DeepLearning/GAN/MNIST/train_MNIST_DCGAN.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import time\n",
    "import torch\n",
    "from tqdm import tqdm\n",
    "from torch import nn, optim\n",
    "from torch.utils.data import DataLoader\n",
    "from torchvision import datasets\n",
    "from torchvision import transforms\n",
    "from torchvision.utils import save_image\n",
    "import sys \n",
    "from pathlib import Path\n",
    "sys.path.append(os.path.abspath(r\"../..\"))\n",
    "from model.generator import Generator, Generator_Transpose\n",
    "from model.discriminator import DiscriminatorResnet, DiscriminatorLinear, DiscriminatorConv\n",
    "from utils_.utils import weights_init, weight_init\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "from PIL import Image"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. 加载和预处理MNIST数据"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((60000, 784), (60000,))"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from mlxtend.data import loadlocal_mnist\n",
    "train_data_path = \"../../../../data/MNIST/train-images.idx3-ubyte\"\n",
    "train_label_path = \"../../../../data/MNIST/train-labels.idx1-ubyte\"\n",
    "test_data_path = \"../../../../data/MNIST/t10k-images.idx3-ubyte\"\n",
    "test_label_path = \"../../../../data/MNIST/t10k-labels.idx1-ubyte\"\n",
    "\n",
    "train_data,train_label = loadlocal_mnist(\n",
    "    images_path = train_data_path,\n",
    "    labels_path = train_label_path\n",
    ")\n",
    "train_data.shape,train_label.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 900x900 with 9 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "img,ax = plt.subplots(3,3,figsize=(9,9))\n",
    "plt.subplots_adjust(hspace=0.4,wspace=0.4)\n",
    "for i in range(3):\n",
    "    for j in range(3):\n",
    "        num = np.random.randint(0,train_label.shape[0])\n",
    "        ax[i][j].imshow(train_data[num].reshape((28,28)),cmap=\"gray\")\n",
    "        ax[i][j].set_title(train_label[num],fontdict={\"fontsize\":20})\n",
    "plt.show()\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Dataset and Model parameter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "input_size = [1, 28, 28]\n",
    "batch_size = 128\n",
    "Epoch = 1000\n",
    "GenEpoch = 1\n",
    "in_channel = 64"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch.utils.data import Dataset,DataLoader\n",
    "import numpy as np \n",
    "from mlxtend.data import loadlocal_mnist\n",
    "import torchvision.transforms as transforms\n",
    "\n",
    "class MNIST_Dataset(Dataset):\n",
    "    def __init__(self,train_data_path,train_label_path,transform=None):\n",
    "        train_data,train_label = loadlocal_mnist(\n",
    "            images_path = train_data_path,\n",
    "            labels_path = train_label_path\n",
    "            )\n",
    "        self.train_data = train_data\n",
    "        self.train_label = train_label.reshape(-1)\n",
    "        self.transform=transform\n",
    "        \n",
    "    def __len__(self):\n",
    "        return self.train_label.shape[0] \n",
    "    \n",
    "    def __getitem__(self,index):\n",
    "        if torch.is_tensor(index):\n",
    "            index = index.tolist()\n",
    "        images = self.train_data[index,:].reshape((28,28))\n",
    "        labels = self.train_label[index]\n",
    "        if self.transform:\n",
    "            images = self.transform(images)\n",
    "        return images,labels\n",
    "\n",
    "train_data_path = \"../../../../data/MNIST/train-images.idx3-ubyte\"\n",
    "train_label_path = \"../../../../data/MNIST/train-labels.idx1-ubyte\"\n",
    "transform_dataset =transforms.Compose([\n",
    "    transforms.ToTensor()]\n",
    ")\n",
    "MNIST_dataset = MNIST_Dataset(train_data_path=train_data_path,\n",
    "                               train_label_path=train_label_path,\n",
    "                               transform=transform_dataset)  \n",
    "MNIST_dataloader = DataLoader(dataset=MNIST_dataset,\n",
    "                              batch_size=batch_size,\n",
    "                              shuffle=True,drop_last=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 900x900 with 9 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "img,ax = plt.subplots(3,3,figsize=(9,9))\n",
    "plt.subplots_adjust(hspace=0.4,wspace=0.4)\n",
    "for i in range(3):\n",
    "    for j in range(3):\n",
    "        num = np.random.randint(0,train_label.shape[0])\n",
    "        ax[i][j].imshow(MNIST_dataset[num][0].reshape((28,28)),cmap=\"gray\")\n",
    "        ax[i][j].set_title(MNIST_dataset[num][1],fontdict={\"fontsize\":20})\n",
    "plt.show()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Result save path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "time_now = time.strftime('DCGAN_%Y-%m-%d-%H_%M_%S', time.localtime(time.time()))\n",
    "log_path = f'./log/{time_now}'\n",
    "os.makedirs(log_path)\n",
    "os.makedirs(f'{log_path}/image')\n",
    "os.makedirs(f'{log_path}/image/image_all')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "using device: cuda\n"
     ]
    }
   ],
   "source": [
    "device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
    "print(f'using device: {device}')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Model define"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Discriminator(\n",
       "   (dis): Sequential(\n",
       "     (0): Conv2d(1, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "     (1): LeakyReLU(negative_slope=0.2, inplace=True)\n",
