res-net-20-56.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "fc450b37",
   "metadata": {
    "papermill": {
     "duration": 0.005262,
     "end_time": "2023-12-04T10:23:05.554251",
     "exception": false,
     "start_time": "2023-12-04T10:23:05.548989",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h1>モジュールのインストール</h1>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "99292bfd",
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2023-12-04T10:23:05.565683Z",
     "iopub.status.busy": "2023-12-04T10:23:05.565325Z",
     "iopub.status.idle": "2023-12-04T10:23:23.715458Z",
     "shell.execute_reply": "2023-12-04T10:23:23.714514Z"
    },
    "papermill": {
     "duration": 18.158527,
     "end_time": "2023-12-04T10:23:23.717964",
     "exception": false,
     "start_time": "2023-12-04T10:23:05.559437",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: torch in /opt/conda/lib/python3.10/site-packages (2.0.0)\r\n",
      "Requirement already satisfied: torchvision in /opt/conda/lib/python3.10/site-packages (0.15.1)\r\n",
      "Requirement already satisfied: filelock in /opt/conda/lib/python3.10/site-packages (from torch) (3.12.2)\r\n",
      "Requirement already satisfied: typing-extensions in /opt/conda/lib/python3.10/site-packages (from torch) (4.5.0)\r\n",
      "Requirement already satisfied: sympy in /opt/conda/lib/python3.10/site-packages (from torch) (1.12)\r\n",
      "Requirement already satisfied: networkx in /opt/conda/lib/python3.10/site-packages (from torch) (3.1)\r\n",
      "Requirement already satisfied: jinja2 in /opt/conda/lib/python3.10/site-packages (from torch) (3.1.2)\r\n",
      "Requirement already satisfied: numpy in /opt/conda/lib/python3.10/site-packages (from torchvision) (1.24.3)\r\n",
      "Requirement already satisfied: requests in /opt/conda/lib/python3.10/site-packages (from torchvision) (2.31.0)\r\n",
      "Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in /opt/conda/lib/python3.10/site-packages (from torchvision) (10.1.0)\r\n",
      "Requirement already satisfied: MarkupSafe>=2.0 in /opt/conda/lib/python3.10/site-packages (from jinja2->torch) (2.1.3)\r\n",
      "Requirement already satisfied: charset-normalizer<4,>=2 in /opt/conda/lib/python3.10/site-packages (from requests->torchvision) (3.2.0)\r\n",
      "Requirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.10/site-packages (from requests->torchvision) (3.4)\r\n",
      "Requirement already satisfied: urllib3<3,>=1.21.1 in /opt/conda/lib/python3.10/site-packages (from requests->torchvision) (1.26.15)\r\n",
      "Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.10/site-packages (from requests->torchvision) (2023.7.22)\r\n",
      "Requirement already satisfied: mpmath>=0.19 in /opt/conda/lib/python3.10/site-packages (from sympy->torch) (1.3.0)\r\n"
     ]
    }
   ],
   "source": [
    "!pip install torch torchvision\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import torch\n",
    "import torchvision\n",
    "import torchvision.transforms as transforms\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "import time"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1f515d27",
   "metadata": {
    "papermill": {
     "duration": 0.005148,
     "end_time": "2023-12-04T10:23:23.729021",
     "exception": false,
     "start_time": "2023-12-04T10:23:23.723873",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h1>データの準備</h1>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "f5fcd09e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-12-04T10:23:23.741193Z",
     "iopub.status.busy": "2023-12-04T10:23:23.740765Z",
     "iopub.status.idle": "2023-12-04T10:23:29.719307Z",
     "shell.execute_reply": "2023-12-04T10:23:29.718281Z"
    },
    "papermill": {
     "duration": 5.987486,
     "end_time": "2023-12-04T10:23:29.721820",
     "exception": false,
     "start_time": "2023-12-04T10:23:23.734334",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to ./cifar-10/cifar-10-python.tar.gz\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 170498071/170498071 [00:02<00:00, 72018624.31it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./cifar-10/cifar-10-python.tar.gz to ./cifar-10\n",
      "Files already downloaded and verified\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "import torchvision\n",
    "from torchvision import transforms\n",
    "\n",
    "transform_train = transforms.Compose([\n",
    "    transforms.RandomHorizontalFlip(),\n",
    "    transforms.ToTensor(),\n",
    "    transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))\n",
    "])\n",
    "\n",
    "\n",
    "transform_test = transforms.Compose([\n",
    "    transforms.ToTensor(),\n",
    "    transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))\n",
    "])\n",
    "\n",
    "\n",
    "\n",
    "train_dataset = torchvision.datasets. CIFAR10(\n",
    "    './cifar-10', train=True, download=True,\n",
    "    transform=transform_train)\n",
    "\n",
    "\n",
    "test_data = torchvision.datasets. CIFAR10(\n",
    "    './cifar-10', train=False, download=True,\n",
    "    transform=transform_test)\n",
    "\n",
    "# train_dataをshuffleしてから分割\n",
    "\n",
    "train_data, val_data = torch.utils.data.random_split(train_dataset, [45000, 5000])\n",
    "\n",
    "train_loader = torch.utils.data.DataLoader(train_data,  batch_size=128, shuffle = True, num_workers = 2)\n",
    "val_loader = torch.utils.data.DataLoader(val_data, batch_size = 128)\n",
    "\n",
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c9ffa09d",
   "metadata": {
    "papermill": {
     "duration": 0.007638,
     "end_time": "2023-12-04T10:23:29.737451",
     "exception": false,
     "start_time": "2023-12-04T10:23:29.729813",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h1>Res Net 20 アーキテクチャ</h1>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "b799f4c4",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-12-04T10:23:29.754355Z",
     "iopub.status.busy": "2023-12-04T10:23:29.754008Z",
     "iopub.status.idle": "2023-12-04T10:23:33.019352Z",
     "shell.execute_reply": "2023-12-04T10:23:33.018535Z"
    },
    "papermill": {
     "duration": 3.27669,
     "end_time": "2023-12-04T10:23:33.021792",
     "exception": false,
     "start_time": "2023-12-04T10:23:29.745102",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "#シードの固定\n",
    "\n",
    "torch.manual_seed(4931)\n",
    "\n",
    "def identity(x,channels):\n",
    "        \n",
    "    x = F.interpolate(x, size=(x.size(2)//2, x.size(3)//2), mode='nearest')\n",
    "        \n",
    "    #テンソルの次元は [batch_size, channels, x , yの順番]\n",
    "        \n",
    "    zero_padding = torch.zeros(x.size(0),channels,x.size(2),x.size(3), device = device)\n",
    "    \n",
    "    x = torch.cat((x,zero_padding), dim = 1)\n",
    "        \n",
    "    return x\n",
    "    \n",
    "\n",
    "\n",
    "class ResNet20(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(ResNet20, self).__init__()\n",
    "        # Input size: [batch_size, 3, 32, 32] -> Output Size : [batch_size, 16, 32, 32]\n",
    "        self.conv1 = nn.Conv2d(in_channels=3, out_channels=16, kernel_size= 3, padding = 1)\n",
    "\n",
    "        self.bn1 = nn.BatchNorm2d(16)\n",
    "\n",
    "        # Input size: [batch_size, 16, 32, 32] -> Output Size : [batch_size, 16, 32, 32]\n",
    "        self.conv2_1 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_2 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_3 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_4 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_5 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_6 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "\n",
    "        self.bn2_1 = nn.BatchNorm2d(16)\n",
    "        self.bn2_2 = nn.BatchNorm2d(16)\n",
    "        self.bn2_3 = nn.BatchNorm2d(16)\n",
    "        self.bn2_4 = nn.BatchNorm2d(16)\n",
    "        self.bn2_5 = nn.BatchNorm2d(16)\n",
    "        self.bn2_6 = nn.BatchNorm2d(16)\n",
    "\n",
    "        # Input size: [batch_size, 16, 32, 32] -> Output Size : [batch_size, 32, 16, 16]\n",
    "        self.conv3 = nn.Conv2d(in_channels=16, out_channels=32, kernel_size=3, padding=1,stride = 2)\n",
    "\n",
    "        self.bn3 = nn.BatchNorm2d(32)\n",
    "        \n",
    "        \n",
    "\n",
    "        # Input size: [batch_size, 32, 16, 16] -> Output Size : [batch_size, 32, 16, 16]\n",
    "        self.conv4_1 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_2= nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_3 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_4 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_5 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "\n",
    "        self.bn4_1 = nn.BatchNorm2d(32)\n",
    "        self.bn4_2 = nn.BatchNorm2d(32)\n",
    "        self.bn4_3 = nn.BatchNorm2d(32)\n",
    "        self.bn4_4 = nn.BatchNorm2d(32)\n",
    "        self.bn4_5 = nn.BatchNorm2d(32)\n",
    "\n",
    "\n",
    "        # Input size: [batch_size, 32, 16, 16] -> Output Size : [batch_size, 64, 8, 8]\n",
    "        self.conv5 = nn.Conv2d(in_channels=32, out_channels=64, kernel_size=3, padding=1, stride = 2)\n",
    "\n",
    "        self.bn5 = nn.BatchNorm2d(64)\n",
    "\n",
    "        # Input size: [batch_size, 64, 8, 8] -> Output Size : [batch_size, 64, 8, 8]\n",
    "        self.conv6_1 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_2 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_3 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_4 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_5 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "\n",
    "\n",
    "        self.bn6_1 = nn.BatchNorm2d(64)\n",
    "        self.bn6_2 = nn.BatchNorm2d(64)\n",
    "        self.bn6_3 = nn.BatchNorm2d(64)\n",
    "        self.bn6_4 = nn.BatchNorm2d(64)\n",
    "\n",
    "\n",
    "\n",
    "        # Global Average Pooling\n",
    "        self.avgpool = nn.AdaptiveAvgPool2d(output_size=(1,1))\n",
    "        # Size after pooling: [batch_size, 64, 1, 1]\n",
    "\n",
    "        self.fc = nn.Linear(in_features=64, out_features=10)\n",
    "        \n",
    "        \n",
    "    \n",
    "\n",
    "\n",
    "    def forward(self, x):\n",
    "\n",
    "        x_1 = F.relu(self.bn1(self.conv1(x)))\n",
    "\n",
    "        x_2 = F.relu(self.bn2_1(self.conv2_1(x_1)))\n",
    "        x_3 = F.relu(self.bn2_2(self.conv2_2(x_2)) + x_1)\n",
    "        \n",
