py2.[실습02]MLP-학습.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "6830965f",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.optim as optim\n",
    "import torchvision.datasets as dset\n",
    "import torchvision.transforms as transforms\n",
    "from torch.utils.data import DataLoader\n",
    "from torch.autograd import Variable\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fbd91b96",
   "metadata": {},
   "source": [
    "### 1. MNIST train, test dataset 가져오기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "f9c4688b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# \"\": 현재 폴더에 MNIST 있음\n",
    "mnist_train = dset.MNIST(\"\", train=True, transform=transforms.ToTensor(), target_transform=None, download=True)  # train용으로 쓰겠다.\n",
    "mnist_test = dset.MNIST(\"\", train=False, transform=transforms.ToTensor(), target_transform=None, download=True)  # test용으로 쓰겠다."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c8437689",
   "metadata": {},
   "source": [
    "### 2. 대략적인 데이터 형태"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "4316dc68",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mnist_train 길이: 60000\n",
      "mnist_test 길이: 10000\n",
      "image data 형태: torch.Size([1, 28, 28])\n",
      "label: 5\n"
     ]
    },
    {
     "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": [
    "print \"mnist_train 길이:\", len(mnist_train)\n",
    "print \"mnist_test 길이:\", len(mnist_test)\n",
    "\n",
    "# 데이터 하나 형태\n",
    "image, label = mnist_train.__getitem__(0)  # 0번째 데이터\n",
    "print \"image data 형태:\", image.size()\n",
    "print \"label:\", label\n",
    "\n",
    "# 그리기\n",
    "img = image.numpy()  # image 타입을 numpy 로 변환(1,28,28)\n",
    "plt.title(\"label: %d\" %label)\n",
    "plt.imshow(img[0], cmap='gray')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "337a9920",
   "metadata": {},
   "source": [
    "### 3. 데이터 로드함수\n",
    "학습시킬 때 batch_size 단위로 끊어서 로드하기 위함"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "35b17e43",
   "metadata": {},
   "outputs": [],
   "source": [
    "# hyper parameters\n",
    "batch_size = 1024\n",
    "learning_rate = 0.01\n",
    "num_epoch = 400"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "3cb7b4bb",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_loader = torch.utils.data.DataLoader(mnist_train, batch_size=batch_size,  # mnist_train를 train 시키자.\n",
    "                                      shuffle=True, num_workers=2,\n",
    "                                      drop_last=True)  # batch_size 만큼 나눌 때 나머지는 버려라.\n",
    "test_loader = torch.utils.data.DataLoader(mnist_test, batch_size=batch_size,  # mnist_train를 train 시키자.\n",
    "                                      shuffle=False, num_workers=2,\n",
    "                                      drop_last=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e3b1cfba",
   "metadata": {},
   "source": [
    "### 데이터 로드함수 이해하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "a3b3f63d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0]\n",
      "한 번에 로드되는 데이터 크기: 1024\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1]\n",
      "한 번에 로드되는 데이터 크기: 1024\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2]\n",
      "한 번에 로드되는 데이터 크기: 1024\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[3]\n",
      "한 번에 로드되는 데이터 크기: 1024\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[4]\n",
      "한 번에 로드되는 데이터 크기: 1024\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[5]\n",
      "한 번에 로드되는 데이터 크기: 1024\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "n = 3  # 샘플로 그려볼 데이터 개수\n",
    "for i, [imgs, labels] in enumerate(test_loader):  # batch_size 만큼\n",
    "    if i>5:\n",
    "        break\n",
    "    \n",
    "    print \"[%d]\" %i\n",
    "    print \"한 번에 로드되는 데이터 크기:\", len(imgs)\n",
    "    \n",
    "    # 그리기\n",
    "    for j in range(n):\n",
    "        img = imgs[j].numpy()  # image 타입을 numpy로 변환 (1,28,28)\n",
    "        img = img.reshape((img.shape[1], img.shape[2]))  # (1,28,28) -> (28,28)\n",
    "        \n",
    "        plt.subplot(1,n,j+1)  # (1,3) 형태 플랏의 j번째 자리에 그리겠다.\n",
    "        plt.imshow(img, cmap='gray')\n",
    "        plt.title(\"label: %d\" %labels[j])\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aa029494",
   "metadata": {},
   "source": [
    "### 4. 모델 선언"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "5d2950f7",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 모델 선언\n",
    "# * 퍼셉트론(2 hidden layer) *\n",
    "model = nn.Sequential(\n",
    "    nn.Linear(28*28,256),\n",
    "    nn.Sigmoid(),\n",
    "    nn.Linear(256,128),\n",
    "    nn.Linear(128,10))\n",
    "# 파라미터 보기\n",
    "# print(list(model.parameters()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "40408569",