       "     (2): Conv2d(64, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "     (3): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "     (4): LeakyReLU(negative_slope=0.2, inplace=True)\n",
       "     (5): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), bias=False)\n",
       "     (6): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "     (7): LeakyReLU(negative_slope=0.2, inplace=True)\n",
       "     (8): Conv2d(256, 1, kernel_size=(1, 1), stride=(3, 3), bias=False)\n",
       "     (9): Sigmoid()\n",
       "   )\n",
       " ),\n",
       " Generator(\n",
       "   (gen): Sequential(\n",
       "     (0): ConvTranspose2d(64, 1024, kernel_size=(3, 3), stride=(1, 1), bias=False)\n",
       "     (1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "     (2): ReLU(inplace=True)\n",
       "     (3): ConvTranspose2d(1024, 512, kernel_size=(3, 3), stride=(2, 2), bias=False)\n",
       "     (4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "     (5): ReLU(inplace=True)\n",
       "     (6): ConvTranspose2d(512, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "     (7): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "     (8): ReLU(inplace=True)\n",
       "     (9): ConvTranspose2d(256, 1, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "     (10): Tanh()\n",
       "   )\n",
       " ))"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import torch\n",
    "from torch import nn \n",
    "\n",
    "def _weight_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)\n",
    "        \n",
    "\n",
    "class Discriminator(nn.Module):\n",
    "    def __init__(self,input_size,inplace=True):\n",
    "        super(Discriminator,self).__init__()\n",
    "        c,h,w = input_size\n",
    "        self.dis = nn.Sequential(\n",
    "            # b,1,28,28 -> b,64,14,14\n",
    "            nn.Conv2d(c,64,(4,4),2,1,bias=False),\n",
    "            nn.LeakyReLU(0.2,inplace=True),\n",
    "            \n",
    "            # b,64,14,14 -> b,128,7,7\n",
    "            nn.Conv2d(64,128,(4,4),2,1,bias=False),\n",
    "            nn.BatchNorm2d(128),\n",
    "            nn.LeakyReLU(0.2,inplace=True),\n",
    "            \n",
    "            # b,128,7,7 -> b,256,3,3\n",
    "            nn.Conv2d(128,256,(3,3),2,0,bias=False),\n",
    "            nn.BatchNorm2d(256),\n",
    "            nn.LeakyReLU(0.2,inplace=True),\n",
    "            \n",
    "            # b,128,3,3 -> b,1,1,1\n",
    "            nn.Conv2d(256,1,1,3,0,bias=False),\n",
    "            nn.Sigmoid()\n",
    "        )\n",
    "        \n",
    "        \n",
    "    def forward(self,x):\n",
    "        x = self.dis(x)\n",
    "        x = x.view(-1)\n",
    "        return x \n",
    "\n",
    "class Generator(nn.Module):\n",
    "    def __init__(self,in_channel):\n",
    "        super(Generator,self).__init__() # 调用父类的构造方法\n",
    "        self.gen = nn.Sequential(\n",
    "            nn.ConvTranspose2d(in_channel,1024,(3,3),1,0,bias=False),\n",
    "            nn.BatchNorm2d(1024),\n",
    "            nn.ReLU(True),\n",
    "            \n",
    "            nn.ConvTranspose2d(1024,512,(3,3),2,0,bias=False),\n",
    "            nn.BatchNorm2d(512),\n",
    "            nn.ReLU(True),\n",
    "            \n",
    "            nn.ConvTranspose2d(512,256,(4,4),2,1,bias=False),\n",
    "            nn.BatchNorm2d(256),\n",
    "            nn.ReLU(True),\n",
    "            \n",
    "            nn.ConvTranspose2d(256,1,(4,4),2,1,bias=False),\n",
    "            nn.Tanh()            \n",
    "        )\n",
    "\n",
    "    def forward(self,x):\n",
    "        res = self.gen(x)\n",
    "        return res.view(x.size()[0],1,28,28)\n",
    "\n",
    "D = Discriminator(input_size=input_size)\n",
    "G = Generator(in_channel=in_channel)\n",
    "D.apply(_weight_init)\n",
    "G.apply(_weight_init)\n",
    "D.to(device)\n",
    "G.to(device)\n",
    "D,G"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 6. Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 1/1000: 100%|██████████| 469/469 [00:25<00:00, 18.28it/s, dis-loss=0.195, gen-loss=2.76]\n",
      "Epoch 2/1000: 100%|██████████| 469/469 [00:24<00:00, 19.39it/s, dis-loss=0.133, gen-loss=3.38] \n",
      "Epoch 3/1000: 100%|██████████| 469/469 [00:24<00:00, 18.78it/s, dis-loss=0.135, gen-loss=3.44]\n",
      "Epoch 4/1000: 100%|██████████| 469/469 [00:25<00:00, 18.36it/s, dis-loss=0.144, gen-loss=3.39]\n",
      "Epoch 5/1000: 100%|██████████| 469/469 [00:24<00:00, 18.77it/s, dis-loss=0.201, gen-loss=2.97]\n",
      "Epoch 6/1000: 100%|██████████| 469/469 [00:24<00:00, 18.80it/s, dis-loss=0.201, gen-loss=2.9] \n",
      "Epoch 7/1000: 100%|██████████| 469/469 [00:24<00:00, 18.83it/s, dis-loss=0.235, gen-loss=2.64]\n",
      "Epoch 8/1000: 100%|██████████| 469/469 [00:24<00:00, 19.33it/s, dis-loss=0.25, gen-loss=2.53] \n",
      "Epoch 9/1000: 100%|██████████| 469/469 [00:24<00:00, 19.18it/s, dis-loss=0.248, gen-loss=2.57]\n",
      "Epoch 10/1000: 100%|██████████| 469/469 [00:23<00:00, 19.90it/s, dis-loss=0.26, gen-loss=2.47] \n",
      "Epoch 11/1000: 100%|██████████| 469/469 [00:23<00:00, 19.62it/s, dis-loss=0.256, gen-loss=2.57]\n",
      "Epoch 12/1000: 100%|██████████| 469/469 [00:23<00:00, 19.69it/s, dis-loss=0.244, gen-loss=2.76]\n",
      "Epoch 13/1000: 100%|██████████| 469/469 [00:23<00:00, 19.96it/s, dis-loss=0.302, gen-loss=2.31]\n",
      "Epoch 14/1000: 100%|██████████| 469/469 [00:23<00:00, 20.03it/s, dis-loss=0.234, gen-loss=2.7] \n",
      "Epoch 15/1000: 100%|██████████| 469/469 [00:23<00:00, 19.97it/s, dis-loss=0.278, gen-loss=2.4] \n",