    "        x_4 = F.relu(self.bn2_3(self.conv2_3(x_3)))\n",
    "        x_5 = F.relu(self.bn2_4(self.conv2_4(x_4)) + x_3)\n",
    "        \n",
    "        x_6 = F.relu(self.bn2_5(self.conv2_5(x_5)))\n",
    "        x_7 = F.relu(self.bn2_6(self.conv2_6(x_6))+ x_5)\n",
    "        \n",
    "\n",
    "        \n",
    "        x_8 = F.relu(self.bn3(self.conv3(x_7)))\n",
    "        x_9 = F.relu(self.bn4_1(self.conv4_1(x_8))+ identity(x = x_7,channels = 16)) #恒等写像A\n",
    "        \n",
    "        \n",
    "        x_10 = F.relu(self.bn4_2(self.conv4_2(x_9)))\n",
    "        x_11 = F.relu(self.bn4_3(self.conv4_3(x_10)) + x_9)\n",
    "        \n",
    "        x_12 = F.relu(self.bn4_4(self.conv4_4(x_11)))\n",
    "        x_13 = F.relu(self.bn4_5(self.conv4_5(x_12)) + x_11)\n",
    "\n",
    "\n",
    "        x_14 = F.relu(self.bn5(self.conv5(x_13))) \n",
    "        x_15 = F.relu(self.bn6_1(self.conv6_1(x_14)) + identity(x = x_13,channels = 32)) #恒等写像A\n",
    "        \n",
    "        \n",
    "        x_16 = F.relu(self.bn6_2(self.conv6_2(x_15)))\n",
    "        x_17 = F.relu(self.bn6_3(self.conv6_3(x_16)) + x_15)\n",
    "        \n",
    "        x_18 = F.relu(self.bn6_4(self.conv6_4(x_17)))\n",
    "        x_19 = F.relu(self.conv6_5(x_18) + x_17)\n",
    "\n",
    "        x_19 = self.avgpool(x_19)\n",
    "\n",
    "        x_19 = x_19.view(-1, 64)\n",
    "        x = self.fc(x_19)\n",
    "\n",
    "        return x\n",
    "\n",
    "\n",
    "\n",
    "# Instantiate the network\n",
    "res_net20 = ResNet20()\n",
    "\n",
    "# データ並列学習 ON\n",
    "\n",
    "res_net20.to(device)\n",
    "\n",
    "res_net20 = nn.DataParallel(res_net20)\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d25d5396",
   "metadata": {
    "papermill": {
     "duration": 0.007273,
     "end_time": "2023-12-04T10:23:33.036979",
     "exception": false,
     "start_time": "2023-12-04T10:23:33.029706",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h1>Res Net 56 アーキテクチャ</h1>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "1d88dffe",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-12-04T10:23:33.053710Z",
     "iopub.status.busy": "2023-12-04T10:23:33.053423Z",
     "iopub.status.idle": "2023-12-04T10:23:33.144380Z",
     "shell.execute_reply": "2023-12-04T10:23:33.143579Z"
    },
    "papermill": {
     "duration": 0.102,
     "end_time": "2023-12-04T10:23:33.146543",
     "exception": false,
     "start_time": "2023-12-04T10:23:33.044543",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "#シードの固定\n",
    "\n",
    "torch.manual_seed(4931)\n",
    "\n",
    "\n",
    "class ResNet56(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(ResNet56, self).__init__()\n",
    "        # Input size: [batch_size, 3, 32, 32] -> Output Size : [batch_size, 16, 32, 32]\n",
    "        self.conv1 = nn.Conv2d(in_channels=3, out_channels=16, kernel_size= 3, padding = 1)\n",
    "\n",
    "        self.bn1 = nn.BatchNorm2d(16)\n",
    "\n",
    "        # Input size: [batch_size, 16, 32, 32] -> Output Size : [batch_size, 16, 32, 32]\n",
    "        self.conv2_1 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_2 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_3 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_4 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_5 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_6 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_7 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_8 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_9 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_10 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_11 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_12 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_13 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_14 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_15 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_16 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_17 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "        self.conv2_18 = nn.Conv2d(in_channels=16, out_channels=16, kernel_size=3, padding=1)\n",
    "\n",
    "        self.bn2_1 = nn.BatchNorm2d(16)\n",
    "        self.bn2_2 = nn.BatchNorm2d(16)\n",
    "        self.bn2_3 = nn.BatchNorm2d(16)\n",
    "        self.bn2_4 = nn.BatchNorm2d(16)\n",
    "        self.bn2_5 = nn.BatchNorm2d(16)\n",
    "        self.bn2_6 = nn.BatchNorm2d(16)\n",
    "        self.bn2_7 = nn.BatchNorm2d(16)\n",
    "        self.bn2_8 = nn.BatchNorm2d(16)\n",
    "        self.bn2_9 = nn.BatchNorm2d(16)\n",
    "        self.bn2_10 = nn.BatchNorm2d(16)\n",
    "        self.bn2_11 = nn.BatchNorm2d(16)\n",
    "        self.bn2_12 = nn.BatchNorm2d(16)\n",
    "        self.bn2_13 = nn.BatchNorm2d(16)\n",
    "        self.bn2_14 = nn.BatchNorm2d(16)\n",
    "        self.bn2_15 = nn.BatchNorm2d(16)\n",
    "        self.bn2_16 = nn.BatchNorm2d(16)\n",
    "        self.bn2_17 = nn.BatchNorm2d(16)\n",
    "        self.bn2_18 = nn.BatchNorm2d(16)\n",
    "\n",
    "        # Input size: [batch_size, 16, 32, 32] -> Output Size : [batch_size, 32, 16, 16]\n",
    "        self.conv3 = nn.Conv2d(in_channels=16, out_channels=32, kernel_size=3, padding=1,stride = 2)\n",
    "\n",
    "        self.bn3 = nn.BatchNorm2d(32)\n",
    "\n",
    "        # Input size: [batch_size, 32, 16, 16] -> Output Size : [batch_size, 32, 16, 16]\n",
    "        self.conv4_1 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_2= nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_3 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_4 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_5 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_6 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_7= nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_8 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_9 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_10 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_11 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_12= nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_13 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_14 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_15 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_16 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.conv4_17 = nn.Conv2d(in_channels=32, out_channels=32, kernel_size=3, padding=1)\n",
    "\n",
    "\n",
    "        self.bn4_1 = nn.BatchNorm2d(32)\n",
    "        self.bn4_2 = nn.BatchNorm2d(32)\n",
    "        self.bn4_3 = nn.BatchNorm2d(32)\n",
    "        self.bn4_4 = nn.BatchNorm2d(32)\n",
    "        self.bn4_5 = nn.BatchNorm2d(32)\n",
    "        self.bn4_6 = nn.BatchNorm2d(32)\n",
    "        self.bn4_7 = nn.BatchNorm2d(32)\n",
    "        self.bn4_8 = nn.BatchNorm2d(32)\n",
    "        self.bn4_9 = nn.BatchNorm2d(32)\n",
    "        self.bn4_10 = nn.BatchNorm2d(32)\n",
    "        self.bn4_11 = nn.BatchNorm2d(32)\n",
    "        self.bn4_12 = nn.BatchNorm2d(32)\n",
    "        self.bn4_13 = nn.BatchNorm2d(32)\n",
    "        self.bn4_14 = nn.BatchNorm2d(32)\n",
    "        self.bn4_15 = nn.BatchNorm2d(32)\n",
    "        self.bn4_16 = nn.BatchNorm2d(32)\n",
    "        self.bn4_17 = nn.BatchNorm2d(32)\n",
    "\n",
    "\n",
    "        # Input size: [batch_size, 32, 16, 16] -> Output Size : [batch_size, 64, 8, 8]\n",
    "        self.conv5 = nn.Conv2d(in_channels=32, out_channels=64, kernel_size=3, padding=1, stride = 2)\n",
    "\n",
    "        self.bn5 = nn.BatchNorm2d(64)\n",
    "\n",
    "        # Input size: [batch_size, 64, 8, 8] -> Output Size : [batch_size, 64, 8, 8]\n",
    "        self.conv6_1 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_2 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_3 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_4 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_5 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_6 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_7 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_8 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_9 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_10 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_11 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_12 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_13 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_14 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_15 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_16 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "        self.conv6_17 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=1)\n",
    "\n",
    "\n",
    "        self.bn6_1 = nn.BatchNorm2d(64)\n",
    "        self.bn6_2 = nn.BatchNorm2d(64)\n",
    "        self.bn6_3 = nn.BatchNorm2d(64)\n",
    "        self.bn6_4 = nn.BatchNorm2d(64)\n",
    "        self.bn6_5 = nn.BatchNorm2d(64)\n",
    "        self.bn6_6 = nn.BatchNorm2d(64)\n",
    "        self.bn6_7 = nn.BatchNorm2d(64)\n",
    "        self.bn6_8 = nn.BatchNorm2d(64)\n",
    "        self.bn6_9 = nn.BatchNorm2d(64)\n",
    "        self.bn6_10 = nn.BatchNorm2d(64)\n",
    "        self.bn6_11 = nn.BatchNorm2d(64)\n",
    "        self.bn6_12 = nn.BatchNorm2d(64)\n",
    "        self.bn6_13 = nn.BatchNorm2d(64)\n",
    "        self.bn6_14 = nn.BatchNorm2d(64)\n",
    "        self.bn6_15 = nn.BatchNorm2d(64)\n",
    "        self.bn6_16 = nn.BatchNorm2d(64)\n",
    "\n",
    "\n",
    "        # Global Average Pooling\n",
    "        self.avgpool = nn.AdaptiveAvgPool2d(output_size=(1,1))\n",
    "        # Size after pooling: [batch_size, 64, 1, 1]\n",
    "\n",
    "        self.fc = nn.Linear(in_features=64, out_features=10)\n",
    "        \n",
    "\n",
    "    def forward(self, x):\n",
    "    \n",
    "        x_1 = F.relu(self.bn1(self.conv1(x)))\n",
    "\n",
    "        x_2 = F.relu(self.bn2_1(self.conv2_1(x_1)))\n",
    "        x_3 = F.relu(self.bn2_2(self.conv2_2(x_2)) + x_1)\n",
    "        \n",
    "        x_4 = F.relu(self.bn2_3(self.conv2_3(x_3)))\n",
    "        x_5 = F.relu(self.bn2_4(self.conv2_4(x_4)) + x_3)\n",
    "        \n",
    "        x_6 = F.relu(self.bn2_5(self.conv2_5(x_5)))\n",
    "        x_7 = F.relu(self.bn2_6(self.conv2_6(x_6)) + x_5)\n",
    "        \n",
    "        x_8 = F.relu(self.bn2_7(self.conv2_7(x_7)))\n",
    "        x_9 = F.relu(self.bn2_8(self.conv2_8(x_8) + x_7))\n",
