lab-11_2_1_charseq.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.optim as optim\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<torch._C.Generator at 0x1118cdb50>"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Random seed to make results deterministic and reproducible\n",
    "torch.manual_seed(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "sample = \" if you want you\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'y': 0, 't': 1, 'o': 2, 'i': 3, 'f': 4, 'w': 5, 'u': 6, 'a': 7, 'n': 8, ' ': 9}\n"
     ]
    }
   ],
   "source": [
    "# make dictionary\n",
    "char_set = list(set(sample))\n",
    "char_dic = {c: i for i, c in enumerate(char_set)}\n",
    "print(char_dic)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# hyper parameters\n",
    "dic_size = len(char_dic)\n",
    "hidden_size = len(char_dic)\n",
    "learning_rate = 0.1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# data setting\n",
    "sample_idx = [char_dic[c] for c in sample]\n",
    "x_data = [sample_idx[:-1]]\n",
    "x_one_hot = [np.eye(dic_size)[x] for x in x_data]\n",
    "y_data = [sample_idx[1:]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# transform as torch tensor variable\n",
    "X = torch.FloatTensor(x_one_hot)\n",
    "Y = torch.LongTensor(y_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# declare RNN\n",
    "rnn = torch.nn.RNN(dic_size, hidden_size, batch_first=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# loss & optimizer setting\n",
    "criterion = torch.nn.CrossEntropyLoss()\n",
    "optimizer = optim.Adam(rnn.parameters(), learning_rate)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 loss:  2.3307039737701416 prediction:  [[2 4 2 9 2 3 2 9 9 4 4 4 9 2 3]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  ofo oio  fff oi\n",
      "1 loss:  2.0587525367736816 prediction:  [[2 2 9 5 2 2 2 5 6 8 9 2 2 2 2]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  oo wooowun oooo\n",
      "2 loss:  1.8148916959762573 prediction:  [[0 6 9 5 2 6 9 5 6 8 9 9 2 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yu wou wun  oou\n",
      "3 loss:  1.6277488470077515 prediction:  [[0 6 9 0 2 6 9 5 0 8 9 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yu you wyn  you\n",
      "4 loss:  1.521714448928833 prediction:  [[0 6 9 0 2 6 9 5 7 8 9 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yu you wan  you\n",
      "5 loss:  1.438066840171814 prediction:  [[0 4 9 0 2 6 9 0 7 8 9 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yf you yan  you\n",
      "6 loss:  1.3512088060379028 prediction:  [[0 8 9 0 2 6 9 0 7 8 9 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yn you yan  you\n",
      "7 loss:  1.2912108898162842 prediction:  [[0 4 9 0 2 6 9 0 7 8 9 0 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yf you yan yyou\n",
      "8 loss:  1.2324522733688354 prediction:  [[0 4 9 0 2 6 9 0 7 8 1 0 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yf you yantyyou\n",
      "9 loss:  1.1849303245544434 prediction:  [[0 4 9 0 2 6 9 0 7 8 1 0 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yf you yantyyou\n",
      "10 loss:  1.1469171047210693 prediction:  [[0 4 9 0 2 6 9 0 7 8 1 0 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yf you yantyyou\n",
      "11 loss:  1.1173083782196045 prediction:  [[0 4 9 0 2 6 9 0 7 8 1 0 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yf you yantyyou\n",
      "12 loss:  1.0898127555847168 prediction:  [[0 4 9 0 2 6 9 0 7 8 1 0 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yf you yantyyou\n",
      "13 loss:  1.0637227296829224 prediction:  [[0 4 9 0 2 6 9 0 7 8 1 0 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  yf you yantyyou\n",
      "14 loss:  1.0346168279647827 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "15 loss:  1.0096397399902344 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "16 loss:  0.9882010817527771 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "17 loss:  0.9665720462799072 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "18 loss:  0.9490536451339722 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "19 loss:  0.9347025752067566 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "20 loss:  0.9210413694381714 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "21 loss:  0.9097891449928284 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "22 loss:  0.9012717008590698 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "23 loss:  0.8941951394081116 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "24 loss:  0.8881015181541443 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "25 loss:  0.8832717537879944 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "26 loss:  0.8788976669311523 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "27 loss:  0.8751237988471985 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "28 loss:  0.8722966313362122 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "29 loss:  0.8697354793548584 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "30 loss:  0.8674185872077942 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "31 loss:  0.8656530380249023 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "32 loss:  0.863921046257019 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "33 loss:  0.8623018860816956 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "34 loss:  0.8610238432884216 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "35 loss:  0.8597375750541687 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "36 loss:  0.8585304021835327 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "37 loss:  0.8575419187545776 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "38 loss:  0.8564358353614807 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "39 loss:  0.8553726077079773 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "40 loss:  0.8543716669082642 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "41 loss:  0.8532596230506897 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "42 loss:  0.8521810173988342 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "43 loss:  0.8511130809783936 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "44 loss:  0.8499239087104797 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "45 loss:  0.8487468957901001 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "46 loss:  0.84747314453125 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "47 loss:  0.8461035490036011 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "48 loss:  0.8447437882423401 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n",
      "49 loss:  0.8433225154876709 prediction:  [[3 4 9 0 2 6 9 0 7 8 1 9 0 2 6]] true Y:  [[3, 4, 9, 0, 2, 6, 9, 5, 7, 8, 1, 9, 0, 2, 6]] prediction str:  if you yant you\n"
     ]
    }
   ],
   "source": [
    "# start training\n",
    "for i in range(50):\n",
    "    optimizer.zero_grad()\n",
    "    outputs, _status = rnn(X)\n",
    "    loss = criterion(outputs.view(-1, dic_size), Y.view(-1))\n",
    "    loss.backward()\n",
    "    optimizer.step()\n",
    "\n",
    "    result = outputs.data.numpy().argmax(axis=2)\n",
    "    result_str = ''.join([char_set[c] for c in np.squeeze(result)])\n",
    "    print(i, \"loss: \", loss.item(), \"prediction: \", result, \"true Y: \", y_data, \"prediction str: \", result_str)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "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": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}