char-rnn-generate.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "import glob\n",
    "import unicodedata\n",
    "import string"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['./data/names/Arabic.txt', './data/names/English.txt', './data/names/German.txt', './data/names/Vietnamese.txt', './data/names/Italian.txt', './data/names/Czech.txt', './data/names/Scottish.txt', './data/names/Dutch.txt', './data/names/Korean.txt', './data/names/Chinese.txt', './data/names/Greek.txt', './data/names/Irish.txt', './data/names/French.txt', './data/names/Portuguese.txt', './data/names/Russian.txt', './data/names/Japanese.txt', './data/names/Polish.txt', './data/names/Spanish.txt']\n"
     ]
    }
   ],
   "source": [
    "def find_files(path): return glob.glob(path)\n",
    "print(find_files('./data/names/*.txt'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Slusarski\n"
     ]
    }
   ],
   "source": [
    "all_letters = string.ascii_letters + \" .,;'-\"\n",
    "n_letters = len(all_letters) + 1 # plus EOS marker\n",
    "\n",
    "# Turn a Unicode string to plain ASCII, thanks to http://stackoverflow.com/a/518232/2809427\n",
    "def unicode2ascii(s):\n",
    "    return ''.join(\n",
    "        c for c in unicodedata.normalize('NFD', s)\n",
    "        if unicodedata.category(c) != 'Mn'\n",
    "        and c in all_letters\n",
    "    )\n",
    "\n",
    "# Read a file and split into lines\n",
    "def read_lines(filename):\n",
    "    lines = open(filename).read().strip().split('\\n')\n",
    "    return [unicode2ascii(line) for line in lines]\n",
    "\n",
    "print(unicode2ascii(\"Ślusàrski\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# Build the category_lines dictionary, a list of names per language\n",
    "category_lines = {}\n",
    "all_categories = []\n",
    "\n",
    "for filename in find_files('./data/names/*.txt'):\n",
    "    category = filename.split('/')[-1].split('.')[0]\n",
    "    all_categories.append(category)\n",
    "    lines = read_lines(filename)\n",
    "    category_lines[category] = lines\n",
    "\n",
    "n_categories = len(all_categories)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# categories:  18 ['Arabic', 'English', 'German', 'Vietnamese', 'Italian', 'Czech', 'Scottish', 'Dutch', 'Korean', 'Chinese', 'Greek', 'Irish', 'French', 'Portuguese', 'Russian', 'Japanese', 'Polish', 'Spanish']\n",
      "Korean Samples ['Ahn', 'Baik', 'Bang', 'Byon', 'Cha']\n"
     ]
    }
   ],
   "source": [
    "print(\"# categories: \", n_categories, all_categories)\n",
    "print(\"Korean Samples\", category_lines['Korean'][:5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "from torch.autograd import Variable\n",
    "\n",
    "class RNN(nn.Module):\n",
    "    def __init__(self, input_size, hidden_size, output_size):\n",
    "        super(RNN, self).__init__()\n",
    "        self.hidden_size = hidden_size\n",
    "        \n",
    "        self.i2h = nn.Linear(n_categories + input_size + hidden_size, hidden_size)\n",
    "        self.i2o = nn.Linear(n_categories + input_size + hidden_size, output_size)\n",
    "        self.o2o = nn.Linear(hidden_size + output_size, output_size)\n",
    "        self.dropout = nn.Dropout(0.1)\n",
    "        self.softmax = nn.LogSoftmax()\n",
    "    \n",
    "    def forward(self, category, input, hidden):\n",
    "        input_combined = torch.cat((category, input, hidden), 1)\n",