    "        \n",
    "        x_10 = F.relu(self.bn2_9(self.conv2_9(x_9)))\n",
    "        x_11 = F.relu(self.bn2_10(self.conv2_10(x_10)+ x_9))\n",
    "        \n",
    "        x_12 = F.relu(self.bn2_11(self.conv2_11(x_11)))\n",
    "        x_13 = F.relu(self.bn2_12(self.conv2_12(x_12)+ x_11))\n",
    "        \n",
    "        x_14 = F.relu(self.bn2_13(self.conv2_13(x_13)))\n",
    "        x_15 = F.relu(self.bn2_14(self.conv2_14(x_14)+ x_13))\n",
    "        \n",
    "        x_16 = F.relu(self.bn2_15(self.conv2_15(x_15)))\n",
    "        x_17 = F.relu(self.bn2_16(self.conv2_16(x_16)+ x_15))\n",
    "        \n",
    "        x_18 = F.relu(self.bn2_17(self.conv2_17(x_17)))\n",
    "        x_19 = F.relu(self.bn2_18(self.conv2_18(x_18)+ x_17))\n",
    "\n",
    "        \n",
    "        x_20 = F.relu(self.bn3(self.conv3(x_19)))\n",
    "        x_21 = F.relu(self.bn4_1(self.conv4_1(x_20))+ identity(x = x_19,channels = 16))\n",
    "              \n",
    "            \n",
    "        x_22 = F.relu(self.bn4_2(self.conv4_2(x_21)))\n",
    "        x_23 = F.relu(self.bn4_3(self.conv4_3(x_22)) + x_21)\n",
    "        \n",
    "        x_24 = F.relu(self.bn4_4(self.conv4_4(x_23)))\n",
    "        x_25 = F.relu(self.bn4_5(self.conv4_5(x_24)) + x_23)\n",
    "        \n",
    "        x_26 = F.relu(self.bn4_6(self.conv4_6(x_25)))\n",
    "        x_27 = F.relu(self.bn4_7(self.conv4_7(x_26)) + x_25)\n",
    "        \n",
    "        x_28 = F.relu(self.bn4_8(self.conv4_8(x_27)))\n",
    "        x_29 = F.relu(self.bn4_9(self.conv4_9(x_28) + x_27))\n",
    "        \n",
    "        \n",
    "        x_30 = F.relu(self.bn4_10(self.conv4_10(x_29)))\n",
    "        x_31 = F.relu(self.bn4_11(self.conv4_11(x_30)+ x_29))\n",
    "        \n",
    "        x_32 = F.relu(self.bn4_12(self.conv4_12(x_31)))\n",
    "        x_33 = F.relu(self.bn4_13(self.conv4_13(x_32)+ x_31))\n",
    "        \n",
    "        x_34 = F.relu(self.bn4_14(self.conv4_14(x_33)))\n",
    "        x_35 = F.relu(self.bn4_15(self.conv4_15(x_34)+ x_33))\n",
    "        \n",
    "        x_36 = F.relu(self.bn4_16(self.conv4_16(x_35)))\n",
    "        x_37 = F.relu(self.bn4_17(self.conv4_17(x_36)+ x_35))\n",
    "        \n",
    "        x_38 = F.relu(self.bn5(self.conv5(x_37)))\n",
    "        x_39 = F.relu(self.bn6_1(self.conv6_1(x_38))+ identity(x = x_37,channels = 32))\n",
    "        \n",
    "        x_40 = F.relu(self.bn6_2(self.conv6_2(x_39)))\n",
    "        x_41 = F.relu(self.bn6_3(self.conv6_3(x_40)+ x_39))\n",
    "        \n",
    "        x_42 = F.relu(self.bn6_4(self.conv6_4(x_41)))\n",
    "        x_43 = F.relu(self.bn6_5(self.conv6_5(x_42)+ x_41))\n",
    "        \n",
    "        x_44 = F.relu(self.bn6_6(self.conv6_6(x_43)))\n",
    "        x_45 = F.relu(self.bn6_7(self.conv6_7(x_44)+ x_43))\n",
    "        \n",
    "        x_46 = F.relu(self.bn6_8(self.conv6_8(x_45)))\n",
    "        x_47 = F.relu(self.bn6_9(self.conv6_9(x_46)+ x_45))\n",
    "        \n",
    "        x_48 = F.relu(self.bn6_10(self.conv6_10(x_47)))\n",
    "        x_49 = F.relu(self.bn6_11(self.conv6_11(x_48)+ x_47))\n",
    "        \n",
    "        x_50 = F.relu(self.bn6_12(self.conv6_12(x_49)))\n",
    "        x_51 = F.relu(self.bn6_13(self.conv6_13(x_50)+ x_49))\n",
    "        \n",
    "        x_52 = F.relu(self.bn6_14(self.conv6_14(x_51)))\n",
    "        x_53 = F.relu(self.bn6_15(self.conv6_15(x_52)+ x_51))\n",
    "        \n",
    "        x_54 = F.relu(self.bn6_16(self.conv6_16(x_53)))\n",
    "\n",
    "\n",
    "        x_55 = F.relu(self.conv6_17(x_54))\n",
    "\n",
    "        x_55 = self.avgpool(x_55)\n",
    "\n",
    "        x_55 = x_55.view(-1, 64)\n",
    "        x = self.fc(x_55)\n",
    "\n",
    "        return x\n",
    "\n",
    "# Instantiate the network\n",
    "res_net56 = ResNet56()\n",
    "\n",
    "\n",
    "\n",
    "# データ並列学習 ON\n",
    "res_net56.to(device)\n",
    "\n",
    "res_net56 = nn.DataParallel(res_net56)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bfd55e5a",
   "metadata": {
    "papermill": {
     "duration": 0.007646,
     "end_time": "2023-12-04T10:23:33.162115",
     "exception": false,
     "start_time": "2023-12-04T10:23:33.154469",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h1>損失関数の定義</h1>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "a39e10ff",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-12-04T10:23:33.177984Z",
     "iopub.status.busy": "2023-12-04T10:23:33.177701Z",
     "iopub.status.idle": "2023-12-04T10:23:33.184021Z",
     "shell.execute_reply": "2023-12-04T10:23:33.183129Z"
    },
    "papermill": {
     "duration": 0.016608,
     "end_time": "2023-12-04T10:23:33.186067",
     "exception": false,
     "start_time": "2023-12-04T10:23:33.169459",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# 交差エントロピー\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "# 確率的勾配降下法\n",
    "optimizer_20 = optim.SGD(res_net20.parameters(), lr=0.1, momentum=0.9, weight_decay = 0.0001)\n",
    "optimizer_56 = optim.SGD(res_net56.parameters(), lr=0.1, momentum=0.9, weight_decay = 0.0001)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b9ad11a1",
   "metadata": {
    "papermill": {
     "duration": 0.007161,
     "end_time": "2023-12-04T10:23:33.200573",
     "exception": false,
     "start_time": "2023-12-04T10:23:33.193412",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h1>Res Net 20 学習</h1>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "c6147e11",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-12-04T10:23:33.216785Z",
     "iopub.status.busy": "2023-12-04T10:23:33.216472Z",
     "iopub.status.idle": "2023-12-04T11:05:07.711119Z",
     "shell.execute_reply": "2023-12-04T11:05:07.710049Z"
    },
    "papermill": {
     "duration": 2494.526545,
     "end_time": "2023-12-04T11:05:07.734523",
     "exception": false,
     "start_time": "2023-12-04T10:23:33.207978",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/182\n",
      "Training loss: 1.521, Training accuracy: 43.564%\n",
      "Validation loss: 1.293, Validation accuracy: 54.660%\n",
      "Epoch 2/182\n",
      "Training loss: 1.063, Training accuracy: 61.960%\n",
      "Validation loss: 1.294, Validation accuracy: 57.360%\n",
      "Epoch 3/182\n",
      "Training loss: 0.846, Training accuracy: 70.242%\n",
      "Validation loss: 0.795, Validation accuracy: 72.060%\n",
      "Epoch 4/182\n",
      "Training loss: 0.706, Training accuracy: 75.133%\n",
      "Validation loss: 0.932, Validation accuracy: 68.000%\n",
      "Epoch 5/182\n",
      "Training loss: 0.620, Training accuracy: 78.429%\n",
      "Validation loss: 0.737, Validation accuracy: 74.580%\n",
      "Epoch 6/182\n",
      "Training loss: 0.563, Training accuracy: 80.436%\n",
      "Validation loss: 0.638, Validation accuracy: 77.800%\n",
      "Epoch 7/182\n",
      "Training loss: 0.507, Training accuracy: 82.478%\n",
      "Validation loss: 0.565, Validation accuracy: 80.320%\n",
      "Epoch 8/182\n",
      "Training loss: 0.477, Training accuracy: 83.538%\n",
      "Validation loss: 0.665, Validation accuracy: 78.280%\n",
      "Epoch 9/182\n",
      "Training loss: 0.443, Training accuracy: 84.682%\n",
      "Validation loss: 0.620, Validation accuracy: 78.340%\n",
      "Epoch 10/182\n",
      "Training loss: 0.413, Training accuracy: 85.553%\n",
      "Validation loss: 0.599, Validation accuracy: 79.660%\n",
      "Epoch 11/182\n",
      "Training loss: 0.387, Training accuracy: 86.644%\n",
      "Validation loss: 0.569, Validation accuracy: 80.980%\n",
      "Epoch 12/182\n",
      "Training loss: 0.375, Training accuracy: 86.951%\n",
      "Validation loss: 0.498, Validation accuracy: 82.840%\n",
      "Epoch 13/182\n",
      "Training loss: 0.350, Training accuracy: 87.844%\n",
      "Validation loss: 0.499, Validation accuracy: 83.020%\n",
      "Epoch 14/182\n",
      "Training loss: 0.341, Training accuracy: 88.182%\n",
      "Validation loss: 0.468, Validation accuracy: 84.440%\n",
      "Epoch 15/182\n",
      "Training loss: 0.324, Training accuracy: 88.684%\n",
      "Validation loss: 0.522, Validation accuracy: 82.800%\n",
      "Epoch 16/182\n",
      "Training loss: 0.313, Training accuracy: 89.067%\n",
      "Validation loss: 0.532, Validation accuracy: 82.060%\n",
      "Epoch 17/182\n",
      "Training loss: 0.299, Training accuracy: 89.542%\n",
      "Validation loss: 0.554, Validation accuracy: 82.400%\n",
      "Epoch 18/182\n",
      "Training loss: 0.290, Training accuracy: 89.818%\n",
      "Validation loss: 0.746, Validation accuracy: 75.920%\n",
      "Epoch 19/182\n",
      "Training loss: 0.282, Training accuracy: 90.120%\n",
      "Validation loss: 0.606, Validation accuracy: 80.500%\n",
      "Epoch 20/182\n",
      "Training loss: 0.270, Training accuracy: 90.389%\n",
      "Validation loss: 0.484, Validation accuracy: 83.820%\n",
      "Epoch 21/182\n",
      "Training loss: 0.259, Training accuracy: 90.896%\n",
      "Validation loss: 0.523, Validation accuracy: 83.540%\n",
      "Epoch 22/182\n",
      "Training loss: 0.251, Training accuracy: 91.267%\n",
      "Validation loss: 0.550, Validation accuracy: 83.200%\n",
      "Epoch 23/182\n",
      "Training loss: 0.243, Training accuracy: 91.433%\n",
      "Validation loss: 0.573, Validation accuracy: 82.860%\n",
      "Epoch 24/182\n",
      "Training loss: 0.243, Training accuracy: 91.389%\n",
      "Validation loss: 0.613, Validation accuracy: 79.380%\n",
      "Epoch 25/182\n",
      "Training loss: 0.232, Training accuracy: 91.864%\n",
      "Validation loss: 0.504, Validation accuracy: 82.960%\n",
      "Epoch 26/182\n",
      "Training loss: 0.228, Training accuracy: 91.873%\n",
      "Validation loss: 0.504, Validation accuracy: 84.320%\n",
      "Epoch 27/182\n",
      "Training loss: 0.222, Training accuracy: 92.140%\n",
      "Validation loss: 0.499, Validation accuracy: 83.880%\n",
      "Epoch 28/182\n",
      "Training loss: 0.216, Training accuracy: 92.467%\n",
      "Validation loss: 0.475, Validation accuracy: 85.400%\n",
      "Epoch 29/182\n",
      "Training loss: 0.216, Training accuracy: 92.289%\n",
      "Validation loss: 0.515, Validation accuracy: 83.640%\n",
      "Epoch 30/182\n",
      "Training loss: 0.212, Training accuracy: 92.558%\n",
      "Validation loss: 0.562, Validation accuracy: 83.680%\n",
      "Epoch 31/182\n",
      "Training loss: 0.204, Training accuracy: 92.718%\n",
