, 1]\n",
        "        return preds\n",
        "      \n",
        "      \n",
        "    def init_hidden(self):\n",
        "        # Before we've done anything, we dont have any hidden state.\n",
        "        # The axes semantics are (num_layers*2, minibatch_size, hidden_dim)\n",
        "        return (torch.zeros(self.num_layers*2, self.batch_size, self.hidden_dim).to(device),\n",
        "                torch.zeros(self.num_layers*2, self.batch_size, self.hidden_dim).to(device))\n",
        "        "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "77d6a83a-9570-4244-8615-d97a96076da8",
        "id": "ZSlP0-Pfe3eh",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 106
        }
      },
      "cell_type": "code",
      "source": [
        "model = BiLSTM_Model(vocab_size=len(TEXT.vocab), embedding_dim=300, hidden_dim=128, batch_size=BATCH_SIZE)\n",
        "model.to(device)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "BiLSTM_Model(\n",
              "  (word_embeddings): Embedding(25002, 300)\n",
              "  (bilstm): LSTM(300, 128, bidirectional=True)\n",
              "  (fc): Linear(in_features=256, out_features=1, bias=True)\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 52
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "UZDLKsOte3el",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "optimizer = optim.Adam(model.parameters(), lr=1e-3, amsgrad=True)\n",
        "criterion = nn.BCEWithLogitsLoss()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "1fGIOutGe3en",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Create training and validation dataloaders\n",
        "dataloaders_dict = {'train': train_iterator, \n",
        "                    'val': val_iterator}"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "OSPorkKge3er",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def train_model(model, dataloaders, criterion, optimizer, num_epochs, batch_size=BATCH_SIZE):\n",
        "    since = time.time()\n",
        "\n",
        "    history = dict()\n",
        "\n",
        "    best_model_wts = copy.deepcopy(model.state_dict())\n",
        "    best_acc = 0.0\n",
        "    skip_count = 0\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        print('Epoch {}/{}'.format(epoch, num_epochs - 1))\n",
        "        print('-' * 10)\n",
        "\n",
        "        # Each epoch has a training and validation phase\n",
        "        for phase in ['train', 'val']:\n",
        "            if phase == 'train':\n",
        "                model.train()  # Set model to training mode\n",
        "            else:\n",
        "                model.eval()   # Set model to evaluate mode\n",
        "\n",
        "            running_loss = 0.0\n",
        "            running_corrects = 0\n",
        "            \n",
        "\n",
        "            # Iterate over data.\n",
        "            for data in dataloaders[phase]:\n",
        "                inputs, labels = data.text, data.label\n",
        "                inputs = inputs.to(device)\n",
        "                labels = labels.to(device)\n",
        "                # zero the parameter gradients\n",
        "                optimizer.zero_grad()\n",
        "\n",
        "                # forward\n",
        "                # track history if only in train\n",
        "                with torch.set_grad_enabled(phase == 'train'):\n",
        "                    # Get model outputs and calculate loss\n",
        "\n",
        "                    # backward + optimize only if in training phase\n",
        "                    if phase == 'train':\n",
        "                        # we need to clear out the hidden state of the LSTM,\n",
        "                        # detaching it from its history on the last instance.\n",
        "                        model.batch_size = inputs.shape[1]\n",
        "                        model.hidden = model.init_hidden()\n",
        "                        \n",
        "                        outputs = model(inputs).squeeze(1)\n",
        "                        loss = criterion(outputs, labels)\n",
        "                        loss.backward()\n",
        "                        optimizer.step()\n",
        "                        \n",
        "                    else:\n",
        "                        model.batch_size = inputs.shape[1]\n",
        "                        model.hidden = model.init_hidden()\n",
        "                        \n",