   "metadata": {},
   "outputs": [],
   "source": [
    "def ComputeAccur(dloader, imodel):\n",
    "    correct = 0\n",
    "    total = 0\n",
    "    \n",
    "    for j, [imgs,labels] in enumerate(dloader):  # batch_size 만큼\n",
    "        img = imgs\n",
    "        label = Variable(labels)\n",
    "        \n",
    "        # (batch_size,1,28,28) -> (batch_size,28,28)\n",
    "        img = img.reshape((img.shape[0], img.shape[2], img.shape[3]))\n",
    "        # (batch_size,28,28) -> (batch_size,28*28)\n",
    "        img = img.reshape((img.shape[0], img.shape[1]*img.shape[2]))\n",
    "        img = Variable(img, requires_grad=False)\n",
    "        \n",
    "        output = imodel(img)\n",
    "        _, output_index = torch.max(output, 1)\n",
    "        \n",
    "        total += label.size(0)         \n",
    "        correct += (output_index==label).sum().float()\n",
    "        \n",
    "    print(\"Accuracy of Test Data: {}\".format(100*correct/total))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "f19ed8eb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy of Test Data: 9.63541698456\n"
     ]
    }
   ],
   "source": [
    "ComputeAccur(test_loader, model)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2ab9e250",
   "metadata": {},
   "source": [
    "### 5. loss, optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "fe5430b8",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_func = nn.CrossEntropyLoss()  # logit(# of classes), target(1)\n",
    "optimizer = optim.SGD(model.parameters(), lr=learning_rate)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9ff9c7b8",
   "metadata": {},
   "source": [
    "### 6. 학습"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "5259c094",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0..\n",
      "Accuracy of Test Data: 11.2955732346\n",
      "tensor(2.2962, grad_fn=<NllLossBackward>)\n",
      "50..\n",
      "Accuracy of Test Data: 81.922744751\n",
      "tensor(0.6807, grad_fn=<NllLossBackward>)\n",
      "100..\n",
      "Accuracy of Test Data: 88.3138046265\n",
      "tensor(0.4667, grad_fn=<NllLossBackward>)\n",
      "150..\n",
      "Accuracy of Test Data: 89.832901001\n",
      "tensor(0.4193, grad_fn=<NllLossBackward>)\n",
      "200..\n",
      "Accuracy of Test Data: 90.6792526245\n",
      "tensor(0.3102, grad_fn=<NllLossBackward>)\n",
      "250..\n",
      "Accuracy of Test Data: 91.221786499\n",
      "tensor(0.3431, grad_fn=<NllLossBackward>)\n",
      "300..\n",
      "Accuracy of Test Data: 91.579864502\n",
      "tensor(0.3072, grad_fn=<NllLossBackward>)\n",
      "350..\n",
      "Accuracy of Test Data: 91.8619766235\n",
      "tensor(0.2768, grad_fn=<NllLossBackward>)\n"
     ]
    }
   ],
   "source": [
    "num_epoch = 400\n",
    "for i in range(num_epoch):\n",
    "    for j, [imgs, labels] in enumerate(train_loader):  # batch_size 만큼\n",
    "        img = imgs\n",
    "        label = Variable(labels)\n",
    "        \n",
    "        # (batch_size,1,28,28) -> (batch_size,28,28)\n",
    "        img = img.reshape((img.shape[0], img.shape[2], img.shape[3]))\n",
    "        # (batch_size,28,28) -> (batch_size, 28*28)\n",
    "        img = img.reshape((img.shape[0], img.shape[1]*img.shape[2]))\n",
    "        img = Variable(img, requires_grad=True)\n",
    "        # img = Variable(img, requires_grad=True).cuda()\n",
    "        \n",
    "        optimizer.zero_grad()\n",
    "        output = model(img)  # forward prop.\n",
    "        loss = loss_func(output, label)  # logit(# of classes), target(1)\n",
    "        \n",
    "        loss.backward()  # back prop.\n",
    "        optimizer.step()  # weight 조정\n",
    "    \n",
    "    if i%50==0:\n",
    "        print(\"%d..\" %i)\n",
    "        ComputeAccur(test_loader, model)\n",
    "        print loss"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7304df5c",
   "metadata": {},
   "source": [
    "### 7. 테스트"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "13347155",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy of Test Data: 92.111541748\n"
     ]
    }
   ],
   "source": [
    "ComputeAccur(test_loader, model)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "702b64d8",
   "metadata": {},
   "source": [
    "### 8. 학습된 파라미터 저장"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "3b741c7d",
   "metadata": {},
   "outputs": [],
   "source": [
    "netname = './nets/mlp_weight.pkl'\n",
    "torch.save(model, netname, )\n",
    "\n",
    "# model = torch.load(netname)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "conda_pytorch_p27",
   "language": "python",
   "name": "conda_pytorch_p27"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}