      "Epoch 16/1000: 100%|██████████| 469/469 [00:23<00:00, 19.85it/s, dis-loss=0.336, gen-loss=2.1] \n",
      "Epoch 17/1000: 100%|██████████| 469/469 [00:23<00:00, 20.12it/s, dis-loss=0.349, gen-loss=1.98]\n",
      "Epoch 18/1000: 100%|██████████| 469/469 [00:22<00:00, 20.44it/s, dis-loss=0.315, gen-loss=2.09]\n",
      "Epoch 19/1000: 100%|██████████| 469/469 [00:23<00:00, 19.96it/s, dis-loss=0.389, gen-loss=1.73]\n",
      "Epoch 20/1000: 100%|██████████| 469/469 [00:23<00:00, 20.24it/s, dis-loss=0.394, gen-loss=1.69]\n",
      "Epoch 21/1000: 100%|██████████| 469/469 [00:23<00:00, 19.61it/s, dis-loss=0.396, gen-loss=1.66]\n",
      "Epoch 22/1000: 100%|██████████| 469/469 [00:24<00:00, 19.46it/s, dis-loss=0.343, gen-loss=1.98]\n",
      "Epoch 23/1000: 100%|██████████| 469/469 [00:24<00:00, 18.84it/s, dis-loss=0.41, gen-loss=1.63] \n",
      "Epoch 24/1000: 100%|██████████| 469/469 [00:23<00:00, 20.01it/s, dis-loss=0.412, gen-loss=1.63]\n",
      "Epoch 25/1000: 100%|██████████| 469/469 [00:22<00:00, 20.44it/s, dis-loss=0.403, gen-loss=1.67]\n",
      "Epoch 26/1000: 100%|██████████| 469/469 [00:23<00:00, 20.05it/s, dis-loss=0.388, gen-loss=1.76]\n",
      "Epoch 27/1000: 100%|██████████| 469/469 [00:23<00:00, 20.37it/s, dis-loss=0.386, gen-loss=1.75]\n",
      "Epoch 28/1000: 100%|██████████| 469/469 [00:22<00:00, 20.50it/s, dis-loss=0.438, gen-loss=1.57]\n",
      "Epoch 29/1000: 100%|██████████| 469/469 [00:22<00:00, 20.47it/s, dis-loss=0.445, gen-loss=1.54]\n",
      "Epoch 30/1000: 100%|██████████| 469/469 [00:23<00:00, 19.98it/s, dis-loss=0.441, gen-loss=1.53]\n",
      "Epoch 31/1000: 100%|██████████| 469/469 [00:23<00:00, 20.23it/s, dis-loss=0.439, gen-loss=1.54]\n",
      "Epoch 32/1000: 100%|██████████| 469/469 [00:23<00:00, 20.33it/s, dis-loss=0.447, gen-loss=1.53]\n",
      "Epoch 33/1000: 100%|██████████| 469/469 [00:23<00:00, 20.22it/s, dis-loss=0.448, gen-loss=1.52]\n",
      "Epoch 34/1000: 100%|██████████| 469/469 [00:23<00:00, 19.92it/s, dis-loss=0.448, gen-loss=1.51]\n",
      "Epoch 35/1000: 100%|██████████| 469/469 [00:22<00:00, 20.57it/s, dis-loss=0.446, gen-loss=1.51]\n",
      "Epoch 36/1000: 100%|██████████| 469/469 [00:23<00:00, 19.94it/s, dis-loss=0.457, gen-loss=1.48]\n",
      "Epoch 37/1000: 100%|██████████| 469/469 [00:23<00:00, 19.81it/s, dis-loss=0.459, gen-loss=1.48]\n",
      "Epoch 38/1000: 100%|██████████| 469/469 [00:23<00:00, 20.24it/s, dis-loss=0.464, gen-loss=1.47]\n",
      "Epoch 39/1000: 100%|██████████| 469/469 [00:23<00:00, 20.20it/s, dis-loss=0.466, gen-loss=1.46]\n",
      "Epoch 40/1000: 100%|██████████| 469/469 [00:23<00:00, 20.19it/s, dis-loss=0.466, gen-loss=1.46]\n",
      "Epoch 41/1000: 100%|██████████| 469/469 [00:23<00:00, 20.32it/s, dis-loss=0.473, gen-loss=1.43]\n",
      "Epoch 42/1000: 100%|██████████| 469/469 [00:23<00:00, 20.30it/s, dis-loss=0.473, gen-loss=1.43]\n",
      "Epoch 43/1000: 100%|██████████| 469/469 [00:23<00:00, 20.27it/s, dis-loss=0.476, gen-loss=1.44]\n",
      "Epoch 44/1000: 100%|██████████| 469/469 [00:23<00:00, 20.30it/s, dis-loss=0.475, gen-loss=1.43]\n",
      "Epoch 45/1000: 100%|██████████| 469/469 [00:22<00:00, 20.55it/s, dis-loss=0.477, gen-loss=1.42]\n",
      "Epoch 46/1000: 100%|██████████| 469/469 [00:22<00:00, 20.49it/s, dis-loss=0.479, gen-loss=1.42]\n",
      "Epoch 47/1000: 100%|██████████| 469/469 [00:23<00:00, 20.36it/s, dis-loss=0.484, gen-loss=1.41]\n",
      "Epoch 48/1000: 100%|██████████| 469/469 [00:22<00:00, 20.49it/s, dis-loss=0.484, gen-loss=1.4] \n",
      "Epoch 49/1000: 100%|██████████| 469/469 [00:23<00:00, 20.37it/s, dis-loss=0.484, gen-loss=1.41]\n",
      "Epoch 50/1000: 100%|██████████| 469/469 [00:23<00:00, 20.32it/s, dis-loss=0.484, gen-loss=1.41]\n",
      "Epoch 51/1000: 100%|██████████| 469/469 [00:22<00:00, 20.49it/s, dis-loss=0.495, gen-loss=1.38]\n",
      "Epoch 52/1000: 100%|██████████| 469/469 [00:22<00:00, 20.46it/s, dis-loss=0.486, gen-loss=1.39]\n",
      "Epoch 53/1000: 100%|██████████| 469/469 [00:22<00:00, 20.51it/s, dis-loss=0.496, gen-loss=1.38]\n",
      "Epoch 54/1000: 100%|██████████| 469/469 [00:22<00:00, 20.42it/s, dis-loss=0.488, gen-loss=1.38]\n",
      "Epoch 55/1000: 100%|██████████| 469/469 [00:22<00:00, 20.48it/s, dis-loss=0.495, gen-loss=1.38]\n",
      "Epoch 56/1000: 100%|██████████| 469/469 [00:23<00:00, 19.63it/s, dis-loss=0.489, gen-loss=1.38]\n",
      "Epoch 57/1000: 100%|██████████| 469/469 [00:24<00:00, 19.47it/s, dis-loss=0.498, gen-loss=1.37]\n",
      "Epoch 58/1000: 100%|██████████| 469/469 [00:22<00:00, 20.48it/s, dis-loss=0.492, gen-loss=1.37]\n",
      "Epoch 59/1000: 100%|██████████| 469/469 [00:23<00:00, 19.77it/s, dis-loss=0.493, gen-loss=1.38]\n",
      "Epoch 60/1000: 100%|██████████| 469/469 [00:23<00:00, 20.16it/s, dis-loss=0.491, gen-loss=1.37]\n",
      "Epoch 61/1000: 100%|██████████| 469/469 [00:23<00:00, 20.22it/s, dis-loss=0.501, gen-loss=1.38]\n",
      "Epoch 62/1000: 100%|██████████| 469/469 [00:23<00:00, 20.12it/s, dis-loss=0.491, gen-loss=1.37]\n",
      "Epoch 63/1000: 100%|██████████| 469/469 [00:23<00:00, 20.08it/s, dis-loss=0.497, gen-loss=1.37]\n",
      "Epoch 64/1000: 100%|██████████| 469/469 [00:23<00:00, 19.89it/s, dis-loss=0.484, gen-loss=1.38]\n",
      "Epoch 65/1000: 100%|██████████| 469/469 [00:23<00:00, 20.04it/s, dis-loss=0.494, gen-loss=1.37]\n",
      "Epoch 66/1000: 100%|██████████| 469/469 [00:24<00:00, 19.34it/s, dis-loss=0.495, gen-loss=1.37]\n",
      "Epoch 67/1000: 100%|██████████| 469/469 [00:23<00:00, 19.66it/s, dis-loss=0.491, gen-loss=1.38]\n",