        "                        outputs = model(inputs).squeeze(1)\n",
        "                        loss = criterion(outputs, labels)\n",
        "\n",
        "\n",
        "                # statistics\n",
        "                running_loss += loss.item()\n",
        "                outputs = torch.round(torch.sigmoid(outputs))\n",
        "                corrects = (outputs == labels).float()\n",
        "                acc = corrects.sum()/len(corrects)\n",
        "                running_corrects += acc.item()\n",
        "\n",
        "            epoch_loss = running_loss / len(dataloaders[phase])\n",
        "            epoch_acc = running_corrects / len(dataloaders[phase])\n",
        "\n",
        "            print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))\n",
        "\n",
        "            # deep copy the model\n",
        "            if phase == 'val' and epoch_acc > best_acc:\n",
        "                best_acc = epoch_acc\n",
        "                best_model_wts = copy.deepcopy(model.state_dict())\n",
        "            \n",
        "            if phase+'_acc' in history:\n",
        "                # append the new number to the existing array at this slot\n",
        "                history[phase+'_acc'].append(epoch_acc)\n",
        "            else:\n",
        "                # create a new array in this slot\n",
        "                history[phase+'_acc'] = [epoch_acc]\n",
        "            \n",
        "            if phase+'_loss' in history:\n",
        "                # append the new number to the existing array at this slot\n",
        "                history[phase+'_loss'].append(epoch_loss)\n",
        "            else:\n",
        "                # create a new array in this slot\n",
        "                history[phase+'_loss'] = [epoch_loss]            \n",
        "\n",
        "    time_elapsed = time.time() - since\n",
        "    print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))\n",
        "    print('Best val Acc: {:4f}'.format(best_acc))\n",
        "\n",
        "    # load best model weights\n",
        "    model.load_state_dict(best_model_wts)\n",
        "    return model, history\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "e21d92bb-7072-47ab-d962-869c139ae962",
        "id": "DYZKLbSHe3es",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 767
        }
      },
      "cell_type": "code",
      "source": [
        "model, history = train_model(model, dataloaders_dict, criterion, optimizer, num_epochs=10)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 0/9\n",
            "----------\n",
            "train Loss: 0.6944 Acc: 0.5044\n",
            "val Loss: 0.6929 Acc: 0.5105\n",
            "Epoch 1/9\n",
            "----------\n",
            "train Loss: 0.6928 Acc: 0.4961\n",
            "val Loss: 0.6942 Acc: 0.5196\n",
            "Epoch 2/9\n",
            "----------\n",
            "train Loss: 0.6911 Acc: 0.5059\n",
            "val Loss: 0.6975 Acc: 0.5235\n",
            "Epoch 3/9\n",
            "----------\n",
            "train Loss: 0.6892 Acc: 0.5055\n",
            "val Loss: 0.7004 Acc: 0.5216\n",
            "Epoch 4/9\n",
            "----------\n",
            "train Loss: 0.6877 Acc: 0.5073\n",
            "val Loss: 0.7012 Acc: 0.5368\n",
            "Epoch 5/9\n",
            "----------\n",
            "train Loss: 0.6790 Acc: 0.5638\n",
            "val Loss: 0.6614 Acc: 0.5963\n",
            "Epoch 6/9\n",
            "----------\n",
            "train Loss: 0.5491 Acc: 0.7439\n",
            "val Loss: 0.6189 Acc: 0.7551\n",
            "Epoch 7/9\n",
            "----------\n",
            "train Loss: 0.4026 Acc: 0.8378\n",
            "val Loss: 0.4393 Acc: 0.8172\n",
            "Epoch 8/9\n",
            "----------\n",
            "train Loss: 0.2977 Acc: 0.8925\n",
            "val Loss: 0.4131 Acc: 0.8331\n",
            "Epoch 9/9\n",
            "----------\n",
            "train Loss: 0.2157 Acc: 0.9284\n",
            "val Loss: 0.4260 Acc: 0.8420\n",
            "Training complete in 18m 0s\n",
            "Best val Acc: 0.841970\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "7a37a225-6e19-4a6b-a092-b79aaaaea8ff",
        "id": "6EjJEQ8se3ez",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 571
        }
      },
      "cell_type": "code",
      "source": [
        "# summarize history for accuracy\n",
        "plt.plot(history['train_acc'])\n",