      "Validation loss: 0.450, Validation accuracy: 85.500%\n",
      "Epoch 32/182\n",
      "Training loss: 0.205, Training accuracy: 92.713%\n",
      "Validation loss: 0.591, Validation accuracy: 82.880%\n",
      "Epoch 33/182\n",
      "Training loss: 0.196, Training accuracy: 93.080%\n",
      "Validation loss: 0.517, Validation accuracy: 83.760%\n",
      "Epoch 34/182\n",
      "Training loss: 0.196, Training accuracy: 92.918%\n",
      "Validation loss: 0.503, Validation accuracy: 84.460%\n",
      "Epoch 35/182\n",
      "Training loss: 0.186, Training accuracy: 93.349%\n",
      "Validation loss: 0.532, Validation accuracy: 84.020%\n",
      "Epoch 36/182\n",
      "Training loss: 0.190, Training accuracy: 93.189%\n",
      "Validation loss: 0.553, Validation accuracy: 83.980%\n",
      "Epoch 37/182\n",
      "Training loss: 0.187, Training accuracy: 93.422%\n",
      "Validation loss: 0.608, Validation accuracy: 83.200%\n",
      "Epoch 38/182\n",
      "Training loss: 0.181, Training accuracy: 93.642%\n",
      "Validation loss: 0.472, Validation accuracy: 85.600%\n",
      "Epoch 39/182\n",
      "Training loss: 0.176, Training accuracy: 93.738%\n",
      "Validation loss: 0.521, Validation accuracy: 83.920%\n",
      "Epoch 40/182\n",
      "Training loss: 0.178, Training accuracy: 93.702%\n",
      "Validation loss: 0.725, Validation accuracy: 78.980%\n",
      "Epoch 41/182\n",
      "Training loss: 0.179, Training accuracy: 93.749%\n",
      "Validation loss: 0.635, Validation accuracy: 82.500%\n",
      "Epoch 42/182\n",
      "Training loss: 0.177, Training accuracy: 93.716%\n",
      "Validation loss: 0.545, Validation accuracy: 83.360%\n",
      "Epoch 43/182\n",
      "Training loss: 0.164, Training accuracy: 94.153%\n",
      "Validation loss: 0.619, Validation accuracy: 81.920%\n",
      "Epoch 44/182\n",
      "Training loss: 0.166, Training accuracy: 94.091%\n",
      "Validation loss: 0.676, Validation accuracy: 81.420%\n",
      "Epoch 45/182\n",
      "Training loss: 0.167, Training accuracy: 94.016%\n",
      "Validation loss: 0.562, Validation accuracy: 83.280%\n",
      "Epoch 46/182\n",
      "Training loss: 0.161, Training accuracy: 94.258%\n",
      "Validation loss: 0.578, Validation accuracy: 83.600%\n",
      "Epoch 47/182\n",
      "Training loss: 0.155, Training accuracy: 94.422%\n",
      "Validation loss: 0.493, Validation accuracy: 85.020%\n",
      "Epoch 48/182\n",
      "Training loss: 0.163, Training accuracy: 94.242%\n",
      "Validation loss: 0.581, Validation accuracy: 83.900%\n",
      "Epoch 49/182\n",
      "Training loss: 0.164, Training accuracy: 94.131%\n",
      "Validation loss: 0.607, Validation accuracy: 82.620%\n",
      "Epoch 50/182\n",
      "Training loss: 0.155, Training accuracy: 94.591%\n",
      "Validation loss: 0.488, Validation accuracy: 85.380%\n",
      "Epoch 51/182\n",
      "Training loss: 0.160, Training accuracy: 94.458%\n",
      "Validation loss: 0.702, Validation accuracy: 80.360%\n",
      "Epoch 52/182\n",
      "Training loss: 0.153, Training accuracy: 94.622%\n",
      "Validation loss: 0.665, Validation accuracy: 81.940%\n",
      "Epoch 53/182\n",
      "Training loss: 0.162, Training accuracy: 94.200%\n",
      "Validation loss: 0.556, Validation accuracy: 84.460%\n",
      "Epoch 54/182\n",
      "Training loss: 0.150, Training accuracy: 94.698%\n",
      "Validation loss: 0.633, Validation accuracy: 82.540%\n",
      "Epoch 55/182\n",
      "Training loss: 0.153, Training accuracy: 94.547%\n",
      "Validation loss: 0.516, Validation accuracy: 85.340%\n",
      "Epoch 56/182\n",
      "Training loss: 0.152, Training accuracy: 94.616%\n",
      "Validation loss: 0.530, Validation accuracy: 84.700%\n",
      "Epoch 57/182\n",
      "Training loss: 0.142, Training accuracy: 95.038%\n",
      "Validation loss: 0.481, Validation accuracy: 86.340%\n",
      "Epoch 58/182\n",
      "Training loss: 0.148, Training accuracy: 94.844%\n",
      "Validation loss: 0.533, Validation accuracy: 85.300%\n",
      "Epoch 59/182\n",
      "Training loss: 0.146, Training accuracy: 94.816%\n",
      "Validation loss: 0.682, Validation accuracy: 82.040%\n",
      "Epoch 60/182\n",
      "Training loss: 0.148, Training accuracy: 94.702%\n",
      "Validation loss: 0.556, Validation accuracy: 83.860%\n",
      "Epoch 61/182\n",
      "Training loss: 0.148, Training accuracy: 94.833%\n",
      "Validation loss: 0.533, Validation accuracy: 84.960%\n",
      "Epoch 62/182\n",
      "Training loss: 0.139, Training accuracy: 95.000%\n",
      "Validation loss: 0.573, Validation accuracy: 84.160%\n",
      "Epoch 63/182\n",
      "Training loss: 0.144, Training accuracy: 94.904%\n",
      "Validation loss: 0.530, Validation accuracy: 84.340%\n",
      "Epoch 64/182\n",
      "Training loss: 0.145, Training accuracy: 94.840%\n",
      "Validation loss: 0.655, Validation accuracy: 82.580%\n",
      "Epoch 65/182\n",
      "Training loss: 0.152, Training accuracy: 94.589%\n",
      "Validation loss: 0.560, Validation accuracy: 83.980%\n",
      "Epoch 66/182\n",
      "Training loss: 0.140, Training accuracy: 95.002%\n",
      "Validation loss: 0.787, Validation accuracy: 80.460%\n",
      "Epoch 67/182\n",
      "Training loss: 0.148, Training accuracy: 94.687%\n",
      "Validation loss: 0.591, Validation accuracy: 83.460%\n",
      "Epoch 68/182\n",
      "Training loss: 0.149, Training accuracy: 94.829%\n",
      "Validation loss: 0.499, Validation accuracy: 85.700%\n",
      "Epoch 69/182\n",
      "Training loss: 0.135, Training accuracy: 95.204%\n",
      "Validation loss: 0.621, Validation accuracy: 81.620%\n",
      "Epoch 70/182\n",
      "Training loss: 0.134, Training accuracy: 95.304%\n",
      "Validation loss: 0.701, Validation accuracy: 82.040%\n",
      "Epoch 71/182\n",
      "Training loss: 0.138, Training accuracy: 95.184%\n",
      "Validation loss: 0.555, Validation accuracy: 84.220%\n",
      "Epoch 72/182\n",
      "Training loss: 0.134, Training accuracy: 95.333%\n",
      "Validation loss: 0.517, Validation accuracy: 85.880%\n",
      "Epoch 73/182\n",
      "Training loss: 0.134, Training accuracy: 95.309%\n",
      "Validation loss: 0.568, Validation accuracy: 85.160%\n",
      "Epoch 74/182\n",
      "Training loss: 0.139, Training accuracy: 95.007%\n",
      "Validation loss: 0.579, Validation accuracy: 84.580%\n",
      "Epoch 75/182\n",
      "Training loss: 0.135, Training accuracy: 95.200%\n",
      "Validation loss: 0.525, Validation accuracy: 84.960%\n",
      "Epoch 76/182\n",
      "Training loss: 0.137, Training accuracy: 95.071%\n",
      "Validation loss: 0.607, Validation accuracy: 83.380%\n",
      "Epoch 77/182\n",
      "Training loss: 0.138, Training accuracy: 95.096%\n",
      "Validation loss: 0.559, Validation accuracy: 84.340%\n",
      "Epoch 78/182\n",
      "Training loss: 0.136, Training accuracy: 95.187%\n",
      "Validation loss: 0.601, Validation accuracy: 84.480%\n",
      "Epoch 79/182\n",
      "Training loss: 0.133, Training accuracy: 95.320%\n",
      "Validation loss: 0.513, Validation accuracy: 85.380%\n",
      "Epoch 80/182\n",
      "Training loss: 0.127, Training accuracy: 95.484%\n",
      "Validation loss: 0.545, Validation accuracy: 85.040%\n",
      "Epoch 81/182\n",
      "Training loss: 0.140, Training accuracy: 95.091%\n",
      "Validation loss: 0.601, Validation accuracy: 83.400%\n",
      "Epoch 82/182\n",
      "Training loss: 0.131, Training accuracy: 95.398%\n",
      "Validation loss: 0.584, Validation accuracy: 84.060%\n",
      "Epoch 83/182\n",
      "Training loss: 0.132, Training accuracy: 95.298%\n",
      "Validation loss: 0.565, Validation accuracy: 83.460%\n",
      "Epoch 84/182\n",
      "Training loss: 0.134, Training accuracy: 95.204%\n",
      "Validation loss: 0.603, Validation accuracy: 84.320%\n",
      "Epoch 85/182\n",
      "Training loss: 0.139, Training accuracy: 95.036%\n",
      "Validation loss: 0.562, Validation accuracy: 83.540%\n",
      "Epoch 86/182\n",
      "Training loss: 0.133, Training accuracy: 95.338%\n",
      "Validation loss: 0.587, Validation accuracy: 84.520%\n",
      "Epoch 87/182\n",
      "Training loss: 0.131, Training accuracy: 95.307%\n",
      "Validation loss: 0.621, Validation accuracy: 84.140%\n",
      "Epoch 88/182\n",
      "Training loss: 0.130, Training accuracy: 95.444%\n",
      "Validation loss: 0.617, Validation accuracy: 83.040%\n",
      "Epoch 89/182\n",
      "Training loss: 0.129, Training accuracy: 95.371%\n",
      "Validation loss: 0.627, Validation accuracy: 83.420%\n",
      "Epoch 90/182\n",
      "Training loss: 0.132, Training accuracy: 95.271%\n",
      "Validation loss: 0.806, Validation accuracy: 80.740%\n",
      "Epoch 91/182\n",
      "Training loss: 0.134, Training accuracy: 95.187%\n",
      "Validation loss: 0.638, Validation accuracy: 83.200%\n",
      "Epoch 92/182\n",
      "Training loss: 0.128, Training accuracy: 95.547%\n",
      "Validation loss: 0.498, Validation accuracy: 85.220%\n",
      "Epoch 93/182\n",
      "Training loss: 0.127, Training accuracy: 95.596%\n",
      "Validation loss: 0.562, Validation accuracy: 84.600%\n",
      "Epoch 94/182\n",
      "Training loss: 0.126, Training accuracy: 95.573%\n",
      "Validation loss: 0.542, Validation accuracy: 84.500%\n",
      "Epoch 95/182\n",
      "Training loss: 0.132, Training accuracy: 95.364%\n",
      "Validation loss: 0.715, Validation accuracy: 82.520%\n",
      "Epoch 96/182\n",
      "Training loss: 0.124, Training accuracy: 95.720%\n",
      "Validation loss: 0.542, Validation accuracy: 85.140%\n",
      "Epoch 97/182\n",
      "Training loss: 0.122, Training accuracy: 95.684%\n",
      "Validation loss: 0.593, Validation accuracy: 83.520%\n",
      "Epoch 98/182\n",
      "Training loss: 0.126, Training accuracy: 95.509%\n",
      "Validation loss: 0.498, Validation accuracy: 85.820%\n",
      "Epoch 99/182\n",
      "Training loss: 0.128, Training accuracy: 95.398%\n",
      "Validation loss: 0.814, Validation accuracy: 80.160%\n",
      "Epoch 100/182\n",
      "Training loss: 0.133, Training accuracy: 95.269%\n",
      "Validation loss: 0.522, Validation accuracy: 84.980%\n",
      "Epoch 101/182\n",
      "Training loss: 0.126, Training accuracy: 95.540%\n",
      "Validation loss: 0.630, Validation accuracy: 84.080%\n",
      "Epoch 102/182\n",
      "Training loss: 0.128, Training accuracy: 95.464%\n",
      "Validation loss: 0.543, Validation accuracy: 85.460%\n",