      "Epoch 68/1000: 100%|██████████| 469/469 [00:23<00:00, 20.05it/s, dis-loss=0.493, gen-loss=1.37]\n",
      "Epoch 69/1000: 100%|██████████| 469/469 [00:23<00:00, 19.98it/s, dis-loss=0.494, gen-loss=1.37]\n",
      "Epoch 70/1000: 100%|██████████| 469/469 [00:24<00:00, 18.99it/s, dis-loss=0.498, gen-loss=1.37]\n",
      "Epoch 71/1000: 100%|██████████| 469/469 [00:24<00:00, 19.35it/s, dis-loss=0.488, gen-loss=1.38]\n",
      "Epoch 72/1000: 100%|██████████| 469/469 [00:23<00:00, 19.87it/s, dis-loss=0.497, gen-loss=1.37]\n",
      "Epoch 73/1000: 100%|██████████| 469/469 [00:23<00:00, 20.31it/s, dis-loss=0.488, gen-loss=1.37]\n",
      "Epoch 74/1000: 100%|██████████| 469/469 [00:23<00:00, 20.26it/s, dis-loss=0.492, gen-loss=1.38]\n",
      "Epoch 75/1000: 100%|██████████| 469/469 [00:24<00:00, 19.38it/s, dis-loss=0.488, gen-loss=1.38]\n",
      "Epoch 76/1000: 100%|██████████| 469/469 [00:24<00:00, 18.79it/s, dis-loss=0.483, gen-loss=1.38]\n",
      "Epoch 77/1000: 100%|██████████| 469/469 [00:22<00:00, 20.44it/s, dis-loss=0.484, gen-loss=1.4] \n",
      "Epoch 78/1000: 100%|██████████| 469/469 [00:23<00:00, 20.20it/s, dis-loss=0.492, gen-loss=1.39]\n",
      "Epoch 79/1000: 100%|██████████| 469/469 [00:23<00:00, 20.27it/s, dis-loss=0.484, gen-loss=1.39]\n",
      "Epoch 80/1000: 100%|██████████| 469/469 [00:23<00:00, 20.13it/s, dis-loss=0.479, gen-loss=1.41]\n",
      "Epoch 81/1000: 100%|██████████| 469/469 [00:23<00:00, 20.31it/s, dis-loss=0.477, gen-loss=1.42]\n",
      "Epoch 82/1000: 100%|██████████| 469/469 [00:23<00:00, 20.22it/s, dis-loss=0.485, gen-loss=1.39]\n",
      "Epoch 83/1000: 100%|██████████| 469/469 [00:23<00:00, 20.11it/s, dis-loss=0.48, gen-loss=1.41] \n",
      "Epoch 84/1000: 100%|██████████| 469/469 [00:24<00:00, 19.41it/s, dis-loss=0.479, gen-loss=1.41]\n",
      "Epoch 85/1000: 100%|██████████| 469/469 [00:23<00:00, 20.23it/s, dis-loss=0.477, gen-loss=1.42]\n",
      "Epoch 86/1000: 100%|██████████| 469/469 [00:23<00:00, 20.31it/s, dis-loss=0.477, gen-loss=1.42]\n",
      "Epoch 87/1000: 100%|██████████| 469/469 [00:23<00:00, 20.33it/s, dis-loss=0.475, gen-loss=1.42]\n",
      "Epoch 88/1000: 100%|██████████| 469/469 [00:23<00:00, 20.07it/s, dis-loss=0.47, gen-loss=1.43] \n",
      "Epoch 89/1000: 100%|██████████| 469/469 [00:23<00:00, 20.15it/s, dis-loss=0.475, gen-loss=1.43]\n",
      "Epoch 90/1000: 100%|██████████| 469/469 [00:23<00:00, 20.24it/s, dis-loss=0.471, gen-loss=1.42]\n",
      "Epoch 91/1000: 100%|██████████| 469/469 [00:23<00:00, 19.86it/s, dis-loss=0.473, gen-loss=1.43]\n",
      "Epoch 92/1000: 100%|██████████| 469/469 [00:23<00:00, 19.95it/s, dis-loss=0.472, gen-loss=1.43]\n",
      "Epoch 93/1000: 100%|██████████| 469/469 [00:23<00:00, 20.17it/s, dis-loss=0.47, gen-loss=1.44] \n",
      "Epoch 94/1000: 100%|██████████| 469/469 [00:23<00:00, 20.17it/s, dis-loss=0.464, gen-loss=1.45]\n",
      "Epoch 95/1000: 100%|██████████| 469/469 [00:23<00:00, 20.09it/s, dis-loss=0.468, gen-loss=1.45]\n",
      "Epoch 96/1000: 100%|██████████| 469/469 [00:23<00:00, 20.01it/s, dis-loss=0.462, gen-loss=1.45]\n",
      "Epoch 97/1000: 100%|██████████| 469/469 [00:23<00:00, 19.61it/s, dis-loss=0.467, gen-loss=1.44]\n",
      "Epoch 98/1000: 100%|██████████| 469/469 [00:23<00:00, 19.60it/s, dis-loss=0.459, gen-loss=1.46]\n",
      "Epoch 99/1000: 100%|██████████| 469/469 [00:23<00:00, 20.00it/s, dis-loss=0.464, gen-loss=1.46]\n",
      "Epoch 100/1000: 100%|██████████| 469/469 [00:22<00:00, 20.40it/s, dis-loss=0.467, gen-loss=1.46]\n",
      "Epoch 101/1000: 100%|██████████| 469/469 [00:23<00:00, 20.15it/s, dis-loss=0.461, gen-loss=1.46]\n",
      "Epoch 102/1000: 100%|██████████| 469/469 [00:23<00:00, 20.04it/s, dis-loss=0.458, gen-loss=1.47]\n",
      "Epoch 103/1000: 100%|██████████| 469/469 [00:23<00:00, 20.13it/s, dis-loss=0.454, gen-loss=1.47]\n",
      "Epoch 104/1000: 100%|██████████| 469/469 [00:23<00:00, 20.10it/s, dis-loss=0.448, gen-loss=1.49]\n",
      "Epoch 105/1000: 100%|██████████| 469/469 [00:23<00:00, 20.36it/s, dis-loss=0.454, gen-loss=1.48]\n",
      "Epoch 106/1000: 100%|██████████| 469/469 [00:23<00:00, 19.96it/s, dis-loss=0.45, gen-loss=1.49] \n",
      "Epoch 107/1000: 100%|██████████| 469/469 [00:23<00:00, 20.11it/s, dis-loss=0.445, gen-loss=1.5] \n",
      "Epoch 108/1000: 100%|██████████| 469/469 [00:23<00:00, 20.09it/s, dis-loss=0.448, gen-loss=1.5] \n",
      "Epoch 109/1000: 100%|██████████| 469/469 [00:23<00:00, 20.14it/s, dis-loss=0.452, gen-loss=1.5] \n",
      "Epoch 110/1000: 100%|██████████| 469/469 [00:23<00:00, 20.08it/s, dis-loss=0.442, gen-loss=1.52]\n",
      "Epoch 111/1000: 100%|██████████| 469/469 [00:23<00:00, 20.06it/s, dis-loss=0.444, gen-loss=1.51]\n",
      "Epoch 112/1000: 100%|██████████| 469/469 [00:23<00:00, 20.31it/s, dis-loss=0.435, gen-loss=1.53]\n",
      "Epoch 113/1000: 100%|██████████| 469/469 [00:23<00:00, 19.94it/s, dis-loss=0.441, gen-loss=1.53]\n",
      "Epoch 114/1000: 100%|██████████| 469/469 [00:23<00:00, 20.31it/s, dis-loss=0.444, gen-loss=1.52]\n",
      "Epoch 115/1000: 100%|██████████| 469/469 [00:23<00:00, 20.03it/s, dis-loss=0.431, gen-loss=1.54]\n",
      "Epoch 116/1000: 100%|██████████| 469/469 [00:23<00:00, 20.17it/s, dis-loss=0.431, gen-loss=1.55]\n",
      "Epoch 117/1000: 100%|██████████| 469/469 [00:23<00:00, 20.22it/s, dis-loss=0.435, gen-loss=1.55]\n",
      "Epoch 118/1000: 100%|██████████| 469/469 [00:23<00:00, 20.11it/s, dis-loss=0.435, gen-loss=1.55]\n",
      "Epoch 119/1000: 100%|██████████| 469/469 [00:23<00:00, 20.23it/s, dis-loss=0.429, gen-loss=1.55]\n",
