CoGAN.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from __future__ import print_function\n",
    "#%matplotlib inline\n",
    "import argparse\n",
    "import os\n",
    "import random\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.parallel\n",
    "import torch.backends.cudnn as cudnn\n",
    "import torch.optim as optim\n",
    "import torch.utils.data\n",
    "import torchvision.datasets as dset\n",
    "import torchvision.transforms as transforms\n",
    "import torchvision.utils as vutils\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.animation as animation\n",
    "from IPython.display import HTML\n",
    "from torch.autograd import Variable\n",
    "from torchvision.utils import save_image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Root directory for dataset\n",
    "dataroot = \"data\"\n",
    "\n",
    "# Number of workers for dataloader\n",
    "workers = 2\n",
    "\n",
    "# Batch size during training\n",
    "batch_size = 10\n",
    "\n",
    "# Spatial size of training images. All images will be resized to this\n",
    "#   size using a transformer.\n",
    "image_size = 32\n",
    "\n",
    "# Number of channels in the training images. For color images this is 3\n",
    "nc = 3\n",
    "\n",
    "# Size of z latent vector (i.e. size of generator input)\n",
    "nz = 100\n",
    "\n",
    "# Size of feature maps in generator\n",
    "ngf = 32\n",
    "\n",
    "# Size of feature maps in discriminator\n",
    "ndf = 32\n",
    "\n",
    "# Number of training epochs\n",
    "num_epochs = 3\n",
    "\n",
    "# Learning rate for optimizers\n",
    "lr = 0.0002\n",
    "\n",
    "# Beta1 hyperparam for Adam optimizers\n",
    "beta1 = 0.5\n",
    "\n",
    "# Number of GPUs available. Use 0 for CPU mode.\n",
    "ngpu = 1\n",
    "\n",
    "\n",
    "# The interval between image sampling.\n",
    "sample_interval = 400 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x19512bf9f48>"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# We can use an image folder dataset the way we have it setup.\n",
    "# Create the dataset\n",
    "dataset = dset.ImageFolder(root=dataroot,\n",
    "                           transform=transforms.Compose([\n",
    "                               transforms.Resize(image_size),\n",
    "                               transforms.CenterCrop(image_size),\n",
    "                               transforms.ToTensor(),\n",
    "                               transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),\n",
    "                           ]))\n",
    "\n",
    "\n",
    "# Create the dataloader\n",
    "dataloader1 = torch.utils.data.DataLoader(dataset, batch_size=batch_size,\n",
    "                                         shuffle=True, num_workers=workers)\n",
    "\n",
    "dataloader2 = torch.utils.data.DataLoader(dataset, batch_size=batch_size,\n",
    "                                         shuffle=True, num_workers=workers)\n",
    "\n",
    "# Decide which device we want to run on\n",
    "device = torch.device(\"cuda:0\" if (torch.cuda.is_available() and ngpu > 0) else \"cpu\")\n",
    "\n",
    "# Plot some training images\n",
    "real_batch = next(iter(dataloader1))\n",
    "plt.figure(figsize=(8,8))\n",