        "plt.plot(history['val_acc'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()\n",
        "# summarize history for loss\n",
        "plt.plot(history['train_loss'])\n",
        "plt.plot(history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEVCAYAAADpbDJPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEVCAYAAADpbDJPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQggRpv4/r/inpty3lJ0AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "OG7Hv11oe3e1",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "text",
        "id": "s0ooeOpjfMD0"
      },
      "cell_type": "markdown",
      "source": [
        "## Custom Embedding Layer using GRU"
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "lLgRr9d2fMD0",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "TEXT.build_vocab(train_data, max_size=25000)\n",
        "LABEL.build_vocab(train_data)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "d5201799-583d-4a85-f6ea-dedf46d6e770",
        "id": "z5z6iItgfMD2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        }
      },
      "cell_type": "code",
      "source": [
        "print('Unique tokens in TEXT vocabulary:', len(TEXT.vocab))\n",
        "print('Unique tokens in LABEL vocabulary:',len(LABEL.vocab))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Unique tokens in TEXT vocabulary: 25002\n",
            "Unique tokens in LABEL vocabulary: 2\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "159bb8a3-d1f3-4d3a-c20c-d43bd018723b",
        "id": "ywIHZp0gfMD5",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 55
        }
      },
      "cell_type": "code",
      "source": [
        "print(TEXT.vocab.freqs.most_common(20))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[('the', 231930), (',', 220347), ('.', 188890), ('and', 125520), ('a', 125105), ('of', 115296), ('to', 107349), ('is', 86751), ('in', 70003), ('I', 62140), ('it', 61173), ('that', 56419), ('\"', 50909), (\"'s\", 49412), ('this', 48255), ('-', 42354), ('/><br', 40754), ('was', 40203), ('as', 34887), ('with', 34164)]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "9af20773-3057-4739-f6ac-2832ed4b74f2",
        "id": "ULyCo6KlfMD9",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "cell_type": "code",
      "source": [
        "print(TEXT.vocab.itos[:10])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "['<unk>', '<pad>', 'the', ',', '.', 'and', 'a', 'of', 'to', 'is']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "lHQbtazdfMEI",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "BATCH_SIZE = 64\n",
        "\n",
        "train_iterator, val_iterator, test_iterator = data.BucketIterator.splits(\n",
        "    (train_data, val_data, test_data),\n",
        "    batch_size=BATCH_SIZE,\n",
        "    device=device)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "n1LtxxSqfMEL",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "class GRU_Model(nn.Module):\n",
        "    \n",
        "    def __init__(self, vocab_size, embedding_dim, hidden_dim, batch_size):\n",
        "      \n",
        "      super(GRU_Model, self).__init__()\n",
        "      self.num_layers = 1\n",
        "      self.batch_size = batch_size\n",
        "      self.hidden_dim = hidden_dim\n",
        "      \n",
        "      self.word_embeddings = nn.Embedding(vocab_size, embedding_dim) \n",
        "      # The LSTM takes word embeddings as inputs, and outputs hidden states\n",
        "      # with dimensionality hidden_dim.\n",
        "      self.gru = nn.GRU(embedding_dim, hidden_dim, num_layers=self.num_layers)\n",
        "      self.fc = nn.Linear(hidden_dim, 1)\n",
        "      self.hidden = self.init_hidden()      \n",
        "      \n",
        "    def forward(self, sentence):\n",
        "      \n",
        "        embeds = self.word_embeddings(sentence)\n",
        "        # [sent_len, batch_size] --> [sent_len, batch_size, emb_dim]\n",
        "        gru_out, self.hidden = self.gru(embeds, self.hidden) \n",
        "        # [sent_len, batch_size, emb_dim] --> [seq_len, batch, num_directions*hidden_size]\n",
        "        preds = self.fc(gru_out[-1].squeeze(0))\n",
        "        # [batch, num_directions*hidden_size] --> [batch_size, 1]\n",
        "        return preds\n",
        "      \n",
        "      \n",
        "    def init_hidden(self):\n",
        "        # Before we've done anything, we dont have any hidden state.\n",
        "        # The axes semantics are (num_layers, minibatch_size, hidden_dim)\n",