    "        hidden = self.i2h(input_combined)\n",
    "        output = self.i2o(input_combined)\n",
    "        output_combined = torch.cat((hidden, output), 1)\n",
    "        output = self.o2o(output_combined)\n",
    "        output = self.dropout(output)\n",
    "        output = self.softmax(output)\n",
    "        return output, hidden\n",
    "    \n",
    "    def init_hidden(self):\n",
    "        return Variable(torch.zeros(1, self.hidden_size))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "import random\n",
    "\n",
    "# Random item from a list\n",
    "def randomChoice(l):\n",
    "    return l[random.randint(0, len(l) - 1)]\n",
    "\n",
    "# Get a random category and random line from that category\n",
    "def randomTrainingPair():\n",
    "    category = randomChoice(all_categories)\n",
    "    line = randomChoice(category_lines[category])\n",
    "    return category, line"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# One-hot vector for category\n",
    "def categoryTensor(category):\n",
    "    li = all_categories.index(category)\n",
    "    tensor = torch.zeros(1, n_categories)\n",
    "    tensor[0][li] = 1\n",
    "    return tensor\n",
    "\n",
    "# One-hot matrix of first to last letters (not including EOS) for input\n",
    "def inputTensor(line):\n",
    "    tensor = torch.zeros(len(line), 1, n_letters)\n",
    "    for li in range(len(line)):\n",
    "        letter = line[li]\n",
    "        tensor[li][0][all_letters.find(letter)] = 1\n",
    "    return tensor\n",
    "\n",
    "# LongTensor of second letter to end (EOS) for target\n",
    "def targetTensor(line):\n",
    "    letter_indexes = [all_letters.find(line[li]) for li in range(1, len(line))]\n",
    "    letter_indexes.append(n_letters - 1) # EOS\n",
    "    return torch.LongTensor(letter_indexes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\n",
    "# Make category, input, and target tensors from a random category, line pair\n",
    "def randomTrainingSet():\n",
    "    category, line = randomTrainingPair()\n",
    "    category_tensor = Variable(categoryTensor(category))\n",
    "    input_line_tensor = Variable(inputTensor(line))\n",
    "    target_line_tensor = Variable(targetTensor(line))\n",
    "    return category_tensor, input_line_tensor, target_line_tensor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "criterion = nn.NLLLoss()\n",
    "\n",
    "learning_rate = 0.0005\n",
    "\n",
    "def train(category_tensor, input_line_tensor, target_line_tensor):\n",
    "    hidden = rnn.init_hidden()\n",
    "\n",
    "    rnn.zero_grad()\n",
    "    \n",
    "    loss = 0\n",
    "    \n",
    "    for i in range(input_line_tensor.size()[0]):\n",
    "        output, hidden = rnn(category_tensor, input_line_tensor[i], hidden)\n",
    "        loss += criterion(output, target_line_tensor[i])\n",
    "\n",
    "    loss.backward()\n",
    "\n",
    "    for p in rnn.parameters():\n",
    "        p.data.add_(-learning_rate, p.grad.data)\n",
    "\n",
    "    return output, loss.data[0] / input_line_tensor.size()[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\n",
    "import time\n",
    "import math\n",
    "\n",
    "def timeSince(since):\n",
    "    now = time.time()\n",
    "    s = now - since\n",
    "    m = math.floor(s / 60)\n",
    "    s -= m * 60\n",