    "plt.axis(\"off\")\n",
    "plt.title(\"Training Images\")\n",
    "plt.imshow(np.transpose(vutils.make_grid(real_batch[0].to(device)[:64], padding=2, normalize=True).cpu(),(1,2,0)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# custom weights initialization called on netG and netD\n",
    "def weights_init(m):\n",
    "    classname = m.__class__.__name__\n",
    "    if classname.find('Conv') != -1:\n",
    "        nn.init.normal_(m.weight.data, 0.0, 0.02)\n",
    "    elif classname.find('BatchNorm') != -1:\n",
    "        nn.init.normal_(m.weight.data, 1.0, 0.02)\n",
    "        nn.init.constant_(m.bias.data, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "class CoupledGenerators(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(CoupledGenerators, self).__init__()\n",
    "\n",
    "        self.init_size = image_size // 4\n",
    "        self.fc = nn.Sequential(nn.Linear(nz, 128 * self.init_size ** 2))\n",
    "\n",
    "        self.shared_conv = nn.Sequential(\n",
    "            nn.BatchNorm2d(128),\n",
    "            nn.Upsample(scale_factor=2),\n",
    "            nn.Conv2d(128, 128, 3, stride=1, padding=1),\n",
    "            nn.BatchNorm2d(128, 0.8),\n",
    "            nn.LeakyReLU(0.2, inplace=True),\n",
    "            nn.Upsample(scale_factor=2),\n",
    "        )\n",
    "        self.G1 = nn.Sequential(\n",
    "            nn.Conv2d(128, 64, 3, stride=1, padding=1),\n",
    "            nn.BatchNorm2d(64, 0.8),\n",
    "            nn.LeakyReLU(0.2, inplace=True),\n",
    "            nn.Conv2d(64, nc, 3, stride=1, padding=1),\n",
    "            nn.Tanh(),\n",
    "        )\n",
    "        self.G2 = nn.Sequential(\n",
    "            nn.Conv2d(128, 64, 3, stride=1, padding=1),\n",
    "            nn.BatchNorm2d(64, 0.8),\n",
    "            nn.LeakyReLU(0.2, inplace=True),\n",
    "            nn.Conv2d(64, nc, 3, stride=1, padding=1),\n",
    "            nn.Tanh(),\n",
    "        )\n",
    "\n",
    "    def forward(self, noise):\n",
    "        out = self.fc(noise)\n",
    "        out = out.view(out.shape[0], 128, self.init_size, self.init_size)\n",
    "        img_emb = self.shared_conv(out)\n",
    "        img1 = self.G1(img_emb)\n",
    "        img2 = self.G2(img_emb)\n",
    "        return img1, img2\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "class CoupledDiscriminators(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(CoupledDiscriminators, self).__init__()\n",
    "\n",
    "        def discriminator_block(in_filters, out_filters, bn=True):\n",
    "            block = [nn.Conv2d(in_filters, out_filters, 3, 2, 1)]\n",
    "            if bn:\n",
    "                block.append(nn.BatchNorm2d(out_filters, 0.8))\n",
    "            block.extend([nn.LeakyReLU(0.2, inplace=True), nn.Dropout2d(0.25)])\n",
    "            return block\n",
    "\n",
    "        self.shared_conv = nn.Sequential(\n",
    "            *discriminator_block(nc, 16, bn=False),\n",
    "            *discriminator_block(16, 32),\n",
    "            *discriminator_block(32, 64),\n",
    "            *discriminator_block(64, 128),\n",
    "        )\n",
    "        # The height and width of downsampled image\n",
    "        ds_size = image_size // 2 ** 4\n",