      "Epoch 120/1000: 100%|██████████| 469/469 [00:23<00:00, 20.13it/s, dis-loss=0.431, gen-loss=1.56]\n",
      "Epoch 121/1000: 100%|██████████| 469/469 [00:23<00:00, 20.15it/s, dis-loss=0.424, gen-loss=1.57]\n",
      "Epoch 122/1000: 100%|██████████| 469/469 [00:23<00:00, 19.85it/s, dis-loss=0.435, gen-loss=1.56]\n",
      "Epoch 123/1000: 100%|██████████| 469/469 [00:23<00:00, 20.05it/s, dis-loss=0.424, gen-loss=1.56]\n",
      "Epoch 124/1000: 100%|██████████| 469/469 [00:23<00:00, 20.24it/s, dis-loss=0.429, gen-loss=1.58]\n",
      "Epoch 125/1000: 100%|██████████| 469/469 [00:23<00:00, 20.22it/s, dis-loss=0.417, gen-loss=1.59]\n",
      "Epoch 126/1000: 100%|██████████| 469/469 [00:23<00:00, 19.86it/s, dis-loss=0.422, gen-loss=1.59]\n",
      "Epoch 127/1000: 100%|██████████| 469/469 [00:23<00:00, 19.78it/s, dis-loss=0.411, gen-loss=1.61]\n",
      "Epoch 128/1000: 100%|██████████| 469/469 [00:24<00:00, 19.35it/s, dis-loss=0.428, gen-loss=1.59]\n",
      "Epoch 129/1000: 100%|██████████| 469/469 [00:24<00:00, 19.18it/s, dis-loss=0.416, gen-loss=1.6] \n",
      "Epoch 130/1000: 100%|██████████| 469/469 [00:24<00:00, 19.27it/s, dis-loss=0.409, gen-loss=1.62]\n",
      "Epoch 131/1000: 100%|██████████| 469/469 [00:24<00:00, 19.17it/s, dis-loss=0.413, gen-loss=1.62]\n",
      "Epoch 132/1000: 100%|██████████| 469/469 [00:24<00:00, 19.23it/s, dis-loss=0.415, gen-loss=1.61]\n",
      "Epoch 133/1000: 100%|██████████| 469/469 [00:24<00:00, 19.17it/s, dis-loss=0.412, gen-loss=1.63]\n",
      "Epoch 134/1000: 100%|██████████| 469/469 [00:24<00:00, 19.09it/s, dis-loss=0.41, gen-loss=1.62] \n",
      "Epoch 135/1000: 100%|██████████| 469/469 [00:24<00:00, 19.28it/s, dis-loss=0.416, gen-loss=1.62]\n",
      "Epoch 136/1000: 100%|██████████| 469/469 [00:24<00:00, 18.97it/s, dis-loss=0.406, gen-loss=1.63]\n",
      "Epoch 137/1000: 100%|██████████| 469/469 [00:24<00:00, 19.08it/s, dis-loss=0.398, gen-loss=1.65]\n",
      "Epoch 138/1000: 100%|██████████| 469/469 [00:24<00:00, 19.18it/s, dis-loss=0.41, gen-loss=1.66] \n",
      "Epoch 139/1000: 100%|██████████| 469/469 [00:24<00:00, 19.43it/s, dis-loss=0.402, gen-loss=1.65]\n",
      "Epoch 140/1000: 100%|██████████| 469/469 [00:24<00:00, 19.54it/s, dis-loss=0.402, gen-loss=1.66]\n",
      "Epoch 141/1000: 100%|██████████| 469/469 [00:24<00:00, 19.36it/s, dis-loss=0.398, gen-loss=1.67]\n",
      "Epoch 142/1000: 100%|██████████| 469/469 [00:24<00:00, 19.26it/s, dis-loss=0.404, gen-loss=1.66]\n",
      "Epoch 143/1000: 100%|██████████| 469/469 [00:24<00:00, 19.48it/s, dis-loss=0.396, gen-loss=1.67]\n",
      "Epoch 144/1000: 100%|██████████| 469/469 [00:24<00:00, 19.19it/s, dis-loss=0.395, gen-loss=1.67]\n",
      "Epoch 145/1000: 100%|██████████| 469/469 [00:24<00:00, 19.33it/s, dis-loss=0.401, gen-loss=1.67]\n",
      "Epoch 146/1000: 100%|██████████| 469/469 [00:24<00:00, 19.30it/s, dis-loss=0.386, gen-loss=1.69]\n",
      "Epoch 147/1000: 100%|██████████| 469/469 [00:24<00:00, 19.38it/s, dis-loss=0.386, gen-loss=1.7] \n",
      "Epoch 148/1000: 100%|██████████| 469/469 [00:24<00:00, 19.30it/s, dis-loss=0.389, gen-loss=1.7] \n",
      "Epoch 149/1000: 100%|██████████| 469/469 [00:24<00:00, 19.44it/s, dis-loss=0.39, gen-loss=1.7]  \n",
      "Epoch 150/1000: 100%|██████████| 469/469 [00:24<00:00, 19.44it/s, dis-loss=0.382, gen-loss=1.72]\n",
      "Epoch 151/1000: 100%|██████████| 469/469 [00:24<00:00, 19.24it/s, dis-loss=0.389, gen-loss=1.72]\n",
      "Epoch 152/1000: 100%|██████████| 469/469 [00:24<00:00, 19.29it/s, dis-loss=0.381, gen-loss=1.73]\n",
      "Epoch 153/1000: 100%|██████████| 469/469 [00:24<00:00, 19.41it/s, dis-loss=0.399, gen-loss=1.72]\n",
      "Epoch 154/1000: 100%|██████████| 469/469 [00:24<00:00, 19.48it/s, dis-loss=0.373, gen-loss=1.72]\n",
      "Epoch 155/1000: 100%|██████████| 469/469 [00:24<00:00, 19.32it/s, dis-loss=0.373, gen-loss=1.74]\n",
      "Epoch 156/1000: 100%|██████████| 469/469 [00:24<00:00, 19.33it/s, dis-loss=0.383, gen-loss=1.74]\n",
      "Epoch 157/1000: 100%|██████████| 469/469 [00:24<00:00, 19.34it/s, dis-loss=0.375, gen-loss=1.76]\n",
      "Epoch 158/1000: 100%|██████████| 469/469 [00:24<00:00, 19.20it/s, dis-loss=0.374, gen-loss=1.75]\n",
      "Epoch 159/1000: 100%|██████████| 469/469 [00:24<00:00, 19.35it/s, dis-loss=0.369, gen-loss=1.78]\n",
      "Epoch 160/1000: 100%|██████████| 469/469 [00:24<00:00, 19.33it/s, dis-loss=0.373, gen-loss=1.76]\n",
      "Epoch 161/1000: 100%|██████████| 469/469 [00:24<00:00, 19.35it/s, dis-loss=0.377, gen-loss=1.75]\n",
      "Epoch 162/1000: 100%|██████████| 469/469 [00:23<00:00, 19.60it/s, dis-loss=0.368, gen-loss=1.77]\n",
      "Epoch 163/1000:  70%|██████▉   | 328/469 [00:17<00:07, 18.74it/s, dis-loss=0.371, gen-loss=1.77]\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "Cell \u001b[0;32mIn [10], line 44\u001b[0m\n\u001b[1;32m     42\u001b[0m \u001b[39m# backward\u001b[39;00m\n\u001b[1;32m     43\u001b[0m G_optimizer\u001b[39m.\u001b[39mzero_grad()\n\u001b[0;32m---> 44\u001b[0m gen_loss\u001b[39m.\u001b[39;49mbackward()\n\u001b[1;32m     45\u001b[0m G_optimizer\u001b[39m.\u001b[39mstep()\n\u001b[1;32m     46\u001b[0m \u001b[39m# save figure\u001b[39;00m\n",