    "    return '%dm %ds' % (m, s)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0m 19s (5000 5%) 2.6837\n",
      "0m 37s (10000 10%) 1.9344\n",
      "1m 2s (15000 15%) 3.1829\n",
      "1m 20s (20000 20%) 2.0901\n",
      "1m 37s (25000 25%) 2.2600\n",
      "1m 54s (30000 30%) 2.4768\n",
      "2m 11s (35000 35%) 1.9242\n",
      "2m 48s (40000 40%) 3.0706\n",
      "3m 19s (45000 45%) 2.1971\n",
      "3m 56s (50000 50%) 2.3001\n",
      "4m 32s (55000 55%) 2.4356\n",
      "5m 5s (60000 60%) 2.7043\n",
      "5m 41s (65000 65%) 2.8568\n",
      "6m 25s (70000 70%) 2.9467\n",
      "6m 53s (75000 75%) 2.4472\n",
      "7m 26s (80000 80%) 2.0377\n",
      "7m 50s (85000 85%) 2.0354\n",
      "8m 14s (90000 90%) 2.9155\n",
      "8m 37s (95000 95%) 2.5176\n",
      "9m 0s (100000 100%) 3.0529\n"
     ]
    }
   ],
   "source": [
    "rnn = RNN(n_letters, 128, n_letters)\n",
    "\n",
    "n_epochs = 100000\n",
    "print_every = 5000\n",
    "plot_every = 500\n",
    "all_losses = []\n",
    "total_loss = 0 # Reset every plot_every epochs\n",
    "\n",
    "start = time.time()\n",
    "\n",
    "for epoch in range(1, n_epochs + 1):\n",
    "    output, loss = train(*randomTrainingSet())\n",
    "    total_loss += loss\n",
    "    \n",
    "    if epoch % print_every == 0:\n",
    "        print('%s (%d %d%%) %.4f' % (timeSince(start), epoch, epoch / n_epochs * 100, loss))\n",
    "\n",
    "    if epoch % plot_every == 0:\n",
    "        all_losses.append(total_loss / plot_every)\n",
    "        total_loss = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f125e88f898>]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAD8CAYAAABw1c+bAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\njTGmlbHEb4wxrYwlfmOMaWX+H05ko91iwPRJAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f12acacec88>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.ticker as ticker\n",
    "%matplotlib inline\n",
    "\n",
    "plt.figure()\n",
    "plt.plot(all_losses)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "max_length = 20\n",
    "\n",
    "# Sample from a category and starting letter\n",
    "def sample(category, start_letter='A'):\n",
    "    category_tensor = Variable(categoryTensor(category))\n",
    "    input = Variable(inputTensor(start_letter))\n",
    "    hidden = rnn.init_hidden()\n",
    "\n",
    "    output_name = start_letter\n",
    "    \n",
    "    for i in range(max_length):\n",
    "        output, hidden = rnn(category_tensor, input[0], hidden)\n",
    "        topv, topi = output.data.topk(1)\n",
    "        topi = topi[0][0]\n",
    "        if topi == n_letters - 1:\n",
    "            break\n",
    "        else:    \n",
    "            letter = all_letters[topi]\n",
    "            output_name += letter\n",
    "        input = Variable(inputTensor(letter))\n",
    "\n",
    "    return output_name\n",
    "\n",
    "# Get multiple samples from one category and multiple starting letters\n",
    "def samples(category, start_letters='ABC'):\n",
    "    for start_letter in start_letters:\n",
    "        print(sample(category, start_letter))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "Columns 0 to 7 \n",
      " -4.3618  -6.4685  -5.5572  -5.7780  -2.1311  -5.1758  -7.2366  -4.3618\n",
      "\n",
      "Columns 8 to 15 \n",
      " -2.9089  -7.3804  -4.3618  -5.5018  -6.8049  -5.3581  -2.0561  -6.3749\n",
      "\n",
      "Columns 16 to 23 \n",