      "Epoch 103/182\n",
      "Training loss: 0.122, Training accuracy: 95.620%\n",
      "Validation loss: 0.656, Validation accuracy: 82.340%\n",
      "Epoch 104/182\n",
      "Training loss: 0.119, Training accuracy: 95.762%\n",
      "Validation loss: 0.595, Validation accuracy: 84.760%\n",
      "Epoch 105/182\n",
      "Training loss: 0.123, Training accuracy: 95.698%\n",
      "Validation loss: 0.612, Validation accuracy: 84.300%\n",
      "Epoch 106/182\n",
      "Training loss: 0.123, Training accuracy: 95.593%\n",
      "Validation loss: 0.553, Validation accuracy: 84.060%\n",
      "Epoch 107/182\n",
      "Training loss: 0.125, Training accuracy: 95.571%\n",
      "Validation loss: 0.544, Validation accuracy: 84.680%\n",
      "Epoch 108/182\n",
      "Training loss: 0.122, Training accuracy: 95.733%\n",
      "Validation loss: 0.585, Validation accuracy: 83.960%\n",
      "Epoch 109/182\n",
      "Training loss: 0.119, Training accuracy: 95.749%\n",
      "Validation loss: 0.556, Validation accuracy: 85.120%\n",
      "Epoch 110/182\n",
      "Training loss: 0.124, Training accuracy: 95.602%\n",
      "Validation loss: 0.519, Validation accuracy: 85.320%\n",
      "Epoch 111/182\n",
      "Training loss: 0.117, Training accuracy: 95.798%\n",
      "Validation loss: 0.669, Validation accuracy: 82.260%\n",
      "Epoch 112/182\n",
      "Training loss: 0.122, Training accuracy: 95.689%\n",
      "Validation loss: 0.545, Validation accuracy: 84.700%\n",
      "Epoch 113/182\n",
      "Training loss: 0.123, Training accuracy: 95.713%\n",
      "Validation loss: 0.517, Validation accuracy: 85.440%\n",
      "Epoch 114/182\n",
      "Training loss: 0.125, Training accuracy: 95.462%\n",
      "Validation loss: 0.550, Validation accuracy: 85.020%\n",
      "Epoch 115/182\n",
      "Training loss: 0.124, Training accuracy: 95.687%\n",
      "Validation loss: 0.605, Validation accuracy: 83.500%\n",
      "Epoch 116/182\n",
      "Training loss: 0.123, Training accuracy: 95.671%\n",
      "Validation loss: 0.653, Validation accuracy: 82.860%\n",
      "Epoch 117/182\n",
      "Training loss: 0.115, Training accuracy: 95.956%\n",
      "Validation loss: 0.548, Validation accuracy: 85.500%\n",
      "Epoch 118/182\n",
      "Training loss: 0.125, Training accuracy: 95.653%\n",
      "Validation loss: 0.700, Validation accuracy: 81.420%\n",
      "Epoch 119/182\n",
      "Training loss: 0.124, Training accuracy: 95.673%\n",
      "Validation loss: 0.604, Validation accuracy: 83.000%\n",
      "Epoch 120/182\n",
      "Training loss: 0.119, Training accuracy: 95.827%\n",
      "Validation loss: 0.515, Validation accuracy: 85.320%\n",
      "Epoch 121/182\n",
      "Training loss: 0.123, Training accuracy: 95.618%\n",
      "Validation loss: 0.688, Validation accuracy: 80.560%\n",
      "Epoch 122/182\n",
      "Training loss: 0.118, Training accuracy: 95.860%\n",
      "Validation loss: 0.801, Validation accuracy: 79.640%\n",
      "Epoch 123/182\n",
      "Training loss: 0.116, Training accuracy: 95.987%\n",
      "Validation loss: 0.644, Validation accuracy: 84.060%\n",
      "Epoch 124/182\n",
      "Training loss: 0.121, Training accuracy: 95.780%\n",
      "Validation loss: 0.554, Validation accuracy: 85.020%\n",
      "Epoch 125/182\n",
      "Training loss: 0.125, Training accuracy: 95.547%\n",
      "Validation loss: 0.511, Validation accuracy: 85.400%\n",
      "Epoch 126/182\n",
      "Training loss: 0.113, Training accuracy: 96.047%\n",
      "Validation loss: 0.556, Validation accuracy: 85.500%\n",
      "Epoch 127/182\n",
      "Training loss: 0.123, Training accuracy: 95.627%\n",
      "Validation loss: 0.678, Validation accuracy: 83.760%\n",
      "Epoch 128/182\n",
      "Training loss: 0.119, Training accuracy: 95.807%\n",
      "Validation loss: 0.524, Validation accuracy: 85.560%\n",
      "Epoch 129/182\n",
      "Training loss: 0.124, Training accuracy: 95.567%\n",
      "Validation loss: 0.518, Validation accuracy: 85.000%\n",
      "Epoch 130/182\n",
      "Training loss: 0.117, Training accuracy: 95.878%\n",
      "Validation loss: 0.573, Validation accuracy: 85.020%\n",
      "Epoch 131/182\n",
      "Training loss: 0.123, Training accuracy: 95.756%\n",
      "Validation loss: 0.555, Validation accuracy: 84.280%\n",
      "Epoch 132/182\n",
      "Training loss: 0.120, Training accuracy: 95.804%\n",
      "Validation loss: 0.486, Validation accuracy: 86.620%\n",
      "Epoch 133/182\n",
      "Training loss: 0.120, Training accuracy: 95.824%\n",
      "Validation loss: 0.489, Validation accuracy: 86.020%\n",
      "Epoch 134/182\n",
      "Training loss: 0.117, Training accuracy: 95.884%\n",
      "Validation loss: 0.557, Validation accuracy: 85.080%\n",
      "Epoch 135/182\n",
      "Training loss: 0.123, Training accuracy: 95.622%\n",
      "Validation loss: 0.535, Validation accuracy: 84.100%\n",
      "Epoch 136/182\n",
      "Training loss: 0.120, Training accuracy: 95.733%\n",
      "Validation loss: 0.489, Validation accuracy: 86.520%\n",
      "Epoch 137/182\n",
      "Training loss: 0.119, Training accuracy: 95.789%\n",
      "Validation loss: 0.530, Validation accuracy: 84.860%\n",
      "Epoch 138/182\n",
      "Training loss: 0.117, Training accuracy: 95.836%\n",
      "Validation loss: 0.644, Validation accuracy: 84.320%\n",
      "Epoch 139/182\n",
      "Training loss: 0.122, Training accuracy: 95.720%\n",
      "Validation loss: 0.593, Validation accuracy: 84.980%\n",
      "Epoch 140/182\n",
      "Training loss: 0.116, Training accuracy: 96.022%\n",
      "Validation loss: 0.570, Validation accuracy: 85.200%\n",
      "Epoch 141/182\n",
      "Training loss: 0.124, Training accuracy: 95.527%\n",
      "Validation loss: 0.548, Validation accuracy: 84.200%\n",
      "Epoch 142/182\n",
      "Training loss: 0.114, Training accuracy: 96.007%\n",
      "Validation loss: 0.546, Validation accuracy: 84.820%\n",
      "Epoch 143/182\n",
      "Training loss: 0.118, Training accuracy: 95.771%\n",
      "Validation loss: 0.560, Validation accuracy: 85.340%\n",
      "Epoch 144/182\n",
      "Training loss: 0.118, Training accuracy: 95.791%\n",
      "Validation loss: 0.667, Validation accuracy: 83.440%\n",
      "Epoch 145/182\n",
      "Training loss: 0.122, Training accuracy: 95.709%\n",
      "Validation loss: 0.524, Validation accuracy: 85.280%\n",
      "Epoch 146/182\n",
      "Training loss: 0.120, Training accuracy: 95.742%\n",
      "Validation loss: 0.687, Validation accuracy: 83.300%\n",
      "Epoch 147/182\n",
      "Training loss: 0.117, Training accuracy: 95.869%\n",
      "Validation loss: 0.653, Validation accuracy: 81.780%\n",
      "Epoch 148/182\n",
      "Training loss: 0.117, Training accuracy: 95.798%\n",
      "Validation loss: 0.561, Validation accuracy: 85.080%\n",
      "Epoch 149/182\n",
      "Training loss: 0.114, Training accuracy: 95.824%\n",
      "Validation loss: 0.700, Validation accuracy: 83.080%\n",
      "Epoch 150/182\n",
      "Training loss: 0.114, Training accuracy: 95.976%\n",
      "Validation loss: 0.565, Validation accuracy: 84.540%\n",
      "Epoch 151/182\n",
      "Training loss: 0.127, Training accuracy: 95.540%\n",
      "Validation loss: 0.565, Validation accuracy: 85.020%\n",
      "Epoch 152/182\n",
      "Training loss: 0.118, Training accuracy: 95.847%\n",
      "Validation loss: 0.560, Validation accuracy: 84.560%\n",
      "Epoch 153/182\n",
      "Training loss: 0.115, Training accuracy: 95.904%\n",
      "Validation loss: 0.592, Validation accuracy: 83.760%\n",
      "Epoch 154/182\n",
      "Training loss: 0.122, Training accuracy: 95.678%\n",
      "Validation loss: 0.611, Validation accuracy: 83.100%\n",
      "Epoch 155/182\n",
      "Training loss: 0.110, Training accuracy: 96.191%\n",
      "Validation loss: 0.594, Validation accuracy: 84.000%\n",
      "Epoch 156/182\n",
      "Training loss: 0.109, Training accuracy: 96.064%\n",
      "Validation loss: 0.550, Validation accuracy: 85.040%\n",
      "Epoch 157/182\n",
      "Training loss: 0.109, Training accuracy: 96.136%\n",
      "Validation loss: 0.499, Validation accuracy: 86.520%\n",
      "Epoch 158/182\n",
      "Training loss: 0.121, Training accuracy: 95.807%\n",
      "Validation loss: 0.510, Validation accuracy: 85.800%\n",
      "Epoch 159/182\n",
      "Training loss: 0.112, Training accuracy: 96.009%\n",
      "Validation loss: 0.829, Validation accuracy: 79.840%\n",
      "Epoch 160/182\n",
      "Training loss: 0.117, Training accuracy: 95.838%\n",
      "Validation loss: 0.616, Validation accuracy: 83.420%\n",
      "Epoch 161/182\n",
      "Training loss: 0.117, Training accuracy: 95.944%\n",
      "Validation loss: 0.582, Validation accuracy: 83.380%\n",
      "Epoch 162/182\n",
      "Training loss: 0.119, Training accuracy: 95.782%\n",
      "Validation loss: 0.488, Validation accuracy: 85.640%\n",
      "Epoch 163/182\n",
      "Training loss: 0.114, Training accuracy: 95.969%\n",
      "Validation loss: 0.752, Validation accuracy: 81.340%\n",
      "Epoch 164/182\n",
      "Training loss: 0.118, Training accuracy: 95.842%\n",
      "Validation loss: 0.592, Validation accuracy: 83.680%\n",
      "Epoch 165/182\n",
      "Training loss: 0.106, Training accuracy: 96.289%\n",
      "Validation loss: 0.503, Validation accuracy: 86.140%\n",
      "Epoch 166/182\n",
      "Training loss: 0.119, Training accuracy: 95.771%\n",
      "Validation loss: 0.521, Validation accuracy: 85.500%\n",
      "Epoch 167/182\n",
      "Training loss: 0.109, Training accuracy: 96.216%\n",
      "Validation loss: 0.560, Validation accuracy: 85.140%\n",
      "Epoch 168/182\n",
      "Training loss: 0.117, Training accuracy: 95.820%\n",
      "Validation loss: 0.447, Validation accuracy: 87.100%\n",
      "Epoch 169/182\n",
      "Training loss: 0.111, Training accuracy: 96.122%\n",
      "Validation loss: 0.560, Validation accuracy: 84.480%\n",
      "Epoch 170/182\n",
      "Training loss: 0.112, Training accuracy: 96.073%\n",
      "Validation loss: 0.546, Validation accuracy: 84.780%\n",
      "Epoch 171/182\n",
      "Training loss: 0.119, Training accuracy: 95.871%\n",
      "Validation loss: 0.607, Validation accuracy: 84.020%\n",
      "Epoch 172/182\n",
      "Training loss: 0.117, Training accuracy: 95.836%\n",
      "Validation loss: 0.619, Validation accuracy: 84.180%\n",
      "Epoch 173/182\n",
      "Training loss: 0.114, Training accuracy: 95.987%\n",
