manuscript_cpu.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YmT-pZduM_1s"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "_uuid": "b6f19b34ab182e6f55204c91098a1829a36b8670",
    "colab": {},
    "colab_type": "code",
    "id": "CL7q3uDPM9Ki"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "from collections import Counter\n",
    "\n",
    "np.random.seed(42)\n",
    "\n",
    "def make_dictionary(sentences, vocabulary_size=None, initial_words=['<UNK>', '<PAD>', '<SOS>', '<EOS>']):\n",
    "    \"\"\"sentences : list of list\"\"\"\n",
    "    \n",
    "    counter = Counter()\n",
    "    for words in sentences:\n",
    "        counter.update(words)\n",
    "    \n",
    "    if vocabulary_size is None:\n",
    "        vocabulary_size = len(counter.keys())\n",
    "        \n",
    "    vocab_words = counter.most_common(vocabulary_size)\n",
    "    \n",
    "    for initial_word in initial_words:\n",
    "        vocab_words.insert(0, (initial_word, 0))\n",
    "    \n",
    "    word2idx = {word:idx for idx, (word, count) in enumerate(vocab_words)}\n",
    "    idx2word = {idx:word for word, idx in word2idx.items()}\n",
    "    \n",
    "    return word2idx, idx2word\n",
    "\n",
    "\n",
    "def process_sentences(sentences, word2idx, sentence_length=20, padding='<PAD>'):\n",
    "    \"\"\"sentences : list of list\n",
    "    Only paddding. No SOS or EOS\n",
    "    \"\"\"\n",
    "    \n",
    "    sentences_processed = []\n",
    "    for sentence in sentences:\n",
    "        if len(sentence) > sentence_length:\n",
    "            fixed_sentence = sentence[:sentence_length]\n",
    "        else:\n",
    "            fixed_sentence = sentence + [padding]*(sentence_length - len(sentence))\n",
    "        \n",
    "        sentence_idx = [word2idx[word] if word in word2idx.keys() else word2idx['<UNK>'] for word in fixed_sentence]\n",
    "        \n",
    "        sentences_processed.append(sentence_idx)\n",
    "\n",
    "    return sentences_processed\n",
    "\n",
    "def make_mask(sentences, sentence_length=20):\n",
    "    \n",
    "    masks = []\n",
    "    for sentence in sentences:\n",
    "        words_count = len(sentence[:sentence_length])\n",
    "        sentence_mask = np.concatenate([np.ones(words_count-1), np.ones(1), np.zeros(sentence_length-words_count)])\n",
    "        masks.append(sentence_mask)\n",
    "    \n",
    "    mask = np.array(masks)\n",
    "    return mask"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "5unVgwNsM9Kl"
   },
   "outputs": [],
   "source": [
    "import pickle\n",
    "with open('./params.pkl', 'rb') as pkl:\n",
    "    params = pickle.load(pkl)\n",
    "word2idx = params['word2idx']\n",
    "idx2word = params['idx2word']\n",
    "SENTENCE_LENGTH = params['sentence_length']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_uuid": "2de2e89be3197d328785437b56f24c5f2f969645",
    "colab_type": "text",
    "id": "GVTn9KE3M9Ko"
   },
   "source": [
    "# Build model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "_uuid": "96be6d6ef8ec8b533f6e9313e373749f4900023b",
    "colab": {},
    "colab_type": "code",
    "id": "xo1B0GqUM9Ko"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<torch._C.Generator at 0x11b438770>"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import torch\n",
    "from torch import nn\n",
    "from torch.utils.data import Dataset, DataLoader\n",
    "from torch.autograd import Variable\n",
    "from torch.optim import Adam, SGD\n",
    "\n",
    "torch.manual_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "_uuid": "4ff44374048e116e2a2fe88a008262ef9c2d7e34",
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "LX04GL2JM9Kq",
    "outputId": "eadbd251-2fed-4f12-8f43-4aa17a5a9b65"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "False"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "torch.cuda.is_available()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "_uuid": "6fe75e6516f3407b433974d21092173a52cae5a2",