    "        self.D1 = nn.Linear(128 * ds_size ** 2, 1)\n",
    "        self.D2 = nn.Linear(128 * ds_size ** 2, 1)\n",
    "\n",
    "    def forward(self,img1, img2):\n",
    "        # Determine validity of first image\n",
    "        out = self.shared_conv(img1)\n",
    "        out = out.view(out.shape[0], -1)\n",
    "        validity1 = self.D1(out)\n",
    "        # Determine validity of second image\n",
    "        out = self.shared_conv(img2)\n",
    "        out = out.view(out.shape[0], -1)\n",
    "        validity2 = self.D2(out)\n",
    "\n",
    "        return validity1, validity2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CoupledGenerators(\n",
      "  (fc): Sequential(\n",
      "    (0): Linear(in_features=100, out_features=8192, bias=True)\n",
      "  )\n",
      "  (shared_conv): Sequential(\n",
      "    (0): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (1): Upsample(scale_factor=2.0, mode=nearest)\n",
      "    (2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "    (3): BatchNorm2d(128, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (4): LeakyReLU(negative_slope=0.2, inplace=True)\n",
      "    (5): Upsample(scale_factor=2.0, mode=nearest)\n",
      "  )\n",
      "  (G1): Sequential(\n",
      "    (0): Conv2d(128, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "    (1): BatchNorm2d(64, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): LeakyReLU(negative_slope=0.2, inplace=True)\n",
      "    (3): Conv2d(64, 3, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "    (4): Tanh()\n",
      "  )\n",
      "  (G2): Sequential(\n",
      "    (0): Conv2d(128, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "    (1): BatchNorm2d(64, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): LeakyReLU(negative_slope=0.2, inplace=True)\n",
      "    (3): Conv2d(64, 3, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "    (4): Tanh()\n",
      "  )\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "# Create the generators\n",
    "netG = CoupledGenerators().to(device)\n",
    "\n",
    "# Initialize weights\n",
    "netG.apply(weights_init)\n",
    "\n",
    "# Print the model\n",
    "print(netG)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CoupledDiscriminators(\n",
      "  (shared_conv): Sequential(\n",
      "    (0): Conv2d(3, 16, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
      "    (1): LeakyReLU(negative_slope=0.2, inplace=True)\n",
      "    (2): Dropout2d(p=0.25, inplace=False)\n",
      "    (3): Conv2d(16, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
      "    (4): BatchNorm2d(32, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (5): LeakyReLU(negative_slope=0.2, inplace=True)\n",
      "    (6): Dropout2d(p=0.25, inplace=False)\n",
      "    (7): Conv2d(32, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
      "    (8): BatchNorm2d(64, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (9): LeakyReLU(negative_slope=0.2, inplace=True)\n",
      "    (10): Dropout2d(p=0.25, inplace=False)\n",
      "    (11): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
      "    (12): BatchNorm2d(128, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (13): LeakyReLU(negative_slope=0.2, inplace=True)\n",
