lab-11_2_2_hihello.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 0x111e5fb50>"
      ]
     },
     "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": [
    "# declare dictionary\n",
    "char_set = ['h', 'i', 'e', 'l', 'o']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# hyper parameters\n",
    "input_size = len(char_set)\n",
    "hidden_size = len(char_set)\n",
    "learning_rate = 0.1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# data setting\n",
    "x_data = [[0, 1, 0, 2, 3, 3]]\n",
    "x_one_hot = [[[1, 0, 0, 0, 0],\n",
    "              [0, 1, 0, 0, 0],\n",
    "              [1, 0, 0, 0, 0],\n",
    "              [0, 0, 1, 0, 0],\n",
    "              [0, 0, 0, 1, 0],\n",
    "              [0, 0, 0, 1, 0]]]\n",
    "y_data = [[1, 0, 2, 3, 3, 4]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# declare RNN\n",
    "rnn = torch.nn.RNN(input_size, hidden_size, batch_first=True)  # batch_first guarantees the order of output = (B, S, F)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# loss & optimizer setting\n",
    "criterion = torch.nn.CrossEntropyLoss()\n",
    "optimizer = optim.Adam(rnn.parameters(), learning_rate)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 loss:  1.7802648544311523 prediction:  [[1 1 1 1 1 1]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  iiiiii\n",
      "1 loss:  1.4931954145431519 prediction:  [[1 4 1 1 4 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ioiioo\n",
      "2 loss:  1.3337129354476929 prediction:  [[1 3 2 3 1 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilelio\n",
      "3 loss:  1.215295433998108 prediction:  [[2 3 2 3 3 3]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  elelll\n",
      "4 loss:  1.1131411790847778 prediction:  [[2 3 2 3 3 3]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  elelll\n",
      "5 loss:  1.0241888761520386 prediction:  [[2 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  elello\n",
      "6 loss:  0.9573155045509338 prediction:  [[2 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  elello\n",
      "7 loss:  0.9102011322975159 prediction:  [[2 0 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ehello\n",
      "8 loss:  0.8731772899627686 prediction:  [[1 0 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ihello\n",
      "9 loss:  0.8399624824523926 prediction:  [[1 0 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ihello\n",
      "10 loss:  0.8088951706886292 prediction:  [[1 0 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ihello\n",
      "11 loss:  0.7812867760658264 prediction:  [[1 0 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ihello\n",
      "12 loss:  0.7585349082946777 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "13 loss:  0.7401294112205505 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "14 loss:  0.7243587970733643 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "15 loss:  0.709122359752655 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "16 loss:  0.6929269433021545 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "17 loss:  0.6821210980415344 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "18 loss:  0.6735268235206604 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "19 loss:  0.6595444679260254 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "20 loss:  0.6534828543663025 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "21 loss:  0.6465457081794739 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "22 loss:  0.6398184895515442 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "23 loss:  0.6381523013114929 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "24 loss:  0.6326718926429749 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "25 loss:  0.6256727576255798 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "26 loss:  0.6215079426765442 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "27 loss:  0.616705060005188 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "28 loss:  0.6099358201026917 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "29 loss:  0.6030194163322449 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "30 loss:  0.5992398262023926 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "31 loss:  0.5964334607124329 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "32 loss:  0.5916643738746643 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "33 loss:  0.5881562232971191 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "34 loss:  0.5854337811470032 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "35 loss:  0.5813184976577759 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "36 loss:  0.5761863589286804 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "37 loss:  0.5734922289848328 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "38 loss:  0.5727553963661194 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "39 loss:  0.5682081580162048 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "40 loss:  0.5656263828277588 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "41 loss:  0.5647333264350891 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "42 loss:  0.5629464983940125 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "43 loss:  0.5603764057159424 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "44 loss:  0.5588172078132629 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "45 loss:  0.5584632754325867 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "46 loss:  0.5565395355224609 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "47 loss:  0.5548029541969299 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "48 loss:  0.5542733669281006 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "49 loss:  0.5534438490867615 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "50 loss:  0.5520094037055969 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "51 loss:  0.5510937571525574 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "52 loss:  0.5506715178489685 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "53 loss:  0.5493640303611755 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "54 loss:  0.548336923122406 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "55 loss:  0.5478048920631409 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "56 loss:  0.5469381213188171 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "57 loss:  0.545922577381134 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "58 loss:  0.5454025268554688 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "59 loss:  0.544852077960968 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "60 loss:  0.5439804196357727 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "61 loss:  0.5434582233428955 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "62 loss:  0.5429832935333252 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "63 loss:  0.5422741174697876 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "64 loss:  0.5417040586471558 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "65 loss:  0.5413308143615723 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "66 loss:  0.5407487750053406 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "67 loss:  0.5402576923370361 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "68 loss:  0.5399190187454224 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "69 loss:  0.5394622683525085 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "70 loss:  0.5390090942382812 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "71 loss:  0.5387027263641357 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "72 loss:  0.538316011428833 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "73 loss:  0.5379175543785095 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "74 loss:  0.5376288890838623 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "75 loss:  0.5372945666313171 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "76 loss:  0.5369362235069275 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "77 loss:  0.5366637110710144 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "78 loss:  0.5363660454750061 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "79 loss:  0.53604656457901 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "80 loss:  0.5357930064201355 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "81 loss:  0.5355223417282104 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "82 loss:  0.5352355241775513 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "83 loss:  0.5349991917610168 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "84 loss:  0.5347511172294617 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "85 loss:  0.5344937443733215 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "86 loss:  0.5342753529548645 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "87 loss:  0.534046471118927 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "88 loss:  0.5338148474693298 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "89 loss:  0.5336135029792786 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "90 loss:  0.5334023833274841 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "91 loss:  0.533194899559021 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "92 loss:  0.5330093502998352 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "93 loss:  0.53281569480896 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "94 loss:  0.5326292514801025 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "95 loss:  0.5324582457542419 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "96 loss:  0.5322802066802979 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "97 loss:  0.5321123003959656 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "98 loss:  0.5319531559944153 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n",
      "99 loss:  0.5317898392677307 prediction:  [[1 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] prediction str:  ilello\n"
     ]
    }
   ],
   "source": [
    "# start training\n",
    "for i in range(100):\n",
    "    optimizer.zero_grad()\n",
    "    outputs, _status = rnn(X)\n",
    "    loss = criterion(outputs.view(-1, input_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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}