    "colab": {},
    "colab_type": "code",
    "id": "fh5xMpYXM9Kv"
   },
   "outputs": [],
   "source": [
    "class Discriminator(nn.Module):\n",
    "    def __init__(self, vocab_size, embed_size, c_size, kernel_num, kernel_sizes):\n",
    "        super().__init__()\n",
    "        \n",
    "        self.embedding = nn.Embedding(\n",
    "            vocab_size, embed_size, padding_idx=0\n",
    "        )\n",
    "        self.conv_list = [\n",
    "            nn.Conv1d(embed_size, kernel_num, kernel_size=kernel_size) \n",
    "            for kernel_size in kernel_sizes\n",
    "        ]\n",
    "        self.convs = nn.ModuleList(self.conv_list)\n",
    "        \n",
    "        self.maxpools = nn.ModuleList([\n",
    "            nn.MaxPool1d(kernel_size) \n",
    "            for kernel_size in kernel_sizes\n",
    "        ])\n",
    "        \n",
    "        self.linear = nn.Linear(2200, c_size)\n",
    "        self.softmax = nn.LogSoftmax(dim=1)\n",
    "        self.sigmoid = nn.Sigmoid()\n",
    "        self.dropout = nn.Dropout(p=0.2)\n",
    "        \n",
    "    def forward(self, x):\n",
    "        batch_size = x.size(0)\n",
    "        embedded = self.embedding(x)\n",
    "        embedded = embedded.transpose(1,2)\n",
    "        \n",
    "        pools = []\n",
    "        for conv, maxpool in zip(self.convs, self.maxpools):\n",
    "            feature_map = conv(embedded)\n",
    "            pooled = maxpool(feature_map)\n",
    "            pools.append(pooled)\n",
    "            \n",
    "        conv_concat = torch.cat(pools, dim=-1).view(batch_size, -1)\n",
    "        conv_concat = self.dropout(conv_concat)\n",
    "        logits = self.linear(conv_concat)\n",
    "        return self.softmax(logits)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "_uuid": "d02075e3dd4d7403fc8a140fe9fbe260ab8b2217",
    "colab": {},
    "colab_type": "code",
    "id": "LPUHyw1tM9Kx"
   },
   "outputs": [],
   "source": [
    "D = Discriminator(\n",
    "    vocab_size=len(word2idx), \n",
    "    embed_size=128, \n",
    "    c_size=2, \n",
    "    kernel_num=100, \n",
    "    kernel_sizes=[2,3,4,5]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "_uuid": "63b1986939324618dd36fe7cff85f566ad31fdf9",
    "colab": {},
    "colab_type": "code",
    "id": "4bK8Rf5kM9Kz"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "IncompatibleKeys(missing_keys=[], unexpected_keys=[])"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "D.load_state_dict(torch.load('./D_180115.pth', map_location='cpu'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Discriminator(\n",
       "  (embedding): Embedding(9206, 128, padding_idx=0)\n",
       "  (convs): ModuleList(\n",
       "    (0): Conv1d(128, 100, kernel_size=(2,), stride=(1,))\n",
       "    (1): Conv1d(128, 100, kernel_size=(3,), stride=(1,))\n",
       "    (2): Conv1d(128, 100, kernel_size=(4,), stride=(1,))\n",
       "    (3): Conv1d(128, 100, kernel_size=(5,), stride=(1,))\n",
       "  )\n",
       "  (maxpools): ModuleList(\n",
       "    (0): MaxPool1d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "    (1): MaxPool1d(kernel_size=3, stride=3, padding=0, dilation=1, ceil_mode=False)\n",
       "    (2): MaxPool1d(kernel_size=4, stride=4, padding=0, dilation=1, ceil_mode=False)\n",
       "    (3): MaxPool1d(kernel_size=5, stride=5, padding=0, dilation=1, ceil_mode=False)\n",
       "  )\n",
       "  (linear): Linear(in_features=2200, out_features=2, bias=True)\n",
       "  (softmax): LogSoftmax()\n",
       "  (sigmoid): Sigmoid()\n",
       "  (dropout): Dropout(p=0.2)\n",
       ")"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "D.eval()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_uuid": "922af855c632e0e0517b713ff782b25517281e80",
    "colab_type": "text",
    "id": "o-Xe4sZoM9K1"
   },
   "source": [
    "# Inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "_uuid": "55ab3eebd57f1e2145559444b091465dbddd715d",