      "    (14): Dropout2d(p=0.25, inplace=False)\n",
      "  )\n",
      "  (D1): Linear(in_features=512, out_features=1, bias=True)\n",
      "  (D2): Linear(in_features=512, out_features=1, bias=True)\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "# Create the discriminators\n",
    "netD = CoupledDiscriminators().to(device)\n",
    "\n",
    "# Initialize weights\n",
    "netD.apply(weights_init)\n",
    "\n",
    "# Print the model\n",
    "print(netD)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initialize MSELoss function\n",
    "loss = nn.MSELoss()\n",
    "\n",
    "# Create batch of latent vectors that we will use to visualize\n",
    "#  the progression of the generator\n",
    "#fixed_noise = torch.randn(64, nz, 1, 1, device=device)\n",
    "\n",
    "# Establish convention for real and fake labels during training\n",
    "real_label = 1\n",
    "fake_label = 0\n",
    "\n",
    "# Setup Adam optimizers for both G and D\n",
    "optimizerD = optim.Adam(netD.parameters(), lr=lr, betas=(beta1, 0.999))\n",
    "optimizerG = optim.Adam(netG.parameters(), lr=lr, betas=(beta1, 0.999))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[Epoch 0/3] [Batch 0/20260] [D loss: 0.523270] [G loss: 1.046095]\n",
      "[Epoch 0/3] [Batch 400/20260] [D loss: 0.244941] [G loss: 0.260908]\n",
      "[Epoch 0/3] [Batch 800/20260] [D loss: 0.271521] [G loss: 0.269336]\n",
      "[Epoch 0/3] [Batch 1200/20260] [D loss: 0.274514] [G loss: 0.232115]\n",
      "[Epoch 0/3] [Batch 1600/20260] [D loss: 0.277789] [G loss: 0.222223]\n",
      "[Epoch 0/3] [Batch 2000/20260] [D loss: 0.246303] [G loss: 0.274969]\n",
      "[Epoch 0/3] [Batch 2400/20260] [D loss: 0.267887] [G loss: 0.271612]\n",
      "[Epoch 0/3] [Batch 2800/20260] [D loss: 0.255150] [G loss: 0.247815]\n",
      "[Epoch 0/3] [Batch 3200/20260] [D loss: 0.250294] [G loss: 0.252984]\n",
      "[Epoch 0/3] [Batch 3600/20260] [D loss: 0.238657] [G loss: 0.254664]\n",
      "[Epoch 0/3] [Batch 4000/20260] [D loss: 0.257386] [G loss: 0.258784]\n",
      "[Epoch 0/3] [Batch 4400/20260] [D loss: 0.245828] [G loss: 0.255726]\n",
      "[Epoch 0/3] [Batch 4800/20260] [D loss: 0.251969] [G loss: 0.261540]\n",
      "[Epoch 0/3] [Batch 5200/20260] [D loss: 0.250416] [G loss: 0.272092]\n",
      "[Epoch 0/3] [Batch 5600/20260] [D loss: 0.250739] [G loss: 0.251348]\n",
      "[Epoch 0/3] [Batch 6000/20260] [D loss: 0.250093] [G loss: 0.246989]\n",
      "[Epoch 0/3] [Batch 6400/20260] [D loss: 0.250354] [G loss: 0.242821]\n",
      "[Epoch 0/3] [Batch 6800/20260] [D loss: 0.249783] [G loss: 0.243480]\n",
      "[Epoch 0/3] [Batch 7200/20260] [D loss: 0.255547] [G loss: 0.248354]\n",
      "[Epoch 0/3] [Batch 7600/20260] [D loss: 0.248955] [G loss: 0.252777]\n",
      "[Epoch 0/3] [Batch 8000/20260] [D loss: 0.252276] [G loss: 0.258375]\n",
      "[Epoch 0/3] [Batch 8400/20260] [D loss: 0.246543] [G loss: 0.249620]\n",
      "[Epoch 0/3] [Batch 8800/20260] [D loss: 0.257429] [G loss: 0.250741]\n",
      "[Epoch 0/3] [Batch 9200/20260] [D loss: 0.255129] [G loss: 0.246740]\n",
      "[Epoch 0/3] [Batch 9600/20260] [D loss: 0.251263] [G loss: 0.251014]\n",
      "[Epoch 0/3] [Batch 10000/20260] [D loss: 0.246329] [G loss: 0.249285]\n",