      " -7.7400  -5.7679  -3.7401  -5.3510  -5.2909  -4.5763  -8.6225  -7.6426\n",
      "\n",
      "Columns 24 to 31 \n",
      " -3.7029  -6.6118 -10.0374  -4.3618 -10.0299 -10.1536 -10.5617 -10.0057\n",
      "\n",
      "Columns 32 to 39 \n",
      "-10.6159 -10.1043 -10.2568 -10.3292  -9.9136 -10.1090  -9.8665 -10.0643\n",
      "\n",
      "Columns 40 to 47 \n",
      "-10.1542 -10.2233 -10.2552 -10.0955  -4.3618  -4.3618 -10.3225 -10.1535\n",
      "\n",
      "Columns 48 to 55 \n",
      "-10.3798  -4.3618  -9.4846 -10.5004  -7.8044 -10.5546 -10.3649  -4.3618\n",
      "\n",
      "Columns 56 to 58 \n",
      " -4.3618  -9.6946  -0.7309\n",
      "[torch.FloatTensor of size 1x59]\n",
      "\n",
      "Rovakov\n",
      "\n",
      "\n",
      "Columns 0 to 7 \n",
      " -2.6788  -5.9787  -4.0485  -4.0485  -2.2653  -4.7267  -5.9963  -4.4737\n",
      "\n",
      "Columns 8 to 15 \n",
      " -3.3885  -6.3507  -4.4089  -4.0485  -6.3201  -4.9112  -2.5057  -4.0485\n",
      "\n",
      "Columns 16 to 23 \n",
      " -7.1685  -4.9019  -3.3162  -4.0485  -5.5367  -4.3063  -7.7782  -7.0433\n",
      "\n",
      "Columns 24 to 31 \n",
      " -3.7488  -5.9542  -4.0485  -9.5683  -9.4676  -9.5159  -9.8848  -9.4191\n",
      "\n",
      "Columns 32 to 39 \n",
      " -9.8855  -9.4618  -9.6554  -9.6305  -9.2810  -9.4778  -9.2664  -9.3336\n",
      "\n",
      "Columns 40 to 47 \n",
      " -9.4293  -4.0485  -9.6352  -9.4339  -9.8284  -9.3219  -9.5953  -9.3783\n",
      "\n",
      "Columns 48 to 55 \n",
      " -9.6766  -9.3245  -8.7606  -4.0485  -6.9446  -9.9242  -9.6035  -9.6714\n",
      "\n",
      "Columns 56 to 58 \n",
      "-10.4061  -9.1045  -0.8398\n",
      "[torch.FloatTensor of size 1x59]\n",
      "\n",
      "Uantanov\n",
      "\n",
      "\n",
      "Columns 0 to 7 \n",
      " -2.1858  -6.1611  -4.8507  -4.1367  -1.9766  -5.1331  -7.0398  -4.8450\n",
      "\n",
      "Columns 8 to 15 \n",
      " -2.5287  -7.0875  -4.1367  -5.1082  -6.5362  -5.0047  -2.2352  -6.1900\n",
      "\n",
      "Columns 16 to 23 \n",
      " -7.7156  -5.8143  -4.1367  -4.5919  -5.2901  -3.9507  -4.1367  -7.5146\n",
      "\n",
      "Columns 24 to 31 \n",
      " -3.4789  -6.0326  -9.9633 -10.1665  -9.9974 -10.0680 -10.5547 -10.0712\n",
      "\n",
      "Columns 32 to 39 \n",
      "-10.6065  -9.8598 -10.3478 -10.3225  -9.8276 -10.1775  -9.8774 -10.0152\n",
      "\n",
      "Columns 40 to 47 \n",
      "-10.0429 -10.1870 -10.2539 -10.0500  -4.1367  -9.7594 -10.2552 -10.2054\n",
      "\n",
      "Columns 48 to 55 \n",
      "-10.2663 -10.2237  -9.4943 -10.3988  -7.8236 -10.4719 -10.2825 -10.4001\n",
      "\n",
      "Columns 56 to 58 \n",
      "-10.7363  -9.7314  -1.0009\n",
      "[torch.FloatTensor of size 1x59]\n",
      "\n",
      "Shavavav\n"
     ]
    }
   ],
   "source": [
    "samples('Russian', 'RUS')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Garten\n",
      "Eren\n",
      "Roure\n"
     ]
    }
   ],
   "source": [
    "samples('German', 'GER')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Kon\n",
      "Oo\n",
      "Ro\n"
     ]
    }
   ],
   "source": [
    "samples('Korean', 'KOR')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Can\n",
      "Han\n",
      "Iun\n"
     ]
    }
   ],
   "source": [
    "samples('Chinese', 'CHI')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.5.2+"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}