    "colab": {},
    "colab_type": "code",
    "id": "kWaDQtP_M9K2"
   },
   "outputs": [],
   "source": [
    "import itertools\n",
    "import re"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "_uuid": "1cf62c0abe595a7a6d93476fe351de24c66599e3",
    "colab": {},
    "colab_type": "code",
    "id": "Xvizh1VvM9K4"
   },
   "outputs": [],
   "source": [
    "def clean(s):\n",
    "    ss = re.sub('[^가-힣ㄱ-ㅎㅏ-ㅣA-Za-z0-9]+', '', s)\n",
    "    ssss = ''.join(ch if len(list(grouper)) == 1 else ch*2 for ch, grouper in itertools.groupby(ss))\n",
    "    return ssss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "_uuid": "252a6cec1f7b7fb04ccf52992645302e33422162",
    "colab": {},
    "colab_type": "code",
    "id": "lmBWw0anM9K7",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "def do_inference(raw_sentences, print_clean=False):\n",
    "    clean_sentences = [clean(s) for s in raw_sentences]\n",
    "    sentences = [list(''.join(clean_sentence.split())) for clean_sentence in clean_sentences]\n",
    "    infer_sentences_processed = process_sentences(sentences, word2idx, sentence_length=SENTENCE_LENGTH)\n",
    "    data = torch.LongTensor(infer_sentences_processed)\n",
    "    log_probs = D(Variable(data))\n",
    "    probs = log_probs.exp()\n",
    "    return probs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "_uuid": "fbf311c995d720ca2525a7fa6dde143d822f3cfc",
    "colab": {},
    "colab_type": "code",
    "id": "KwwJVdhaM9K9",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "def inference_one(string):\n",
    "    res = do_inference([string])\n",
    "#     return float(int(res[0][1]*100)/100)\n",
    "    return res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "_uuid": "f3bb9883bc035c7917d9921f893fe87c3397a372",
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "LuJ42AwfM9K_",
    "outputId": "e06bf3f7-37dc-451d-9ead-8a34b4ea6855",
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[7.3662e-04, 9.9926e-01]], grad_fn=<ExpBackward>)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "inference_one('ㅅㅂㅅㅂ')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "_uuid": "920f864502fad7b4cb18d9fccd26e1202148d64d",
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "IiDM8xBCM9LC",
    "outputId": "fd72b793-99f8-4ee1-c0fb-e19a3c3786f8",
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[0.8160, 0.1840]], grad_fn=<ExpBackward>)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "inference_one('넘 멋질 것 같아욧 짱이다')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "_uuid": "7590cc3842b9dfda2d4807943a86340f26c0a612",
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "bgKwYsc6M9LE",
    "outputId": "a6c82c75-5f1b-41a2-edcf-6f14326c76aa"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[0.9609, 0.0391]], grad_fn=<ExpBackward>)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "inference_one('안녕하세요')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[0.0241, 0.9759]], grad_fn=<ExpBackward>)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "inference_one('시발')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YGG3IoQ9M9LH"
   },
   "outputs": [],
   "source": [
    "# Lime 검증기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "id": "hAVah9HYNsEI",
    "outputId": "b858d57b-f462-479a-e52a-62eb640d53dc"
   },
   "outputs": [],
   "source": [
    "!pip install -q -U lime --no-deps # Already in Kaggle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Du62-6x2NsnQ"
   },
   "outputs": [],
   "source": [
    "from lime.lime_text import LimeTextExplainer\n",
    "explainer = LimeTextExplainer()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "RtNxFbkWNuo2"
   },
   "outputs": [],
   "source": [
    "def spacing_example(example):\n",
    "    length = len(example.split())\n",
    "    if length < 2:\n",
    "        spaced = ' '.join([c for c in example.replace(' ', '')])\n",
    "    else:\n",
    "        spaced = example\n",
