pretrain autoencoder.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "from torchvision import datasets, transforms\n",
    "from torch.utils.data import DataLoader\n",
    "\n",
    "from torchvision.utils import save_image\n",
    "\n",
    "import pandas as pd\n",
    "\n",
    "import random\n",
    "\n",
    "from tqdm.notebook import tqdm\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.pyplot import imshow\n",
    "\n",
    "from model_defs import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def to_img(x):\n",
    "    x = 0.5 * (x + 1)\n",
    "    x = x.clamp(0, 1)\n",
    "    x = x.view(x.size(0), 1, 28, 28)\n",
    "    return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "transform=transforms.Compose([\n",
    "        transforms.ToTensor(),\n",
    "        transforms.Normalize((0.1722,), (0.3242,))\n",
    "        ])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# dataset1 = datasets.MNIST('./data', train=True, download=True,\n",
    "#                        transform=transform)\n",
    "# dataset2 = datasets.MNIST('./data', train=False, download = True,\n",
    "#                        transform=transform)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# hyper param\n",
    "num_epochs = 60\n",
    "batch_size = 128\n",
    "learning_rate = 1e-3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = datasets.EMNIST('./data', train = True, transform = transforms.ToTensor(), split = 'letters', download = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataloader = DataLoader(dataset, batch_size = batch_size, shuffle = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# model = autoencoder().cuda()\n",
    "model = autoencoder_v2().cuda()\n",
    "criterion = nn.MSELoss()\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate,\n",
    "                             weight_decay=1e-5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "07cc71b13d7946c79a67cd6be57f18be",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=60.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch [1/60], loss:0.0181\n",
      "epoch [2/60], loss:0.0169\n",
      "epoch [3/60], loss:0.0168\n",
      "epoch [4/60], loss:0.0155\n",
      "epoch [5/60], loss:0.0138\n",
      "epoch [6/60], loss:0.0154\n",
      "epoch [7/60], loss:0.0149\n",
      "epoch [8/60], loss:0.0156\n",
      "epoch [9/60], loss:0.0142\n",
      "epoch [10/60], loss:0.0152\n",
      "epoch [11/60], loss:0.0129\n",
      "epoch [12/60], loss:0.0132\n",
      "epoch [13/60], loss:0.0137\n",
      "epoch [14/60], loss:0.0131\n",
      "epoch [15/60], loss:0.0134\n",
      "epoch [16/60], loss:0.0134\n",
      "epoch [17/60], loss:0.0131\n",
      "epoch [18/60], loss:0.0134\n",
      "epoch [19/60], loss:0.0135\n",
      "epoch [20/60], loss:0.0134\n",
      "epoch [21/60], loss:0.0122\n",
      "epoch [22/60], loss:0.0118\n",
      "epoch [23/60], loss:0.0134\n",
      "epoch [24/60], loss:0.0134\n",
      "epoch [25/60], loss:0.0127\n",
      "epoch [26/60], loss:0.0123\n",
      "epoch [27/60], loss:0.0121\n",
      "epoch [28/60], loss:0.0123\n",
      "epoch [29/60], loss:0.0120\n",
      "epoch [30/60], loss:0.0128\n",
      "epoch [31/60], loss:0.0126\n",
      "epoch [32/60], loss:0.0123\n",
      "epoch [33/60], loss:0.0126\n",
      "epoch [34/60], loss:0.0122\n",
      "epoch [35/60], loss:0.0137\n",
      "epoch [36/60], loss:0.0133\n",
      "epoch [37/60], loss:0.0116\n",
      "epoch [38/60], loss:0.0122\n",
      "epoch [39/60], loss:0.0120\n",
      "epoch [40/60], loss:0.0112\n",
      "epoch [41/60], loss:0.0126\n",
      "epoch [42/60], loss:0.0140\n",
      "epoch [43/60], loss:0.0122\n",
      "epoch [44/60], loss:0.0118\n",
      "epoch [45/60], loss:0.0120\n",
      "epoch [46/60], loss:0.0121\n",
      "epoch [47/60], loss:0.0106\n",
      "epoch [48/60], loss:0.0118\n",
      "epoch [49/60], loss:0.0120\n",
      "epoch [50/60], loss:0.0124\n",
      "epoch [51/60], loss:0.0117\n",
      "epoch [52/60], loss:0.0122\n",
      "epoch [53/60], loss:0.0116\n",
      "epoch [54/60], loss:0.0122\n",
      "epoch [55/60], loss:0.0124\n",
      "epoch [56/60], loss:0.0113\n",
      "epoch [57/60], loss:0.0133\n",
      "epoch [58/60], loss:0.0105\n",
      "epoch [59/60], loss:0.0123\n",
      "epoch [60/60], loss:0.0122\n",
      "\n"
     ]
    }
   ],
   "source": [
    "losses = []\n",
    "for epoch in tqdm(range(num_epochs)):\n",
    "    for data in dataloader:\n",
    "        img, l = data\n",
    "        img = img.cuda()\n",
    "        # ===================forward=====================\n",
    "        output = model(img)\n",
    "        loss = criterion(output, img)\n",
    "        # ===================backward====================\n",
    "        optimizer.zero_grad()\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "    # ===================log========================\n",
    "    print('epoch [{}/{}], loss:{:.4f}'\n",
    "          .format(epoch+1, num_epochs, loss.item()))\n",
    "    losses.append(loss.item())\n",
    "    \n",
    "    if epoch % 10 == 0:\n",
    "        pic = to_img(output.cpu().data)\n",
    "        save_image(pic, './dc_img/image_v2_{}.png'.format(epoch))\n",
    "    \n",
    "torch.save(model.state_dict(), './conv_autoencoder_v2.pth')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f76c8047c10>]"
      ]
     },
     "execution_count": 11,
     "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": [
    "plt.plot(losses)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def re_cons2(img):\n",
    "    re_img = encode(img.view(1,1,28,28))\n",
    "    re_img = re_img.cpu()\n",
    "    imshow(re_img.view(28,28).data.numpy())\n",
    "\n",
    "encode = autoencoder_v3()\n",
    "encode.load_state_dict(torch.load('./conv_autoencoder_v3.pth'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def re_cons(img):\n",
    "    re_img = model(img.view(1,1,28,28).cuda())\n",
    "    re_img = re_img.cpu()\n",
    "    imshow(re_img.view(28,28).data.numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "mnist = datasets.MNIST('./data', train=True, download = True, transform = transforms.ToTensor())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "n = random.randint(0,len(mnist))\n",
    "img = mnist[n][0]\n",
    "imshow(img.view(28,-1))\n",
    "print(n)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "re_cons(img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "re_cons2(img)"
   ]
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}