        "        return torch.zeros(self.num_layers, self.batch_size, self.hidden_dim).to(device)\n",
        "        "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "d776d772-5037-4b78-a5f1-91fa1dd4b03d",
        "id": "2FNjRLiLfMEN",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 106
        }
      },
      "cell_type": "code",
      "source": [
        "model = GRU_Model(vocab_size=len(TEXT.vocab), embedding_dim=300, hidden_dim=128, batch_size=BATCH_SIZE)\n",
        "model.to(device)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GRU_Model(\n",
              "  (word_embeddings): Embedding(25002, 300)\n",
              "  (gru): GRU(300, 128)\n",
              "  (fc): Linear(in_features=128, out_features=1, bias=True)\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "Elg_XD8zfMEX",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "optimizer = optim.Adam(model.parameters(), lr=1e-3, amsgrad=True)\n",
        "criterion = nn.BCEWithLogitsLoss()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "T26yqAlLfMEZ",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Create training and validation dataloaders\n",
        "dataloaders_dict = {'train': train_iterator, \n",
        "                    'val': val_iterator}"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "by86t5uSfMEb",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def train_model(model, dataloaders, criterion, optimizer, num_epochs, batch_size=BATCH_SIZE):\n",
        "    since = time.time()\n",
        "\n",
        "    history = dict()\n",
        "\n",
        "    best_model_wts = copy.deepcopy(model.state_dict())\n",
        "    best_acc = 0.0\n",
        "    skip_count = 0\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        print('Epoch {}/{}'.format(epoch, num_epochs - 1))\n",
        "        print('-' * 10)\n",
        "\n",
        "        # Each epoch has a training and validation phase\n",
        "        for phase in ['train', 'val']:\n",
        "            if phase == 'train':\n",
        "                model.train()  # Set model to training mode\n",
        "            else:\n",
        "                model.eval()   # Set model to evaluate mode\n",
        "\n",
        "            running_loss = 0.0\n",
        "            running_corrects = 0\n",
        "            \n",
        "\n",
        "            # Iterate over data.\n",
        "            for data in dataloaders[phase]:\n",
        "                inputs, labels = data.text, data.label\n",
        "                inputs = inputs.to(device)\n",
        "                labels = labels.to(device)\n",
        "                # zero the parameter gradients\n",
        "                optimizer.zero_grad()\n",
        "\n",
        "                # forward\n",
        "                # track history if only in train\n",
        "                with torch.set_grad_enabled(phase == 'train'):\n",
        "                    # Get model outputs and calculate loss\n",
        "\n",
        "                    # backward + optimize only if in training phase\n",
        "                    if phase == 'train':\n",
        "                        # we need to clear out the hidden state of the LSTM,\n",
        "                        # detaching it from its history on the last instance.\n",
        "                        model.batch_size = inputs.shape[1]\n",
        "                        model.hidden = model.init_hidden()\n",
        "                        \n",
        "                        outputs = model(inputs).squeeze(1)\n",
        "                        loss = criterion(outputs, labels)\n",
        "                        loss.backward()\n",
        "                        optimizer.step()\n",
        "                        \n",
        "                    else:\n",
        "                        model.batch_size = inputs.shape[1]\n",
        "                        model.hidden = model.init_hidden()\n",
        "                        \n",
        "                        outputs = model(inputs).squeeze(1)\n",
        "                        loss = criterion(outputs, labels)\n",
        "\n",
        "\n",
        "                # statistics\n",
        "                running_loss += loss.item()\n",
        "                outputs = torch.round(torch.sigmoid(outputs))\n",
        "                corrects = (outputs == labels).float()\n",
        "                acc = corrects.sum()/len(corrects)\n",
        "                running_corrects += acc.item()\n",