      "File \u001b[0;32m~/anaconda3/envs/pytorch/lib/python3.8/site-packages/torch/_tensor.py:396\u001b[0m, in \u001b[0;36mTensor.backward\u001b[0;34m(self, gradient, retain_graph, create_graph, inputs)\u001b[0m\n\u001b[1;32m    387\u001b[0m \u001b[39mif\u001b[39;00m has_torch_function_unary(\u001b[39mself\u001b[39m):\n\u001b[1;32m    388\u001b[0m     \u001b[39mreturn\u001b[39;00m handle_torch_function(\n\u001b[1;32m    389\u001b[0m         Tensor\u001b[39m.\u001b[39mbackward,\n\u001b[1;32m    390\u001b[0m         (\u001b[39mself\u001b[39m,),\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    394\u001b[0m         create_graph\u001b[39m=\u001b[39mcreate_graph,\n\u001b[1;32m    395\u001b[0m         inputs\u001b[39m=\u001b[39minputs)\n\u001b[0;32m--> 396\u001b[0m torch\u001b[39m.\u001b[39;49mautograd\u001b[39m.\u001b[39;49mbackward(\u001b[39mself\u001b[39;49m, gradient, retain_graph, create_graph, inputs\u001b[39m=\u001b[39;49minputs)\n",
      "File \u001b[0;32m~/anaconda3/envs/pytorch/lib/python3.8/site-packages/torch/autograd/__init__.py:173\u001b[0m, in \u001b[0;36mbackward\u001b[0;34m(tensors, grad_tensors, retain_graph, create_graph, grad_variables, inputs)\u001b[0m\n\u001b[1;32m    168\u001b[0m     retain_graph \u001b[39m=\u001b[39m create_graph\n\u001b[1;32m    170\u001b[0m \u001b[39m# The reason we repeat same the comment below is that\u001b[39;00m\n\u001b[1;32m    171\u001b[0m \u001b[39m# some Python versions print out the first line of a multi-line function\u001b[39;00m\n\u001b[1;32m    172\u001b[0m \u001b[39m# calls in the traceback and some print out the last line\u001b[39;00m\n\u001b[0;32m--> 173\u001b[0m Variable\u001b[39m.\u001b[39;49m_execution_engine\u001b[39m.\u001b[39;49mrun_backward(  \u001b[39m# Calls into the C++ engine to run the backward pass\u001b[39;49;00m\n\u001b[1;32m    174\u001b[0m     tensors, grad_tensors_, retain_graph, create_graph, inputs,\n\u001b[1;32m    175\u001b[0m     allow_unreachable\u001b[39m=\u001b[39;49m\u001b[39mTrue\u001b[39;49;00m, accumulate_grad\u001b[39m=\u001b[39;49m\u001b[39mTrue\u001b[39;49;00m)\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "criterion = nn.BCELoss()\n",
    "D_optimizer = torch.optim.Adam(D.parameters(),lr=2e-4, betas=(0.5, 0.999))\n",
    "G_optimizer = torch.optim.Adam(G.parameters(),lr=2e-4, betas=(0.5, 0.999))\n",
    "D.train()\n",
    "G.train()\n",
    "gen_loss_list = []\n",
    "dis_loss_list = []\n",
    "\n",
    "for epoch in range(Epoch):\n",
    "    with tqdm(total=MNIST_dataloader.__len__(),desc=f'Epoch {epoch+1}/{Epoch}')as pbar:\n",
    "        gen_loss_avg = []\n",
    "        dis_loss_avg = []\n",
    "        index = 0\n",
    "        for batch_idx,(img,_) in enumerate(MNIST_dataloader):\n",
    "            img = img.to(device)\n",
    "            # the output label\n",
    "            valid = torch.ones(img.size()[0]).to(device)\n",
    "            fake = torch.zeros(img.size()[0]).to(device)\n",
    "            # Generator input\n",
    "            G_img = torch.randn([img.size()[0],in_channel,1,1],requires_grad=True).to(device)\n",
    "            # ------------------Update Discriminator------------------\n",
    "            # forward\n",
    "            G_pred_gen = G(G_img)\n",
    "            G_pred_dis = D(G_pred_gen.detach())\n",
    "            R_pred_dis = D(img)\n",
    "            # the misfit\n",
    "            G_loss = criterion(G_pred_dis,fake)\n",
    "            R_loss = criterion(R_pred_dis,valid)\n",
    "            dis_loss = (G_loss+R_loss)/2\n",
    "            dis_loss_avg.append(dis_loss.item())\n",
    "            # backward\n",
    "            D_optimizer.zero_grad()\n",
    "            dis_loss.backward()\n",
    "            D_optimizer.step()\n",
    "            # ------------------Update Optimizer------------------\n",
    "            # forward\n",
    "            G_pred_gen = G(G_img)\n",
    "            G_pred_dis = D(G_pred_gen)\n",
    "            # the misfit\n",
    "            gen_loss = criterion(G_pred_dis,valid)\n",
    "            gen_loss_avg.append(gen_loss.item())\n",
    "            # backward\n",
    "            G_optimizer.zero_grad()\n",
    "            gen_loss.backward()\n",
    "            G_optimizer.step()\n",
    "            # save figure\n",
    "            if index % 200 == 0 or index + 1 == MNIST_dataset.__len__():\n",
    "                save_image(G_pred_gen, f'{log_path}/image/image_all/epoch-{epoch}-index-{index}.png')\n",
    "            index += 1\n",
    "            # ------------------进度条更新------------------\n",
    "            pbar.set_postfix(**{\n",
    "                'gen-loss': sum(gen_loss_avg) / len(gen_loss_avg),\n",
    "                'dis-loss': sum(dis_loss_avg) / len(dis_loss_avg)\n",
    "            })\n",
    "            pbar.update(1)\n",
    "        save_image(G_pred_gen, f'{log_path}/image/epoch-{epoch}.png')\n",
    "    filename = 'epoch%d-genLoss%.2f-disLoss%.2f' % (epoch, sum(gen_loss_avg) / len(gen_loss_avg), sum(dis_loss_avg) / len(dis_loss_avg))\n",
    "    torch.save(G.state_dict(), f'{log_path}/{filename}-gen.pth')\n",
    "    torch.save(D.state_dict(), f'{log_path}/{filename}-dis.pth')\n",
    "    # 记录损失\n",
    "    gen_loss_list.append(sum(gen_loss_avg) / len(gen_loss_avg))\n",
    "    dis_loss_list.append(sum(dis_loss_avg) / len(dis_loss_avg))\n",
    "    # 绘制损失图像并保存\n",
    "    plt.figure(0)\n",
    "    plt.plot(range(epoch + 1), gen_loss_list, 'r--', label='gen loss')\n",
    "    plt.plot(range(epoch + 1), dis_loss_list, 'r--', label='dis loss')\n",
    "    plt.legend()\n",
    "    plt.xlabel('epoch')\n",
    "    plt.ylabel('loss')\n",
    "    plt.savefig(f'{log_path}/loss.png', dpi=300)\n",