      "[Epoch 0/3] [Batch 10400/20260] [D loss: 0.249802] [G loss: 0.247157]\n",
      "[Epoch 0/3] [Batch 10800/20260] [D loss: 0.252194] [G loss: 0.250972]\n",
      "[Epoch 0/3] [Batch 11200/20260] [D loss: 0.251467] [G loss: 0.252900]\n",
      "[Epoch 0/3] [Batch 11600/20260] [D loss: 0.255467] [G loss: 0.229702]\n",
      "[Epoch 0/3] [Batch 12000/20260] [D loss: 0.249733] [G loss: 0.244604]\n",
      "[Epoch 0/3] [Batch 12400/20260] [D loss: 0.253274] [G loss: 0.251836]\n",
      "[Epoch 0/3] [Batch 12800/20260] [D loss: 0.257509] [G loss: 0.251967]\n",
      "[Epoch 0/3] [Batch 13200/20260] [D loss: 0.250511] [G loss: 0.244152]\n",
      "[Epoch 0/3] [Batch 13600/20260] [D loss: 0.246939] [G loss: 0.242734]\n",
      "[Epoch 0/3] [Batch 14000/20260] [D loss: 0.250971] [G loss: 0.254066]\n",
      "[Epoch 0/3] [Batch 14400/20260] [D loss: 0.247527] [G loss: 0.247693]\n",
      "[Epoch 0/3] [Batch 14800/20260] [D loss: 0.255561] [G loss: 0.250640]\n",
      "[Epoch 0/3] [Batch 15200/20260] [D loss: 0.253605] [G loss: 0.245902]\n",
      "[Epoch 0/3] [Batch 15600/20260] [D loss: 0.253818] [G loss: 0.251785]\n",
      "[Epoch 0/3] [Batch 16000/20260] [D loss: 0.252732] [G loss: 0.247404]\n",
      "[Epoch 0/3] [Batch 16400/20260] [D loss: 0.247863] [G loss: 0.245268]\n",
      "[Epoch 0/3] [Batch 16800/20260] [D loss: 0.249754] [G loss: 0.247780]\n",
      "[Epoch 0/3] [Batch 17200/20260] [D loss: 0.251026] [G loss: 0.253975]\n",
      "[Epoch 0/3] [Batch 17600/20260] [D loss: 0.244337] [G loss: 0.264587]\n",
      "[Epoch 0/3] [Batch 18000/20260] [D loss: 0.239170] [G loss: 0.239773]\n",
      "[Epoch 0/3] [Batch 18400/20260] [D loss: 0.235615] [G loss: 0.253790]\n",
      "[Epoch 0/3] [Batch 18800/20260] [D loss: 0.243399] [G loss: 0.246778]\n",
      "[Epoch 0/3] [Batch 19200/20260] [D loss: 0.252044] [G loss: 0.230605]\n",
      "[Epoch 0/3] [Batch 19600/20260] [D loss: 0.247933] [G loss: 0.258748]\n",
      "[Epoch 0/3] [Batch 20000/20260] [D loss: 0.249462] [G loss: 0.251511]\n",
      "[Epoch 1/3] [Batch 140/20260] [D loss: 0.252192] [G loss: 0.260783]\n",
      "[Epoch 1/3] [Batch 540/20260] [D loss: 0.244837] [G loss: 0.262041]\n",
      "[Epoch 1/3] [Batch 940/20260] [D loss: 0.246427] [G loss: 0.254119]\n",
      "[Epoch 1/3] [Batch 1340/20260] [D loss: 0.254045] [G loss: 0.251052]\n",
      "[Epoch 1/3] [Batch 1740/20260] [D loss: 0.246977] [G loss: 0.239625]\n",
      "[Epoch 1/3] [Batch 2140/20260] [D loss: 0.246634] [G loss: 0.255501]\n",
      "[Epoch 1/3] [Batch 2540/20260] [D loss: 0.245606] [G loss: 0.256650]\n",
      "[Epoch 1/3] [Batch 2940/20260] [D loss: 0.246576] [G loss: 0.241386]\n",
      "[Epoch 1/3] [Batch 3340/20260] [D loss: 0.244912] [G loss: 0.242888]\n",
      "[Epoch 1/3] [Batch 3740/20260] [D loss: 0.257376] [G loss: 0.253041]\n",
      "[Epoch 1/3] [Batch 4140/20260] [D loss: 0.256699] [G loss: 0.241293]\n",
      "[Epoch 1/3] [Batch 4540/20260] [D loss: 0.250901] [G loss: 0.269736]\n",
      "[Epoch 1/3] [Batch 4940/20260] [D loss: 0.251838] [G loss: 0.246831]\n",
      "[Epoch 1/3] [Batch 5340/20260] [D loss: 0.238414] [G loss: 0.271473]\n",
      "[Epoch 1/3] [Batch 5740/20260] [D loss: 0.253462] [G loss: 0.272276]\n",
      "[Epoch 1/3] [Batch 6140/20260] [D loss: 0.253302] [G loss: 0.237913]\n",