        "\n",
        "            epoch_loss = running_loss / len(dataloaders[phase])\n",
        "            epoch_acc = running_corrects / len(dataloaders[phase])\n",
        "\n",
        "            print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))\n",
        "\n",
        "            # deep copy the model\n",
        "            if phase == 'val' and epoch_acc > best_acc:\n",
        "                best_acc = epoch_acc\n",
        "                best_model_wts = copy.deepcopy(model.state_dict())\n",
        "            \n",
        "            if phase+'_acc' in history:\n",
        "                # append the new number to the existing array at this slot\n",
        "                history[phase+'_acc'].append(epoch_acc)\n",
        "            else:\n",
        "                # create a new array in this slot\n",
        "                history[phase+'_acc'] = [epoch_acc]\n",
        "            \n",
        "            if phase+'_loss' in history:\n",
        "                # append the new number to the existing array at this slot\n",
        "                history[phase+'_loss'].append(epoch_loss)\n",
        "            else:\n",
        "                # create a new array in this slot\n",
        "                history[phase+'_loss'] = [epoch_loss]            \n",
        "\n",
        "    time_elapsed = time.time() - since\n",
        "    print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))\n",
        "    print('Best val Acc: {:4f}'.format(best_acc))\n",
        "\n",
        "    # load best model weights\n",
        "    model.load_state_dict(best_model_wts)\n",
        "    return model, history\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "8d1dc59b-98bd-4d26-8a0a-914fd10fbe02",
        "id": "8JevsEClfMEd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 767
        }
      },
      "cell_type": "code",
      "source": [
        "model, history = train_model(model, dataloaders_dict, criterion, optimizer, num_epochs=10)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 0/9\n",
            "----------\n",
            "train Loss: 0.6971 Acc: 0.4979\n",
            "val Loss: 0.7098 Acc: 0.5095\n",
            "Epoch 1/9\n",
            "----------\n",
            "train Loss: 0.6925 Acc: 0.5104\n",
            "val Loss: 0.7334 Acc: 0.5156\n",
            "Epoch 2/9\n",
            "----------\n",
            "train Loss: 0.5889 Acc: 0.6927\n",
            "val Loss: 0.4999 Acc: 0.7745\n",
            "Epoch 3/9\n",
            "----------\n",
            "train Loss: 0.4295 Acc: 0.8131\n",
            "val Loss: 0.4220 Acc: 0.8167\n",
            "Epoch 4/9\n",
            "----------\n",
            "train Loss: 0.2915 Acc: 0.8843\n",
            "val Loss: 0.3540 Acc: 0.8523\n",
            "Epoch 5/9\n",
            "----------\n",
            "train Loss: 0.1748 Acc: 0.9384\n",
            "val Loss: 0.3429 Acc: 0.8651\n",
            "Epoch 6/9\n",
            "----------\n",
            "train Loss: 0.1056 Acc: 0.9675\n",
            "val Loss: 0.3856 Acc: 0.8665\n",
            "Epoch 7/9\n",
            "----------\n",
            "train Loss: 0.0666 Acc: 0.9813\n",
            "val Loss: 0.3889 Acc: 0.8718\n",
            "Epoch 8/9\n",
            "----------\n",
            "train Loss: 0.0433 Acc: 0.9889\n",
            "val Loss: 0.4231 Acc: 0.8695\n",
            "Epoch 9/9\n",
            "----------\n",
            "train Loss: 0.0315 Acc: 0.9920\n",
            "val Loss: 0.4268 Acc: 0.8695\n",
            "Training complete in 13m 12s\n",
            "Best val Acc: 0.871835\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "outputId": "2959a648-1c30-4dac-ace9-f1d0e9f1e66c",
        "id": "UIHjEoGqfMEh",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 571
        }
      },
      "cell_type": "code",
      "source": [
        "# summarize history for accuracy\n",
        "plt.plot(history['train_acc'])\n",
        "plt.plot(history['val_acc'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()\n",
        "# summarize history for loss\n",
        "plt.plot(history['train_loss'])\n",
        "plt.plot(history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEVCAYAAADpbDJPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGQHUrgoiAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEVCAYAAADpbDJPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "2F9NlsTifMEk",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}