      "Validation loss: 0.540, Validation accuracy: 85.440%\n",
      "Epoch 174/182\n",
      "Training loss: 0.112, Training accuracy: 96.131%\n",
      "Validation loss: 0.522, Validation accuracy: 84.980%\n",
      "Epoch 175/182\n",
      "Training loss: 0.107, Training accuracy: 96.262%\n",
      "Validation loss: 0.538, Validation accuracy: 85.740%\n",
      "Epoch 176/182\n",
      "Training loss: 0.122, Training accuracy: 95.762%\n",
      "Validation loss: 0.509, Validation accuracy: 86.500%\n",
      "Epoch 177/182\n",
      "Training loss: 0.112, Training accuracy: 96.131%\n",
      "Validation loss: 0.563, Validation accuracy: 84.980%\n",
      "Epoch 178/182\n",
      "Training loss: 0.110, Training accuracy: 96.131%\n",
      "Validation loss: 0.625, Validation accuracy: 83.300%\n",
      "Epoch 179/182\n",
      "Training loss: 0.110, Training accuracy: 96.227%\n",
      "Validation loss: 0.584, Validation accuracy: 85.040%\n",
      "Epoch 180/182\n",
      "Training loss: 0.116, Training accuracy: 95.840%\n",
      "Validation loss: 0.622, Validation accuracy: 82.980%\n",
      "Epoch 181/182\n",
      "Training loss: 0.117, Training accuracy: 95.858%\n",
      "Validation loss: 0.557, Validation accuracy: 85.660%\n",
      "Epoch 182/182\n",
      "Training loss: 0.110, Training accuracy: 96.149%\n",
      "Validation loss: 0.558, Validation accuracy: 85.500%\n",
      "Res Net 20 Finished Training\n",
      "Training time(seconds): 2494.4788975715637\n"
     ]
    }
   ],
   "source": [
    "num_epochs = 182\n",
    "\n",
    "train_losses_20 = []\n",
    "val_losses_20 = []\n",
    "train_accuracies_20 = []\n",
    "val_accuracies_20 = []\n",
    "\n",
    "train_losses_56 = []\n",
    "val_losses_56 = []\n",
    "train_accuracies_56 = []\n",
    "val_accuracies_56 = []\n",
    "\n",
    "\n",
    "## Train Res Net 20\n",
    "\n",
    "t0 = time.time()\n",
    "\n",
    "for epoch in range(num_epochs):\n",
    "    # Training\n",
    "    res_net20.train()  # Set the model to training mode\n",
    "    running_loss = 0.0\n",
    "    correct_train = 0\n",
    "    total_train = 0\n",
    "\n",
    "    if num_epochs == 92:\n",
    "\n",
    "        optimizer_20 = optim.SGD(res_net20.parameters(), lr = 0.01, momentum = 0.9, weight_decay = 0.0001)\n",
    "\n",
    "\n",
    "    for data in train_loader:                                     #enumerateでバッチに番号を振り、iでそれを引っ張ってくる。dataは一個ずつ引っ張り出してくるのではなく一気にバッチで学習する。\n",
    "        inputs, labels = data                                     #画像データ,正解ラベルのペアになっているのを分解\n",
    "        inputs, labels = inputs.to(device), labels.to(device)     #GPUに載せる\n",
    "        optimizer_20.zero_grad()                                     #勾配を0に設定\n",
    "        outputs = res_net20(inputs)                                     #インスタンス化したモデルに、入力を加えることで主力を得る\n",
    "        loss = criterion(outputs, labels)                         #指定した損失関数から損失を得る。この時点ではテンソルであることに注意。\n",
    "        loss.backward()                                           #得た損失のテンソルから勾配を計算\n",
    "        optimizer_20.step()                                          #得た勾配からパラメータを更新\n",
    "\n",
    "        running_loss += loss.item()                               #各データの損失を、テンソルから普通の数字(フロート)に変換し、epochごとの損失を得るために足し合わせる。\n",
    "        _, predicted = torch.max(outputs.data, 1)                 # 1:データの各クラス間 の最大値を、確率とその値を返す。\n",
    "        total_train += labels.size(0)                             # バッチに含まれているデータの数を継ぎ足す。\n",
    "        correct_train += (predicted == labels).sum().item()       # 正解した分のみを正答数に継ぎ足す。\n",
    "\n",
    "    train_losses_20.append(running_loss / len(train_loader))         # epochごとに、平均損失リストに加えていく。\n",
    "    train_accuracies_20.append(100 * correct_train / total_train)    # epochごとに正答率リストに加えていく。\n",
    "\n",
    "    # Validation\n",
    "    res_net20.eval()  # Set the model to evaluation mode\n",
    "    val_loss = 0.0\n",
    "    correct_val = 0\n",
    "    total_val = 0\n",
    "    with torch.no_grad():\n",
    "        for data in val_loader:\n",
    "            inputs, labels = data\n",
    "            inputs, labels = inputs.to(device), labels.to(device)\n",
    "            outputs = res_net20(inputs)\n",
    "            loss = criterion(outputs, labels)\n",
    "            val_loss += loss.item()\n",
    "            _, predicted = torch.max(outputs.data, 1)\n",
    "            total_val += labels.size(0)\n",
    "            correct_val += (predicted == labels).sum().item()\n",
    "\n",
    "    val_losses_20.append(val_loss / len(val_loader))\n",
    "    val_accuracies_20.append(100 * correct_val / total_val)\n",
    "\n",
    "    print(f'Epoch {epoch+1}/{num_epochs}')\n",
    "    print(f'Training loss: {train_losses_20[-1]:.3f}, Training accuracy: {train_accuracies_20[-1]:.3f}%')\n",
    "    print(f'Validation loss: {val_losses_20[-1]:.3f}, Validation accuracy: {val_accuracies_20[-1]:.3f}%')\n",
    "\n",
    "print('Res Net 20 Finished Training')\n",
    "print(\"Training time(seconds):\", time.time()-t0)\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f371aa1c",
   "metadata": {
    "papermill": {
     "duration": 0.021173,
     "end_time": "2023-12-04T11:05:07.777589",
     "exception": false,
     "start_time": "2023-12-04T11:05:07.756416",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h1>Res Net 56 学習</h1>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "c5a06830",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-12-04T11:05:07.822781Z",
     "iopub.status.busy": "2023-12-04T11:05:07.822380Z",
     "iopub.status.idle": "2023-12-04T12:33:11.460276Z",
     "shell.execute_reply": "2023-12-04T12:33:11.459055Z"
    },
    "papermill": {
     "duration": 5283.700611,
     "end_time": "2023-12-04T12:33:11.499604",
     "exception": false,
     "start_time": "2023-12-04T11:05:07.798993",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/182\n",
      "Training loss: 1.773, Training accuracy: 31.882%\n",
      "Validation loss: 1.585, Validation accuracy: 41.260%\n",
      "Epoch 2/182\n",
      "Training loss: 1.303, Training accuracy: 52.798%\n",
      "Validation loss: 1.153, Validation accuracy: 59.220%\n",
      "Epoch 3/182\n",
      "Training loss: 0.995, Training accuracy: 64.578%\n",
      "Validation loss: 0.908, Validation accuracy: 67.400%\n",
      "Epoch 4/182\n",
      "Training loss: 0.796, Training accuracy: 72.378%\n",
      "Validation loss: 0.714, Validation accuracy: 74.560%\n",
      "Epoch 5/182\n",
      "Training loss: 0.673, Training accuracy: 76.673%\n",
      "Validation loss: 0.766, Validation accuracy: 73.900%\n",
      "Epoch 6/182\n",
      "Training loss: 0.593, Training accuracy: 79.742%\n",
      "Validation loss: 0.683, Validation accuracy: 77.400%\n",
      "Epoch 7/182\n",
      "Training loss: 0.527, Training accuracy: 81.796%\n",
      "Validation loss: 0.597, Validation accuracy: 80.080%\n",
      "Epoch 8/182\n",
      "Training loss: 0.473, Training accuracy: 83.756%\n",
      "Validation loss: 0.691, Validation accuracy: 78.220%\n",
      "Epoch 9/182\n",
      "Training loss: 0.436, Training accuracy: 85.102%\n",
      "Validation loss: 0.528, Validation accuracy: 82.140%\n",
      "Epoch 10/182\n",
      "Training loss: 0.400, Training accuracy: 86.247%\n",
      "Validation loss: 0.615, Validation accuracy: 79.920%\n",
      "Epoch 11/182\n",
      "Training loss: 0.379, Training accuracy: 86.984%\n",
      "Validation loss: 0.568, Validation accuracy: 81.240%\n",
      "Epoch 12/182\n",
      "Training loss: 0.358, Training accuracy: 87.720%\n",
      "Validation loss: 0.481, Validation accuracy: 83.340%\n",
      "Epoch 13/182\n",
      "Training loss: 0.335, Training accuracy: 88.342%\n",
      "Validation loss: 0.505, Validation accuracy: 83.000%\n",
      "Epoch 14/182\n",
      "Training loss: 0.319, Training accuracy: 88.829%\n",
      "Validation loss: 0.473, Validation accuracy: 84.200%\n",
      "Epoch 15/182\n",
      "Training loss: 0.306, Training accuracy: 89.540%\n",
      "Validation loss: 0.507, Validation accuracy: 83.660%\n",
      "Epoch 16/182\n",
      "Training loss: 0.287, Training accuracy: 90.038%\n",
      "Validation loss: 0.488, Validation accuracy: 83.820%\n",
      "Epoch 17/182\n",
      "Training loss: 0.276, Training accuracy: 90.444%\n",
      "Validation loss: 0.501, Validation accuracy: 84.340%\n",
      "Epoch 18/182\n",
      "Training loss: 0.267, Training accuracy: 90.756%\n",
      "Validation loss: 0.536, Validation accuracy: 83.340%\n",
      "Epoch 19/182\n",
      "Training loss: 0.256, Training accuracy: 91.047%\n",
      "Validation loss: 0.506, Validation accuracy: 83.880%\n",
      "Epoch 20/182\n",
      "Training loss: 0.249, Training accuracy: 91.267%\n",
      "Validation loss: 0.546, Validation accuracy: 83.520%\n",
      "Epoch 21/182\n",
      "Training loss: 0.242, Training accuracy: 91.536%\n",
      "Validation loss: 0.552, Validation accuracy: 82.880%\n",
      "Epoch 22/182\n",
      "Training loss: 0.235, Training accuracy: 91.862%\n",
      "Validation loss: 0.477, Validation accuracy: 84.820%\n",
      "Epoch 23/182\n",
      "Training loss: 0.230, Training accuracy: 91.967%\n",
      "Validation loss: 0.592, Validation accuracy: 82.240%\n",
      "Epoch 24/182\n",
      "Training loss: 0.225, Training accuracy: 92.149%\n",
      "Validation loss: 0.550, Validation accuracy: 83.340%\n",
      "Epoch 25/182\n",
      "Training loss: 0.213, Training accuracy: 92.489%\n",
      "Validation loss: 0.548, Validation accuracy: 83.460%\n",
      "Epoch 26/182\n",
      "Training loss: 0.212, Training accuracy: 92.627%\n",
      "Validation loss: 0.540, Validation accuracy: 83.920%\n",
      "Epoch 27/182\n",
      "Training loss: 0.203, Training accuracy: 92.996%\n",
      "Validation loss: 0.543, Validation accuracy: 84.120%\n",
      "Epoch 28/182\n",
      "Training loss: 0.200, Training accuracy: 93.187%\n",
      "Validation loss: 0.484, Validation accuracy: 85.140%\n",
      "Epoch 29/182\n",
      "Training loss: 0.188, Training accuracy: 93.564%\n",
      "Validation loss: 0.626, Validation accuracy: 82.280%\n",
      "Epoch 30/182\n",
      "Training loss: 0.199, Training accuracy: 93.022%\n",
      "Validation loss: 0.475, Validation accuracy: 85.000%\n",
      "Epoch 31/182\n",
      "Training loss: 0.184, Training accuracy: 93.498%\n",
      "Validation loss: 0.585, Validation accuracy: 83.020%\n",