      "[Epoch 1/3] [Batch 6540/20260] [D loss: 0.256807] [G loss: 0.247472]\n",
      "[Epoch 1/3] [Batch 6940/20260] [D loss: 0.240119] [G loss: 0.275357]\n",
      "[Epoch 1/3] [Batch 7340/20260] [D loss: 0.242410] [G loss: 0.263111]\n",
      "[Epoch 1/3] [Batch 7740/20260] [D loss: 0.249585] [G loss: 0.257870]\n",
      "[Epoch 1/3] [Batch 8140/20260] [D loss: 0.252979] [G loss: 0.250890]\n",
      "[Epoch 1/3] [Batch 8540/20260] [D loss: 0.247991] [G loss: 0.250867]\n",
      "[Epoch 1/3] [Batch 8940/20260] [D loss: 0.280231] [G loss: 0.242820]\n",
      "[Epoch 1/3] [Batch 9340/20260] [D loss: 0.279891] [G loss: 0.255414]\n",
      "[Epoch 1/3] [Batch 9740/20260] [D loss: 0.239548] [G loss: 0.257229]\n",
      "[Epoch 1/3] [Batch 10140/20260] [D loss: 0.261036] [G loss: 0.255641]\n",
      "[Epoch 1/3] [Batch 10540/20260] [D loss: 0.261432] [G loss: 0.220271]\n",
      "[Epoch 1/3] [Batch 10940/20260] [D loss: 0.238238] [G loss: 0.252218]\n",
      "[Epoch 1/3] [Batch 11340/20260] [D loss: 0.243377] [G loss: 0.251853]\n",
      "[Epoch 1/3] [Batch 11740/20260] [D loss: 0.246365] [G loss: 0.244079]\n",
      "[Epoch 1/3] [Batch 12140/20260] [D loss: 0.241897] [G loss: 0.272316]\n",
      "[Epoch 1/3] [Batch 12540/20260] [D loss: 0.269305] [G loss: 0.248010]\n",
      "[Epoch 1/3] [Batch 12940/20260] [D loss: 0.238824] [G loss: 0.275291]\n",
      "[Epoch 1/3] [Batch 13340/20260] [D loss: 0.230337] [G loss: 0.239493]\n",
      "[Epoch 1/3] [Batch 13740/20260] [D loss: 0.242329] [G loss: 0.245154]\n",
      "[Epoch 1/3] [Batch 14140/20260] [D loss: 0.253160] [G loss: 0.257031]\n",
      "[Epoch 1/3] [Batch 14540/20260] [D loss: 0.247339] [G loss: 0.244611]\n",
      "[Epoch 1/3] [Batch 14940/20260] [D loss: 0.262525] [G loss: 0.258924]\n",
      "[Epoch 1/3] [Batch 15340/20260] [D loss: 0.254089] [G loss: 0.240248]\n",
      "[Epoch 1/3] [Batch 15740/20260] [D loss: 0.247446] [G loss: 0.255714]\n",
      "[Epoch 1/3] [Batch 16140/20260] [D loss: 0.251804] [G loss: 0.256263]\n",
      "[Epoch 1/3] [Batch 16540/20260] [D loss: 0.243318] [G loss: 0.239034]\n",
      "[Epoch 1/3] [Batch 16940/20260] [D loss: 0.253238] [G loss: 0.285180]\n",
      "[Epoch 1/3] [Batch 17340/20260] [D loss: 0.254929] [G loss: 0.243144]\n",
      "[Epoch 1/3] [Batch 17740/20260] [D loss: 0.257281] [G loss: 0.272135]\n",
      "[Epoch 1/3] [Batch 18140/20260] [D loss: 0.234422] [G loss: 0.288303]\n",
      "[Epoch 1/3] [Batch 18540/20260] [D loss: 0.238808] [G loss: 0.238137]\n",
      "[Epoch 1/3] [Batch 18940/20260] [D loss: 0.251698] [G loss: 0.263418]\n",
      "[Epoch 1/3] [Batch 19340/20260] [D loss: 0.261644] [G loss: 0.261620]\n",
      "[Epoch 1/3] [Batch 19740/20260] [D loss: 0.229976] [G loss: 0.266421]\n",
      "[Epoch 1/3] [Batch 20140/20260] [D loss: 0.244688] [G loss: 0.239871]\n",
      "[Epoch 2/3] [Batch 280/20260] [D loss: 0.252104] [G loss: 0.240195]\n",
      "[Epoch 2/3] [Batch 680/20260] [D loss: 0.246840] [G loss: 0.283135]\n",
      "[Epoch 2/3] [Batch 1080/20260] [D loss: 0.223125] [G loss: 0.258887]\n",
      "[Epoch 2/3] [Batch 1480/20260] [D loss: 0.244745] [G loss: 0.254244]\n",
      "[Epoch 2/3] [Batch 1880/20260] [D loss: 0.218999] [G loss: 0.257234]\n",
      "[Epoch 2/3] [Batch 2280/20260] [D loss: 0.220925] [G loss: 0.250953]\n",