    "    return spaced"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "OG_m-YRuNwcY"
   },
   "outputs": [],
   "source": [
    "def limer(example):\n",
    "    exp = explainer.explain_instance(spacing_example(example), lambda s: do_inference(s, True).detach().numpy(), top_labels=1)\n",
    "    exp.show_in_notebook()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/roamgom/Dropbox/github/manuscript/.venv/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:211: RuntimeWarning: Glyph 44060 missing from current font.\n",
      "  font.set_text(s, 0.0, flags=flags)\n",
      "/Users/roamgom/Dropbox/github/manuscript/.venv/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:211: RuntimeWarning: Glyph 45180 missing from current font.\n",
      "  font.set_text(s, 0.0, flags=flags)\n",
      "/Users/roamgom/Dropbox/github/manuscript/.venv/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:211: RuntimeWarning: Glyph 49352 missing from current font.\n",
      "  font.set_text(s, 0.0, flags=flags)\n",
      "/Users/roamgom/Dropbox/github/manuscript/.venv/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:176: RuntimeWarning: Glyph 44060 missing from current font.\n",
      "  font.load_char(ord(s), flags=flags)\n",
      "/Users/roamgom/Dropbox/github/manuscript/.venv/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:176: RuntimeWarning: Glyph 45180 missing from current font.\n",
      "  font.load_char(ord(s), flags=flags)\n",
      "/Users/roamgom/Dropbox/github/manuscript/.venv/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:176: RuntimeWarning: Glyph 49352 missing from current font.\n",
      "  font.load_char(ord(s), flags=flags)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "example = '개새끼'\n",
    "exp = explainer.explain_instance(spacing_example(example), lambda s: do_inference(s, True).detach().numpy(), top_labels=1)\n",
    "exp.as_pyplot_figure(label=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "ename": "SyntaxError",
     "evalue": "invalid syntax (<ipython-input-23-a914740aa8cd>, line 3)",
     "output_type": "error",
     "traceback": [
      "\u001b[0;36m  File \u001b[0;32m\"<ipython-input-23-a914740aa8cd>\"\u001b[0;36m, line \u001b[0;32m3\u001b[0m\n\u001b[0;31m    exp.\u001b[0m\n\u001b[0m        ^\u001b[0m\n\u001b[0;31mSyntaxError\u001b[0m\u001b[0;31m:\u001b[0m invalid syntax\n"
     ]
    }
   ],
   "source": [
    "def limer_html(example):\n",
    "    exp = explainer.explain_instance(spacing_example(example), lambda s: do_inference(s, True).detach().numpy(), top_labels=1)\n",
    "    exp."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 250
    },
    "colab_type": "code",
    "id": "FbRfAYC8NyBQ",
    "outputId": "675d4698-4b80-4cfc-e427-1659095aedd8"
   },
   "outputs": [],
   "source": [
    "limer(\"빨갱이들 마지막 발악하네. 다 들고 일어납시다.\t\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 192
    },
    "colab_type": "code",
    "id": "azAbxLTJNzdp",
    "outputId": "03606daf-e423-450e-bf60-bd456c2f9678"
   },
   "outputs": [],
   "source": [
    "limer('진짜 미쳐돌아가네 공산화되기 일보전이네..\t')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 279
    },
    "colab_type": "code",
    "id": "0xv4NrJ2PEhW",
    "outputId": "23f4477f-5319-4c68-b3e7-0fff7d0a148a"
   },
   "outputs": [],
   "source": [
    "limer('뼈때리니 발버둥치는구나 달창 년아. 넌 해골이 시키든?\t')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 250
    },
    "colab_type": "code",
    "id": "im_unCeaPR5G",
    "outputId": "3c5cdaf7-c7bd-44b7-b08b-ae72f18d3b0c"
   },
   "outputs": [],
   "source": [
    "limer('ad21**** 밑바닥 인생 부들부들 하네 ㅋㅋㅋ\t')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 135
    },
    "colab_type": "code",
    "id": "9mKNem7VPYTP",
    "outputId": "9b12c9b5-b55b-495b-fdd5-e4d56b53ae7f"
   },
   "outputs": [],
   "source": [
    "limer('빨갱이 새끼~~~~\t')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "doSeEZ6OPfX_"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "name": "욕탐-cpu-inference.ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}