      "Epoch 32/182\n",
      "Training loss: 0.190, Training accuracy: 93.336%\n",
      "Validation loss: 0.698, Validation accuracy: 80.880%\n",
      "Epoch 33/182\n",
      "Training loss: 0.177, Training accuracy: 93.793%\n",
      "Validation loss: 0.541, Validation accuracy: 84.040%\n",
      "Epoch 34/182\n",
      "Training loss: 0.177, Training accuracy: 93.856%\n",
      "Validation loss: 0.657, Validation accuracy: 80.960%\n",
      "Epoch 35/182\n",
      "Training loss: 0.169, Training accuracy: 94.198%\n",
      "Validation loss: 0.534, Validation accuracy: 85.280%\n",
      "Epoch 36/182\n",
      "Training loss: 0.167, Training accuracy: 94.084%\n",
      "Validation loss: 0.506, Validation accuracy: 84.940%\n",
      "Epoch 37/182\n",
      "Training loss: 0.163, Training accuracy: 94.387%\n",
      "Validation loss: 0.488, Validation accuracy: 86.020%\n",
      "Epoch 38/182\n",
      "Training loss: 0.170, Training accuracy: 94.084%\n",
      "Validation loss: 0.545, Validation accuracy: 84.520%\n",
      "Epoch 39/182\n",
      "Training loss: 0.161, Training accuracy: 94.347%\n",
      "Validation loss: 0.521, Validation accuracy: 84.880%\n",
      "Epoch 40/182\n",
      "Training loss: 0.159, Training accuracy: 94.458%\n",
      "Validation loss: 0.733, Validation accuracy: 81.780%\n",
      "Epoch 41/182\n",
      "Training loss: 0.159, Training accuracy: 94.484%\n",
      "Validation loss: 0.513, Validation accuracy: 84.860%\n",
      "Epoch 42/182\n",
      "Training loss: 0.154, Training accuracy: 94.478%\n",
      "Validation loss: 0.511, Validation accuracy: 85.260%\n",
      "Epoch 43/182\n",
      "Training loss: 0.163, Training accuracy: 94.389%\n",
      "Validation loss: 0.650, Validation accuracy: 81.660%\n",
      "Epoch 44/182\n",
      "Training loss: 0.153, Training accuracy: 94.667%\n",
      "Validation loss: 0.715, Validation accuracy: 81.260%\n",
      "Epoch 45/182\n",
      "Training loss: 0.160, Training accuracy: 94.360%\n",
      "Validation loss: 0.528, Validation accuracy: 84.660%\n",
      "Epoch 46/182\n",
      "Training loss: 0.154, Training accuracy: 94.653%\n",
      "Validation loss: 0.574, Validation accuracy: 84.020%\n",
      "Epoch 47/182\n",
      "Training loss: 0.149, Training accuracy: 94.818%\n",
      "Validation loss: 0.556, Validation accuracy: 84.680%\n",
      "Epoch 48/182\n",
      "Training loss: 0.154, Training accuracy: 94.702%\n",
      "Validation loss: 0.498, Validation accuracy: 85.920%\n",
      "Epoch 49/182\n",
      "Training loss: 0.141, Training accuracy: 95.093%\n",
      "Validation loss: 0.568, Validation accuracy: 84.000%\n",
      "Epoch 50/182\n",
      "Training loss: 0.139, Training accuracy: 95.200%\n",
      "Validation loss: 0.538, Validation accuracy: 85.180%\n",
      "Epoch 51/182\n",
      "Training loss: 0.143, Training accuracy: 95.167%\n",
      "Validation loss: 0.532, Validation accuracy: 85.460%\n",
      "Epoch 52/182\n",
      "Training loss: 0.146, Training accuracy: 94.900%\n",
      "Validation loss: 0.589, Validation accuracy: 84.960%\n",
      "Epoch 53/182\n",
      "Training loss: 0.138, Training accuracy: 95.216%\n",
      "Validation loss: 0.535, Validation accuracy: 85.160%\n",
      "Epoch 54/182\n",
      "Training loss: 0.130, Training accuracy: 95.427%\n",
      "Validation loss: 0.596, Validation accuracy: 83.400%\n",
      "Epoch 55/182\n",
      "Training loss: 0.144, Training accuracy: 95.020%\n",
      "Validation loss: 0.599, Validation accuracy: 84.420%\n",
      "Epoch 56/182\n",
      "Training loss: 0.137, Training accuracy: 95.191%\n",
      "Validation loss: 0.533, Validation accuracy: 85.640%\n",
      "Epoch 57/182\n",
      "Training loss: 0.137, Training accuracy: 95.253%\n",
      "Validation loss: 0.555, Validation accuracy: 84.880%\n",
      "Epoch 58/182\n",
      "Training loss: 0.130, Training accuracy: 95.409%\n",
      "Validation loss: 0.584, Validation accuracy: 84.960%\n",
      "Epoch 59/182\n",
      "Training loss: 0.135, Training accuracy: 95.289%\n",
      "Validation loss: 0.552, Validation accuracy: 84.420%\n",
      "Epoch 60/182\n",
      "Training loss: 0.141, Training accuracy: 95.093%\n",
      "Validation loss: 0.514, Validation accuracy: 85.000%\n",
      "Epoch 61/182\n",
      "Training loss: 0.131, Training accuracy: 95.538%\n",
      "Validation loss: 0.651, Validation accuracy: 83.060%\n",
      "Epoch 62/182\n",
      "Training loss: 0.130, Training accuracy: 95.420%\n",
      "Validation loss: 0.545, Validation accuracy: 85.460%\n",
      "Epoch 63/182\n",
      "Training loss: 0.127, Training accuracy: 95.640%\n",
      "Validation loss: 0.553, Validation accuracy: 85.380%\n",
      "Epoch 64/182\n",
      "Training loss: 0.130, Training accuracy: 95.400%\n",
      "Validation loss: 0.530, Validation accuracy: 86.220%\n",
      "Epoch 65/182\n",
      "Training loss: 0.132, Training accuracy: 95.438%\n",
      "Validation loss: 0.677, Validation accuracy: 83.080%\n",
      "Epoch 66/182\n",
      "Training loss: 0.139, Training accuracy: 95.209%\n",
      "Validation loss: 0.474, Validation accuracy: 86.500%\n",
      "Epoch 67/182\n",
      "Training loss: 0.125, Training accuracy: 95.653%\n",
      "Validation loss: 0.588, Validation accuracy: 84.360%\n",
      "Epoch 68/182\n",
      "Training loss: 0.132, Training accuracy: 95.480%\n",
      "Validation loss: 0.553, Validation accuracy: 84.900%\n",
      "Epoch 69/182\n",
      "Training loss: 0.125, Training accuracy: 95.722%\n",
      "Validation loss: 0.563, Validation accuracy: 84.920%\n",
      "Epoch 70/182\n",
      "Training loss: 0.131, Training accuracy: 95.529%\n",
      "Validation loss: 0.545, Validation accuracy: 85.860%\n",
      "Epoch 71/182\n",
      "Training loss: 0.124, Training accuracy: 95.702%\n",
      "Validation loss: 0.596, Validation accuracy: 85.280%\n",
      "Epoch 72/182\n",
      "Training loss: 0.122, Training accuracy: 95.756%\n",
      "Validation loss: 0.565, Validation accuracy: 85.820%\n",
      "Epoch 73/182\n",
      "Training loss: 0.125, Training accuracy: 95.618%\n",
      "Validation loss: 0.548, Validation accuracy: 85.500%\n",
      "Epoch 74/182\n",
      "Training loss: 0.126, Training accuracy: 95.667%\n",
      "Validation loss: 0.697, Validation accuracy: 82.420%\n",
      "Epoch 75/182\n",
      "Training loss: 0.125, Training accuracy: 95.602%\n",
      "Validation loss: 0.603, Validation accuracy: 83.900%\n",
      "Epoch 76/182\n",
      "Training loss: 0.123, Training accuracy: 95.669%\n",
      "Validation loss: 0.550, Validation accuracy: 85.060%\n",
      "Epoch 77/182\n",
      "Training loss: 0.127, Training accuracy: 95.473%\n",
      "Validation loss: 0.535, Validation accuracy: 86.060%\n",
      "Epoch 78/182\n",
      "Training loss: 0.120, Training accuracy: 95.831%\n",
      "Validation loss: 0.585, Validation accuracy: 85.260%\n",
      "Epoch 79/182\n",
      "Training loss: 0.125, Training accuracy: 95.676%\n",
      "Validation loss: 0.639, Validation accuracy: 84.260%\n",
      "Epoch 80/182\n",
      "Training loss: 0.127, Training accuracy: 95.538%\n",
      "Validation loss: 0.550, Validation accuracy: 85.640%\n",
      "Epoch 81/182\n",
      "Training loss: 0.122, Training accuracy: 95.707%\n",
      "Validation loss: 0.495, Validation accuracy: 86.000%\n",
      "Epoch 82/182\n",
      "Training loss: 0.114, Training accuracy: 96.069%\n",
      "Validation loss: 0.578, Validation accuracy: 84.620%\n",
      "Epoch 83/182\n",
      "Training loss: 0.119, Training accuracy: 95.938%\n",
      "Validation loss: 0.541, Validation accuracy: 85.660%\n",
      "Epoch 84/182\n",
      "Training loss: 0.124, Training accuracy: 95.698%\n",
      "Validation loss: 0.559, Validation accuracy: 85.340%\n",
      "Epoch 85/182\n",
      "Training loss: 0.120, Training accuracy: 95.747%\n",
      "Validation loss: 0.527, Validation accuracy: 85.580%\n",
      "Epoch 86/182\n",
      "Training loss: 0.122, Training accuracy: 95.751%\n",
      "Validation loss: 0.518, Validation accuracy: 85.800%\n",
      "Epoch 87/182\n",
      "Training loss: 0.125, Training accuracy: 95.718%\n",
      "Validation loss: 0.537, Validation accuracy: 84.840%\n",
      "Epoch 88/182\n",
      "Training loss: 0.115, Training accuracy: 95.993%\n",
      "Validation loss: 0.539, Validation accuracy: 85.560%\n",
      "Epoch 89/182\n",
      "Training loss: 0.116, Training accuracy: 95.982%\n",
      "Validation loss: 0.534, Validation accuracy: 85.440%\n",
      "Epoch 90/182\n",
      "Training loss: 0.122, Training accuracy: 95.836%\n",
      "Validation loss: 0.552, Validation accuracy: 85.720%\n",
      "Epoch 91/182\n",
      "Training loss: 0.122, Training accuracy: 95.849%\n",
      "Validation loss: 0.489, Validation accuracy: 85.720%\n",
      "Epoch 92/182\n",
      "Training loss: 0.120, Training accuracy: 95.840%\n",
      "Validation loss: 0.502, Validation accuracy: 86.480%\n",
      "Epoch 93/182\n",
      "Training loss: 0.109, Training accuracy: 96.162%\n",
      "Validation loss: 0.543, Validation accuracy: 85.480%\n",
      "Epoch 94/182\n",
      "Training loss: 0.116, Training accuracy: 96.056%\n",
      "Validation loss: 0.499, Validation accuracy: 86.700%\n",
      "Epoch 95/182\n",
      "Training loss: 0.118, Training accuracy: 95.862%\n",
      "Validation loss: 0.522, Validation accuracy: 85.740%\n",
      "Epoch 96/182\n",
      "Training loss: 0.117, Training accuracy: 95.922%\n",
      "Validation loss: 0.626, Validation accuracy: 84.200%\n",
      "Epoch 97/182\n",
      "Training loss: 0.115, Training accuracy: 96.051%\n",
      "Validation loss: 0.543, Validation accuracy: 85.600%\n",
      "Epoch 98/182\n",
      "Training loss: 0.111, Training accuracy: 96.178%\n",
      "Validation loss: 0.661, Validation accuracy: 82.620%\n",
      "Epoch 99/182\n",
      "Training loss: 0.116, Training accuracy: 95.880%\n",
      "Validation loss: 0.500, Validation accuracy: 86.540%\n",
      "Epoch 100/182\n",
      "Training loss: 0.113, Training accuracy: 95.987%\n",
      "Validation loss: 0.546, Validation accuracy: 85.080%\n",
      "Epoch 101/182\n",
      "Training loss: 0.116, Training accuracy: 96.098%\n",
      "Validation loss: 0.527, Validation accuracy: 85.960%\n",
      "Epoch 102/182\n",
      "Training loss: 0.116, Training accuracy: 95.980%\n",
      "Validation loss: 0.533, Validation accuracy: 85.640%\n",
      "Epoch 103/182\n",