      "[Epoch 2/3] [Batch 2680/20260] [D loss: 0.241198] [G loss: 0.240503]\n",
      "[Epoch 2/3] [Batch 3080/20260] [D loss: 0.281247] [G loss: 0.197886]\n",
      "[Epoch 2/3] [Batch 3480/20260] [D loss: 0.251395] [G loss: 0.277131]\n",
      "[Epoch 2/3] [Batch 3880/20260] [D loss: 0.240743] [G loss: 0.233054]\n",
      "[Epoch 2/3] [Batch 4280/20260] [D loss: 0.267198] [G loss: 0.245151]\n",
      "[Epoch 2/3] [Batch 4680/20260] [D loss: 0.249429] [G loss: 0.274396]\n",
      "[Epoch 2/3] [Batch 5080/20260] [D loss: 0.240691] [G loss: 0.295580]\n",
      "[Epoch 2/3] [Batch 5480/20260] [D loss: 0.247490] [G loss: 0.255301]\n",
      "[Epoch 2/3] [Batch 5880/20260] [D loss: 0.271155] [G loss: 0.257575]\n",
      "[Epoch 2/3] [Batch 6280/20260] [D loss: 0.235728] [G loss: 0.241953]\n",
      "[Epoch 2/3] [Batch 6680/20260] [D loss: 0.255893] [G loss: 0.229624]\n",
      "[Epoch 2/3] [Batch 7080/20260] [D loss: 0.221834] [G loss: 0.275896]\n",
      "[Epoch 2/3] [Batch 7480/20260] [D loss: 0.243157] [G loss: 0.349707]\n",
      "[Epoch 2/3] [Batch 7880/20260] [D loss: 0.238941] [G loss: 0.291050]\n",
      "[Epoch 2/3] [Batch 8280/20260] [D loss: 0.254335] [G loss: 0.233099]\n",
      "[Epoch 2/3] [Batch 8680/20260] [D loss: 0.241453] [G loss: 0.291545]\n",
      "[Epoch 2/3] [Batch 9080/20260] [D loss: 0.264004] [G loss: 0.289830]\n",
      "[Epoch 2/3] [Batch 9480/20260] [D loss: 0.258928] [G loss: 0.226303]\n",
      "[Epoch 2/3] [Batch 9880/20260] [D loss: 0.276069] [G loss: 0.270794]\n",
      "[Epoch 2/3] [Batch 10280/20260] [D loss: 0.248386] [G loss: 0.203056]\n",
      "[Epoch 2/3] [Batch 10680/20260] [D loss: 0.242969] [G loss: 0.318646]\n",
      "[Epoch 2/3] [Batch 11080/20260] [D loss: 0.245089] [G loss: 0.218827]\n",
      "[Epoch 2/3] [Batch 11480/20260] [D loss: 0.220807] [G loss: 0.209483]\n",
      "[Epoch 2/3] [Batch 11880/20260] [D loss: 0.225127] [G loss: 0.287033]\n",
      "[Epoch 2/3] [Batch 12280/20260] [D loss: 0.279839] [G loss: 0.330409]\n",
      "[Epoch 2/3] [Batch 12680/20260] [D loss: 0.233283] [G loss: 0.316587]\n",
      "[Epoch 2/3] [Batch 13080/20260] [D loss: 0.274702] [G loss: 0.250234]\n",
      "[Epoch 2/3] [Batch 13480/20260] [D loss: 0.240351] [G loss: 0.304883]\n",
      "[Epoch 2/3] [Batch 13880/20260] [D loss: 0.258551] [G loss: 0.241908]\n",
      "[Epoch 2/3] [Batch 14280/20260] [D loss: 0.276631] [G loss: 0.259602]\n",
      "[Epoch 2/3] [Batch 14680/20260] [D loss: 0.256109] [G loss: 0.230705]\n",
      "[Epoch 2/3] [Batch 15080/20260] [D loss: 0.265805] [G loss: 0.206260]\n",
      "[Epoch 2/3] [Batch 15480/20260] [D loss: 0.243903] [G loss: 0.302223]\n",
      "[Epoch 2/3] [Batch 15880/20260] [D loss: 0.263563] [G loss: 0.223525]\n",
      "[Epoch 2/3] [Batch 16280/20260] [D loss: 0.255707] [G loss: 0.228787]\n",
      "[Epoch 2/3] [Batch 16680/20260] [D loss: 0.252477] [G loss: 0.236425]\n",
      "[Epoch 2/3] [Batch 17080/20260] [D loss: 0.232401] [G loss: 0.311506]\n",
      "[Epoch 2/3] [Batch 17480/20260] [D loss: 0.269204] [G loss: 0.324897]\n",
      "[Epoch 2/3] [Batch 17880/20260] [D loss: 0.247303] [G loss: 0.250830]\n",
      "[Epoch 2/3] [Batch 18280/20260] [D loss: 0.241761] [G loss: 0.307265]\n",