    "    plt.close(0)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 7. predict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "generator-dis: tensor([0.3375], device='cuda:0', grad_fn=<ViewBackward0>)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaAAAAGdCAYAAABU0qcqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAA9hAAAPYQGoP6dpAAAcP0lEQVR4nO3df2xV9f3H8ddtgQtoe7tS2tuuBQv+wFioGYPaqAylo+0WBkIWUbeVzchkhQj4s0ZFt6+pw8Q5HYMs2UAT8VcmMI1j0WJLdIUFlBC22dCuGzXQMkl6bylSWPv5/kG824UWPZd7++4tz0dyMnrvefd+PLvlyem9PfU555wAABhkKdYLAABcnAgQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwMcJ6AWfr6+vT4cOHlZaWJp/PZ70cAIBHzjl1dXUpLy9PKSkDn+cMuQAdPnxYBQUF1ssAAFygtrY25efnD3j/kAtQWlqaJCklJYUzIABIQs459fX1Rf4+H0jCArRu3To9/fTTam9vV3FxsZ5//nnNnDnzC+c+j47P5yNAAJDEvujv8IS8CeHVV1/V6tWrtWbNGn344YcqLi5WeXm5jh49moiHAwAkIV8iroZdUlKiGTNm6Fe/+pWkM28sKCgo0IoVK/TQQw+ddzYcDisQCCg1NZUzIABIQs459fb2KhQKKT09fcD94n4GdOrUKe3du1dlZWX/fZCUFJWVlamxsfGc/Xt6ehQOh6M2AMDwF/cAffrpp+rt7VVOTk7U7Tk5OWpvbz9n/9raWgUCgcjGO+AA4OJg/oOoNTU1CoVCka2trc16SQCAQRD3d8FlZWUpNTVVHR0dUbd3dHQoGAyes7/f75ff74/3MgAAQ1zcz4BGjRql6dOnq66uLnJbX1+f6urqVFpaGu+HAwAkqYT8HNDq1atVVVWlr3/965o5c6aeffZZdXd364c//GEiHg4AkIQSEqBbb71V//73v/XYY4+pvb1d1157rbZv337OGxMAABevhPwc0IXg54AAILmZ/RwQAABfBgECAJggQAAAEwQIAGCCAAEATBAgAIAJAgQAMEGAAAAmCBAAwAQBAgCYIEAAABMECABgggABAEwQIACACQIEADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwQYAAACYIEADABAECAJgYYb0AAF9OVVWV55lnnnkmpsdqbm72PDN//nzPM+3t7Z5nMHxwBgQAMEGAAAAmCBAAwAQBAgCYIEAAABMECABgggABAEwQIACACQIEADBBgAAAJggQAMAEAQIAmOBipMAFcs55nlm/fr3nmYULF3qe8fv9nmckKRQKeZ4pKiryPMPFSC9unAEBAEwQIACACQIEADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwQYAAACa4GClwgX7zm994npk/f77nmbS0NM8zsVz0VJJ+97vfeZ7561//6nnG5/N5nsHwwRkQAMAEAQIAmIh7gB5//HH5fL6obcqUKfF+GABAkkvIa0DXXHON3n333f8+yAheagIAREtIGUaMGKFgMJiITw0AGCYS8hrQwYMHlZeXp0mTJumOO+7QoUOHBty3p6dH4XA4agMADH9xD1BJSYk2bdqk7du3a/369WptbdWNN96orq6ufvevra1VIBCIbAUFBfFeEgBgCIp7gCorK/Xd735X06ZNU3l5ud5++211dnbqtdde63f/mpoahUKhyNbW1hbvJQEAhqCEvzsgIyNDV155pZqbm/u93+/3y+/3J3oZAIAhJuE/B3T8+HG1tLQoNzc30Q8FAEgicQ/Qfffdp4aGBv3zn//Un//8Z91yyy1KTU3VbbfdFu+HAgAksbh/C+6TTz7RbbfdpmPHjmn8+PG64YYbtGvXLo0fPz7eDwUASGI+55yzXsT/CofDCgQCSk1N5UKFGHTr1q3zPHPnnXd6nklJ8f7Nh23btnmeuf322z3PSNLp06djmgMkyTmn3t5ehUIhpaenD7gf14IDAJggQAAAEwQIAGCCAAEATBAgAIAJAgQAMEGAAAAmCBAAwAQBAgCYIEAAABMECABgggABAEwk/BfSARcqKyvL88yzzz4b02MtWrTI80woFPI8E8tFT5988knPM729vZ5ngMHCGRAAwAQBAgCYIEAAABMECABgggABAEwQIACACQIEADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMcDVsDHmXXXaZ55lvf/vbMT3W0aNHPc/86Ec/8jxTV1fneQYYbjgDAgCYIEAAABMECABgggABAEwQIACACQIEADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMcDFSDCqfz+d5pqqqyvOM3+/3PCNJv/zlLz3PfPDBBzE9FnCx4wwIAGCCAAEATBAgAIAJAgQAMEGAAAAmCBAAwAQBAgCYIEAAABMECABgggABAEwQIACACQIEADDhc84560X8r3A4rEAgoNTU1JguXImh7YYbbvA88+abb3qe6e7u9jwjSQUFBZ5nhtiXEGDOOafe3l6FQiGlp6cPuB9nQAAAEwQIAGDCc4B27typefPmKS8vTz6fT1u3bo263zmnxx57TLm5uRozZozKysp08ODBeK0XADBMeA5Qd3e3iouLtW7dun7vX7t2rZ577jlt2LBBu3fv1iWXXKLy8nKdPHnyghcLABg+PP9G1MrKSlVWVvZ7n3NOzz77rB555BHNnz9fkvTiiy8qJydHW7du1eLFiy9stQCAYSOurwG1traqvb1dZWVlkdsCgYBKSkrU2NjY70xPT4/C4XDUBgAY/uIaoPb2dklSTk5O1O05OTmR+85WW1urQCAQ2WJ5GywAIPmYvwuupqZGoVAosrW1tVkvCQAwCOIaoGAwKEnq6OiIur2joyNy39n8fr/S09OjNgDA8BfXABUWFioYDKquri5yWzgc1u7du1VaWhrPhwIAJDnP74I7fvy4mpubIx+3trZq3759yszM1IQJE7Ry5Ur93//9n6644goVFhbq0UcfVV5enhYsWBDPdQMAkpznAO3Zs0c33XRT5OPVq1dLkqqqqrRp0yY98MAD6u7u1tKlS9XZ2akbbrhB27dv1+jRo+O3agBA0uNipBhUn/98mBevvPKK55l9+/Z5npE0pL9VHMuXaqxfQ4P5WBh+uBgpAGBII0AAABMECABgggABAEwQIACACQIEADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAnPv44BuBD33nuv55mUFO//TnrjjTc8z0hSb2+v55msrCzPM1VVVZ5nrrvuOs8z3d3dnmek2K5A3tXV5XnmH//4h+eZo0ePep7B0MQZEADABAECAJggQAAAEwQIAGCCAAEATBAgAIAJAgQAMEGAAAAmCBAAwAQBAgCYIEAAABMECABggouRYlCNHTvW80xfX5/nmYyMDM8zkpSfn+955qWXXvI8M23aNM8z6enpnmd8Pp/nGUn6/ve/73nm1KlTnmc++OADzzMPPvig55l9+/Z5nkHicQYEADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwQYAAACYIEADABAECAJjgYqQYVHV1dZ5niouLPc9UVlZ6npGkH//4x55nAoFATI/l1YkTJzzPjB49OqbH6u3t9TyTkuL937M333yz55kf/OAHnme4GOnQxBkQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwQYAAACYIEADABAECAJggQAAAEwQIAGCCi5FiUKWnpw/K40ydOjWmuVAo5HnmD3/4g+eZQ4cOeZ7p6OjwPLNhwwbPM5J06aWXep6pqanxPLN06VLPM0VFRZ5nrr32Ws8zEhcxTTTOgAAAJggQAMCE5wDt3LlT8+bNU15ennw+n7Zu3Rp1/5IlS+Tz+aK2ioqKeK0XADBMeA5Qd3e3iouLtW7dugH3qaio0JEjRyLbyy+/fEGLBAAMP57fhFBZWfmFv23S7/crGAzGvCgAwPCXkNeA6uvrlZ2drauuukrLli3TsWPHBty3p6dH4XA4agMADH9xD1BFRYVefPFF1dXV6ec//7kaGhpUWVk54O+Yr62tVSAQiGwFBQXxXhIAYAiK+88BLV68OPLnqVOnatq0aZo8ebLq6+s1Z86cc/avqanR6tWrIx+Hw2EiBAAXgYS/DXvSpEnKyspSc3Nzv/f7/X6lp6dHbQCA4S/hAfrkk0907Ngx5ebmJvqhAABJxPO34I4fPx51NtPa2qp9+/YpMzNTmZmZeuKJJ7Ro0SIFg0G1tLTogQce0OWXX67y8vK4LhwAkNw8B2jPnj266aabIh9//vpNVVWV1q9fr/379+uFF15QZ2en8vLyNHfuXP3sZz+T3++P36oBAEnPc4Bmz54t59yA9//pT3+6oAVheOvs7BzSj7NixQrPM7H8oPWIEd7f/3O+r7uBpKTE9l32WH4cYtWqVZ5nYrkY6fjx4z3PYGjiWnAAABMECABgggABAEwQIACACQIEADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwEfdfyQ2czwsvvOB55p577vE88/DDD3uekaTXXnstpjmvent7Pc8M1hW0JWnkyJGeZ5YtWxbTY3mVn5/veSaWY4fE4wwIAGCCAAEATBAgAIAJAgQAMEGAAAAmCBAAwAQBAgCYIEAAABMECABgggABAEwQIACACQIEADDBFfowqD7++GPPM1dffbXnme985zueZ6TYLt4Zy4VFU1NTPc/EsrZYL0Y6Y8YMzzP3339/TI/lVSwXFj18+HACVoILxRkQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwQYAAACYIEADABAECAJggQAAAEwQIAGCCi5FiyMvPz/c8s2TJkpgeK5aLd27evNnzzBVXXOF5Jisra1AeR5IeeughzzPjxo3zPBPL8f7973/veYaLkQ5NnAEBAEwQIACACQIEADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwQYAAACZ8LparASZQOBxWIBBQamqqfD6f9XIwBDz55JOeZx544IEErKR/n332meeZ0aNHJ2Al5xrMr6HOzk7PMzU1NZ5nNmzY4HlmxAiuuzyYnHPq7e1VKBRSenr6gPtxBgQAMEGAAAAmPAWotrZWM2bMUFpamrKzs7VgwQI1NTVF7XPy5ElVV1dr3LhxuvTSS7Vo0SJ1dHTEddEAgOTnKUANDQ2qrq7Wrl279M477+j06dOaO3euuru7I/usWrVKb775pl5//XU1NDTo8OHDWrhwYdwXDgBIbp5emdu+fXvUx5s2bVJ2drb27t2rWbNmKRQK6be//a02b96sm2++WZK0ceNGXX311dq1a5euu+66+K0cAJDULug1oFAoJEnKzMyUJO3du1enT59WWVlZZJ8pU6ZowoQJamxs7Pdz9PT0KBwOR20AgOEv5gD19fVp5cqVuv7661VUVCRJam9v16hRo5SRkRG1b05Ojtrb2/v9PLW1tQoEApGtoKAg1iUBAJJIzAGqrq7WgQMH9Morr1zQAmpqahQKhSJbW1vbBX0+AEByiOmns5YvX6633npLO3fuVH5+fuT2YDCoU6dOqbOzM+osqKOjQ8FgsN/P5ff75ff7Y1kGACCJeToDcs5p+fLl2rJli3bs2KHCwsKo+6dPn66RI0eqrq4ucltTU5MOHTqk0tLS+KwYADAseDoDqq6u1ubNm7Vt2zalpaVFXtcJBAIaM2aMAoGA7rzzTq1evVqZmZlKT0/XihUrVFpayjvgAABRPAVo/fr1kqTZs2dH3b5x40YtWbJEkvSLX/xCKSkpWrRokXp6elReXq5f//rXcVksAGD44GKkGPLmzZvneaaqqiqmx6qoqPA8E8trmH19fZ5nTpw44Xnmqaee8jwjSf/5z388z7S2tnqeefvttz3PnDx50vMMBhcXIwUADGkECABgggABAEwQIACACQIEADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwwdWwMeSlpHj/d9LYsWNjeqxYfnHi2b+e5MtYvHix55lvfvObnmdaWlo8z0jSyJEjPc/EcgVtDE9cDRsAMKQRIACACQIEADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwQYAAACa4GCkAIK64GCkAYEgjQAAAEwQIAGCCAAEATBAgAIAJAgQAMEGAAAAmCBAAwAQBAgCYIEAAABMECABgggABAEwQIACACQIEADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwQYAAACYIEADABAECAJggQAAAEwQIAGCCAAEATHgKUG1trWbMmKG0tDRlZ2drwYIFampqitpn9uzZ8vl8Udvdd98d10UDAJKfpwA1NDSourpau3bt0jvvvKPTp09r7ty56u7ujtrvrrvu0pEjRyLb2rVr47poAEDyG+Fl5+3bt0d9vGnTJmVnZ2vv3r2aNWtW5PaxY8cqGAzGZ4UAgGHpgl4DCoVCkqTMzMyo21966SVlZWWpqKhINTU1OnHixICfo6enR+FwOGoDAAx/ns6A/ldfX59Wrlyp66+/XkVFRZHbb7/9dk2cOFF5eXnav3+/HnzwQTU1NemNN97o9/PU1tbqiSeeiHUZAIAk5XPOuVgGly1bpj/+8Y96//33lZ+fP+B+O3bs0Jw5c9Tc3KzJkyefc39PT496enoiH4fDYRUUFCg1NVU+ny+WpQEADDnn1Nvbq1AopPT09AH3i+kMaPny5Xrrrbe0c+fO88ZHkkpKSiRpwAD5/X75/f5YlgEASGKeAuSc04oVK7RlyxbV19ersLDwC2f27dsnScrNzY1pgQCA4clTgKqrq7V582Zt27ZNaWlpam9vlyQFAgGNGTNGLS0t2rx5s771rW9p3Lhx2r9/v1atWqVZs2Zp2rRpCfkPAAAkJ0+vAQ30mszGjRu1ZMkStbW16Xvf+54OHDig7u5uFRQU6JZbbtEjjzxy3u8D/q9wOKxAIMBrQACQpL7sa0AxvwkhUQgQACS3LxsgrgUHADBBgAAAJggQAMAEAQIAmCBAAAATBAgAYIIAAQBMECAAgAkCBAAwQYAAACYIEADABAECAJggQAAAEwQIAGCCAAEATBAgAIAJAgQAMEGAAAAmCBAAwAQBAgCYIEAAABMECABgggABAEwQIACACQIEADAxwnoBZ3PORf0vACC5fNm/x4dcgLq6uiRJfX19xisBAFyIrq4uBQKBAe/3uSF2qtHX16fDhw8rLS1NPp8v6r5wOKyCggK1tbUpPT3daIX2OA5ncBzO4DicwXE4YygcB+ecurq6lJeXp5SUgV/pGXJnQCkpKcrPzz/vPunp6Rf1E+xzHIczOA5ncBzO4DicYX0cznfm8znehAAAMEGAAAAmkipAfr9fa9askd/vt16KKY7DGRyHMzgOZ3Aczkim4zDk3oQAALg4JNUZEABg+CBAAAATBAgAYIIAAQBMJE2A1q1bp8suu0yjR49WSUmJ/vKXv1gvadA9/vjj8vl8UduUKVOsl5VwO3fu1Lx585SXlyefz6etW7dG3e+c02OPPabc3FyNGTNGZWVlOnjwoM1iE+iLjsOSJUvOeX5UVFTYLDZBamtrNWPGDKWlpSk7O1sLFixQU1NT1D4nT55UdXW1xo0bp0svvVSLFi1SR0eH0YoT48sch9mzZ5/zfLj77ruNVty/pAjQq6++qtWrV2vNmjX68MMPVVxcrPLych09etR6aYPummuu0ZEjRyLb+++/b72khOvu7lZxcbHWrVvX7/1r167Vc889pw0bNmj37t265JJLVF5erpMnTw7yShPri46DJFVUVEQ9P15++eVBXGHiNTQ0qLq6Wrt27dI777yj06dPa+7cueru7o7ss2rVKr355pt6/fXX1dDQoMOHD2vhwoWGq46/L3McJOmuu+6Kej6sXbvWaMUDcElg5syZrrq6OvJxb2+vy8vLc7W1tYarGnxr1qxxxcXF1sswJclt2bIl8nFfX58LBoPu6aefjtzW2dnp/H6/e/nllw1WODjOPg7OOVdVVeXmz59vsh4rR48edZJcQ0ODc+7M//cjR450r7/+emSfv//9706Sa2xstFpmwp19HJxz7hvf+Ia755577Bb1JQz5M6BTp05p7969Kisri9yWkpKisrIyNTY2Gq7MxsGDB5WXl6dJkybpjjvu0KFDh6yXZKq1tVXt7e1Rz49AIKCSkpKL8vlRX1+v7OxsXXXVVVq2bJmOHTtmvaSECoVCkqTMzExJ0t69e3X69Omo58OUKVM0YcKEYf18OPs4fO6ll15SVlaWioqKVFNToxMnTlgsb0BD7mKkZ/v000/V29urnJycqNtzcnL08ccfG63KRklJiTZt2qSrrrpKR44c0RNPPKEbb7xRBw4cUFpamvXyTLS3t0tSv8+Pz++7WFRUVGjhwoUqLCxUS0uLHn74YVVWVqqxsVGpqanWy4u7vr4+rVy5Utdff72KiooknXk+jBo1ShkZGVH7DufnQ3/HQZJuv/12TZw4UXl5edq/f78efPBBNTU16Y033jBcbbQhHyD8V2VlZeTP06ZNU0lJiSZOnKjXXntNd955p+HKMBQsXrw48uepU6dq2rRpmjx5surr6zVnzhzDlSVGdXW1Dhw4cFG8Dno+Ax2HpUuXRv48depU5ebmas6cOWppadHkyZMHe5n9GvLfgsvKylJqauo572Lp6OhQMBg0WtXQkJGRoSuvvFLNzc3WSzHz+XOA58e5Jk2apKysrGH5/Fi+fLneeustvffee1G/viUYDOrUqVPq7OyM2n+4Ph8GOg79KSkpkaQh9XwY8gEaNWqUpk+frrq6ushtfX19qqurU2lpqeHK7B0/flwtLS3Kzc21XoqZwsJCBYPBqOdHOBzW7t27L/rnxyeffKJjx44Nq+eHc07Lly/Xli1btGPHDhUWFkbdP336dI0cOTLq+dDU1KRDhw4Nq+fDFx2H/uzbt0+ShtbzwfpdEF/GK6+84vx+v9u0aZP729/+5pYuXeoyMjJce3u79dIG1b333uvq6+tda2ur++CDD1xZWZnLyspyR48etV5aQnV1dbmPPvrIffTRR06Se+aZZ9xHH33k/vWvfznnnHvqqadcRkaG27Ztm9u/f7+bP3++KywsdJ999pnxyuPrfMehq6vL3Xfffa6xsdG1tra6d999133ta19zV1xxhTt58qT10uNm2bJlLhAIuPr6enfkyJHIduLEicg+d999t5swYYLbsWOH27NnjystLXWlpaWGq46/LzoOzc3N7qc//anbs2ePa21tddu2bXOTJk1ys2bNMl55tKQIkHPOPf/8827ChAlu1KhRbubMmW7Xrl3WSxp0t956q8vNzXWjRo1yX/3qV92tt97qmpubrZeVcO+9956TdM5WVVXlnDvzVuxHH33U5eTkOL/f7+bMmeOamppsF50A5zsOJ06ccHPnznXjx493I0eOdBMnTnR33XXXsPtHWn///ZLcxo0bI/t89tln7ic/+Yn7yle+4saOHetuueUWd+TIEbtFJ8AXHYdDhw65WbNmuczMTOf3+93ll1/u7r//fhcKhWwXfhZ+HQMAwMSQfw0IADA8ESAAgAkCBAAwQYAAACYIEADABAECAJggQAAAEwQIAGCCAAEATBAgAIAJAgQAMEGAAAAm/h9p4WFMBrisjgAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "input_size = [1, 28, 28]\n",
    "in_channel = 64\n",
    "gen_para_path = './log/DCGAN_2023-02-22-16_24_04/epoch161-genLoss1.77-disLoss0.37-gen.pth'\n",
    "dis_para_path = './log/DCGAN_2023-02-22-16_24_04/epoch161-genLoss1.77-disLoss0.37-dis.pth'\n",
    "device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
    "gen = Generator(in_channel=in_channel).to(device)\n",
    "dis = Discriminator(input_size=input_size).to(device)\n",
    "gen.load_state_dict(torch.load(gen_para_path, map_location=device))\n",
    "gen.eval()\n",
    "# 随机生成一组数据\n",
    "G_img = torch.randn([1, in_channel, 1, 1], requires_grad=False).to(device)\n",
    "# 放入网路\n",
    "G_pred = gen(G_img)\n",
    "G_dis = dis(G_pred)\n",
    "print('generator-dis:', G_dis)\n",
    "# 图像显示\n",
    "G_pred = G_pred[0, ...].cpu().detach()\n",
    "plt.figure()\n",
    "plt.imshow(G_pred.reshape(28,28),cmap=\"gray\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "pytorch",
   "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": "3.8.13 (default, Mar 28 2022, 11:38:47) \n[GCC 7.5.0]"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "971ee1da7255ecde3216791eeee73f2ee8d2956ba6b6da13cdfb2a534ec0c35f"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}