      "[Epoch 2/3] [Batch 18680/20260] [D loss: 0.255416] [G loss: 0.272673]\n",
      "[Epoch 2/3] [Batch 19080/20260] [D loss: 0.259371] [G loss: 0.297096]\n",
      "[Epoch 2/3] [Batch 19480/20260] [D loss: 0.219074] [G loss: 0.237736]\n",
      "[Epoch 2/3] [Batch 19880/20260] [D loss: 0.235614] [G loss: 0.280722]\n"
     ]
    }
   ],
   "source": [
    "Tensor = torch.cuda.FloatTensor\n",
    "for epoch in range(num_epochs):\n",
    "    for i, ((imgs1, _), (imgs2, _)) in enumerate(zip(dataloader1, dataloader2)):\n",
    "\n",
    "        batch_size = imgs1.shape[0]\n",
    "\n",
    "        # Adversarial ground truths\n",
    "        valid = Variable(Tensor(batch_size, 1).fill_(1.0), requires_grad=False)\n",
    "        fake = Variable(Tensor(batch_size, 1).fill_(0.0), requires_grad=False)\n",
    "\n",
    "        # Configure input\n",
    "        imgs1 = Variable(imgs1.type(Tensor).expand(imgs1.size(0), 3, image_size, image_size))\n",
    "        imgs2 = Variable(imgs2.type(Tensor))\n",
    "\n",
    "        # ------------------\n",
    "        #  Train Generators\n",
    "        # ------------------\n",
    "\n",
    "        optimizerG.zero_grad()\n",
    "\n",
    "        # Sample noise as generator input\n",
    "        z = Variable(Tensor(np.random.normal(0, 1, (batch_size, nz))))\n",
    "\n",
    "        # Generate a batch of images\n",
    "        gen_imgs1, gen_imgs2 = netG(z)\n",
    "        # Determine validity of generated images\n",
    "        validity1, validity2 = netD(gen_imgs1, gen_imgs2)\n",
    "\n",
    "        g_loss = (loss(validity1, valid) + loss(validity2, valid)) / 2\n",
    "\n",
    "        g_loss.backward()\n",
    "        optimizerG.step()\n",
    "\n",
    "        # ----------------------\n",
    "        #  Train Discriminators\n",
    "        # ----------------------\n",
    "\n",
    "        optimizerD.zero_grad()\n",
    "\n",
    "        # Determine validity of real and generated images\n",
    "        validity1_real, validity2_real = netD(imgs1, imgs2)\n",
    "#         validity1_fake, validity2_fake = netG(gen_imgs1.detach())\n",
    "#         validity11_fake, validity22_fake = netG(gen_imgs2.detach())\n",
    "\n",
    "        validity1_fake, validity2_fake = netD(gen_imgs1.detach(), gen_imgs2.detach())\n",
    "\n",
    "        d_loss = (\n",
    "            loss(validity1_real, valid)\n",
    "            + loss(validity1_fake, fake)\n",
    "            + loss(validity2_real, valid)\n",
    "            + loss(validity2_fake, fake)\n",
    "        ) / 4\n",
    "\n",
    "        d_loss.backward()\n",
    "        optimizerD.step()\n",
    "\n",
    "        \n",
    "\n",
    "        batches_done = epoch * len(dataloader1) + i\n",
    "        if batches_done % sample_interval == 0:\n",
    "            gen_imgs = torch.cat((gen_imgs1.data, gen_imgs2.data), 0)\n",
    "            save_image(gen_imgs, \"images/%d.png\" % batches_done, nrow=8, normalize=True)\n",
    "            print(\n",
    "            \"[Epoch %d/%d] [Batch %d/%d] [D loss: %f] [G loss: %f]\"\n",
    "            % (epoch, num_epochs, i, len(dataloader1), d_loss.item(), g_loss.item())\n",
    "            )"
   ]
  },
  {
   "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}