      "Training loss: 0.119, Training accuracy: 95.931%\n",
      "Validation loss: 0.671, Validation accuracy: 82.900%\n",
      "Epoch 104/182\n",
      "Training loss: 0.117, Training accuracy: 95.900%\n",
      "Validation loss: 0.599, Validation accuracy: 84.400%\n",
      "Epoch 105/182\n",
      "Training loss: 0.107, Training accuracy: 96.329%\n",
      "Validation loss: 0.548, Validation accuracy: 84.820%\n",
      "Epoch 106/182\n",
      "Training loss: 0.124, Training accuracy: 95.667%\n",
      "Validation loss: 0.517, Validation accuracy: 86.220%\n",
      "Epoch 107/182\n",
      "Training loss: 0.112, Training accuracy: 96.007%\n",
      "Validation loss: 0.528, Validation accuracy: 85.680%\n",
      "Epoch 108/182\n",
      "Training loss: 0.111, Training accuracy: 96.042%\n",
      "Validation loss: 0.604, Validation accuracy: 85.580%\n",
      "Epoch 109/182\n",
      "Training loss: 0.123, Training accuracy: 95.864%\n",
      "Validation loss: 0.555, Validation accuracy: 85.500%\n",
      "Epoch 110/182\n",
      "Training loss: 0.103, Training accuracy: 96.469%\n",
      "Validation loss: 0.538, Validation accuracy: 85.880%\n",
      "Epoch 111/182\n",
      "Training loss: 0.112, Training accuracy: 96.182%\n",
      "Validation loss: 0.545, Validation accuracy: 85.620%\n",
      "Epoch 112/182\n",
      "Training loss: 0.112, Training accuracy: 96.053%\n",
      "Validation loss: 0.554, Validation accuracy: 85.880%\n",
      "Epoch 113/182\n",
      "Training loss: 0.118, Training accuracy: 95.973%\n",
      "Validation loss: 0.600, Validation accuracy: 83.940%\n",
      "Epoch 114/182\n",
      "Training loss: 0.110, Training accuracy: 96.196%\n",
      "Validation loss: 0.487, Validation accuracy: 86.800%\n",
      "Epoch 115/182\n",
      "Training loss: 0.111, Training accuracy: 96.256%\n",
      "Validation loss: 0.540, Validation accuracy: 85.660%\n",
      "Epoch 116/182\n",
      "Training loss: 0.114, Training accuracy: 96.071%\n",
      "Validation loss: 0.576, Validation accuracy: 85.140%\n",
      "Epoch 117/182\n",
      "Training loss: 0.112, Training accuracy: 96.127%\n",
      "Validation loss: 0.579, Validation accuracy: 85.120%\n",
      "Epoch 118/182\n",
      "Training loss: 0.112, Training accuracy: 96.138%\n",
      "Validation loss: 0.530, Validation accuracy: 85.680%\n",
      "Epoch 119/182\n",
      "Training loss: 0.114, Training accuracy: 96.049%\n",
      "Validation loss: 0.606, Validation accuracy: 84.700%\n",
      "Epoch 120/182\n",
      "Training loss: 0.110, Training accuracy: 96.144%\n",
      "Validation loss: 0.533, Validation accuracy: 85.420%\n",
      "Epoch 121/182\n",
      "Training loss: 0.110, Training accuracy: 96.124%\n",
      "Validation loss: 0.715, Validation accuracy: 82.560%\n",
      "Epoch 122/182\n",
      "Training loss: 0.111, Training accuracy: 96.160%\n",
      "Validation loss: 0.536, Validation accuracy: 85.760%\n",
      "Epoch 123/182\n",
      "Training loss: 0.107, Training accuracy: 96.349%\n",
      "Validation loss: 0.498, Validation accuracy: 86.560%\n",
      "Epoch 124/182\n",
      "Training loss: 0.115, Training accuracy: 96.140%\n",
      "Validation loss: 0.502, Validation accuracy: 86.340%\n",
      "Epoch 125/182\n",
      "Training loss: 0.113, Training accuracy: 96.124%\n",
      "Validation loss: 0.455, Validation accuracy: 87.140%\n",
      "Epoch 126/182\n",
      "Training loss: 0.109, Training accuracy: 96.213%\n",
      "Validation loss: 0.547, Validation accuracy: 85.960%\n",
      "Epoch 127/182\n",
      "Training loss: 0.112, Training accuracy: 96.107%\n",
      "Validation loss: 0.474, Validation accuracy: 86.440%\n",
      "Epoch 128/182\n",
      "Training loss: 0.111, Training accuracy: 96.142%\n",
      "Validation loss: 0.517, Validation accuracy: 85.580%\n",
      "Epoch 129/182\n",
      "Training loss: 0.101, Training accuracy: 96.556%\n",
      "Validation loss: 0.667, Validation accuracy: 84.040%\n",
      "Epoch 130/182\n",
      "Training loss: 0.110, Training accuracy: 96.269%\n",
      "Validation loss: 0.483, Validation accuracy: 86.580%\n",
      "Epoch 131/182\n",
      "Training loss: 0.113, Training accuracy: 96.091%\n",
      "Validation loss: 0.520, Validation accuracy: 85.700%\n",
      "Epoch 132/182\n",
      "Training loss: 0.108, Training accuracy: 96.262%\n",
      "Validation loss: 0.516, Validation accuracy: 85.720%\n",
      "Epoch 133/182\n",
      "Training loss: 0.108, Training accuracy: 96.367%\n",
      "Validation loss: 0.564, Validation accuracy: 84.680%\n",
      "Epoch 134/182\n",
      "Training loss: 0.115, Training accuracy: 96.164%\n",
      "Validation loss: 0.505, Validation accuracy: 85.920%\n",
      "Epoch 135/182\n",
      "Training loss: 0.109, Training accuracy: 96.276%\n",
      "Validation loss: 0.532, Validation accuracy: 85.240%\n",
      "Epoch 136/182\n",
      "Training loss: 0.107, Training accuracy: 96.233%\n",
      "Validation loss: 0.601, Validation accuracy: 84.380%\n",
      "Epoch 137/182\n",
      "Training loss: 0.106, Training accuracy: 96.340%\n",
      "Validation loss: 0.485, Validation accuracy: 86.260%\n",
      "Epoch 138/182\n",
      "Training loss: 0.108, Training accuracy: 96.289%\n",
      "Validation loss: 0.564, Validation accuracy: 84.780%\n",
      "Epoch 139/182\n",
      "Training loss: 0.106, Training accuracy: 96.231%\n",
      "Validation loss: 0.555, Validation accuracy: 85.380%\n",
      "Epoch 140/182\n",
      "Training loss: 0.110, Training accuracy: 96.140%\n",
      "Validation loss: 0.503, Validation accuracy: 86.300%\n",
      "Epoch 141/182\n",
      "Training loss: 0.106, Training accuracy: 96.362%\n",
      "Validation loss: 0.524, Validation accuracy: 86.100%\n",
      "Epoch 142/182\n",
      "Training loss: 0.111, Training accuracy: 96.264%\n",
      "Validation loss: 0.558, Validation accuracy: 84.260%\n",
      "Epoch 143/182\n",
      "Training loss: 0.103, Training accuracy: 96.442%\n",
      "Validation loss: 0.537, Validation accuracy: 84.640%\n",
      "Epoch 144/182\n",
      "Training loss: 0.104, Training accuracy: 96.422%\n",
      "Validation loss: 0.613, Validation accuracy: 85.320%\n",
      "Epoch 145/182\n",
      "Training loss: 0.106, Training accuracy: 96.402%\n",
      "Validation loss: 0.595, Validation accuracy: 85.400%\n",
      "Epoch 146/182\n",
      "Training loss: 0.106, Training accuracy: 96.420%\n",
      "Validation loss: 0.553, Validation accuracy: 84.680%\n",
      "Epoch 147/182\n",
      "Training loss: 0.105, Training accuracy: 96.440%\n",
      "Validation loss: 0.602, Validation accuracy: 84.480%\n",
      "Epoch 148/182\n",
      "Training loss: 0.107, Training accuracy: 96.298%\n",
      "Validation loss: 0.539, Validation accuracy: 85.880%\n",
      "Epoch 149/182\n",
      "Training loss: 0.106, Training accuracy: 96.360%\n",
      "Validation loss: 0.486, Validation accuracy: 87.220%\n",
      "Epoch 150/182\n",
      "Training loss: 0.104, Training accuracy: 96.389%\n",
      "Validation loss: 0.583, Validation accuracy: 85.420%\n",
      "Epoch 151/182\n",
      "Training loss: 0.109, Training accuracy: 96.122%\n",
      "Validation loss: 0.605, Validation accuracy: 84.780%\n",
      "Epoch 152/182\n",
      "Training loss: 0.105, Training accuracy: 96.384%\n",
      "Validation loss: 0.569, Validation accuracy: 84.220%\n",
      "Epoch 153/182\n",
      "Training loss: 0.108, Training accuracy: 96.331%\n",
      "Validation loss: 0.539, Validation accuracy: 86.300%\n",
      "Epoch 154/182\n",
      "Training loss: 0.109, Training accuracy: 96.224%\n",
      "Validation loss: 0.583, Validation accuracy: 84.960%\n",
      "Epoch 155/182\n",
      "Training loss: 0.105, Training accuracy: 96.371%\n",
      "Validation loss: 0.511, Validation accuracy: 86.860%\n",
      "Epoch 156/182\n",
      "Training loss: 0.109, Training accuracy: 96.238%\n",
      "Validation loss: 0.518, Validation accuracy: 86.520%\n",
      "Epoch 157/182\n",
      "Training loss: 0.103, Training accuracy: 96.540%\n",
      "Validation loss: 0.496, Validation accuracy: 85.980%\n",
      "Epoch 158/182\n",
      "Training loss: 0.109, Training accuracy: 96.122%\n",
      "Validation loss: 0.519, Validation accuracy: 85.340%\n",
      "Epoch 159/182\n",
      "Training loss: 0.104, Training accuracy: 96.436%\n",
      "Validation loss: 0.531, Validation accuracy: 85.720%\n",
      "Epoch 160/182\n",
      "Training loss: 0.102, Training accuracy: 96.480%\n",
      "Validation loss: 0.602, Validation accuracy: 84.620%\n",
      "Epoch 161/182\n",
      "Training loss: 0.108, Training accuracy: 96.309%\n",
      "Validation loss: 0.473, Validation accuracy: 86.900%\n",
      "Epoch 162/182\n",
      "Training loss: 0.103, Training accuracy: 96.442%\n",
      "Validation loss: 0.546, Validation accuracy: 86.100%\n",
      "Epoch 163/182\n",
      "Training loss: 0.101, Training accuracy: 96.507%\n",
      "Validation loss: 0.534, Validation accuracy: 86.500%\n",
      "Epoch 164/182\n",
      "Training loss: 0.101, Training accuracy: 96.420%\n",
      "Validation loss: 0.552, Validation accuracy: 86.420%\n",
      "Epoch 165/182\n",
      "Training loss: 0.108, Training accuracy: 96.211%\n",
      "Validation loss: 0.527, Validation accuracy: 86.080%\n",
      "Epoch 166/182\n",
      "Training loss: 0.109, Training accuracy: 96.244%\n",
      "Validation loss: 0.626, Validation accuracy: 84.660%\n",
      "Epoch 167/182\n",
      "Training loss: 0.103, Training accuracy: 96.429%\n",
      "Validation loss: 0.487, Validation accuracy: 86.580%\n",
      "Epoch 168/182\n",
      "Training loss: 0.105, Training accuracy: 96.362%\n",
      "Validation loss: 0.601, Validation accuracy: 85.240%\n",
      "Epoch 169/182\n",
      "Training loss: 0.112, Training accuracy: 96.100%\n",
      "Validation loss: 0.464, Validation accuracy: 86.700%\n",
      "Epoch 170/182\n",
      "Training loss: 0.099, Training accuracy: 96.724%\n",
      "Validation loss: 0.491, Validation accuracy: 86.520%\n",
      "Epoch 171/182\n",
      "Training loss: 0.105, Training accuracy: 96.422%\n",
      "Validation loss: 0.471, Validation accuracy: 86.740%\n",
      "Epoch 172/182\n",
      "Training loss: 0.102, Training accuracy: 96.451%\n",
      "Validation loss: 0.557, Validation accuracy: 86.580%\n",
      "Epoch 173/182\n",
      "Training loss: 0.101, Training accuracy: 96.436%\n",
      "Validation loss: 0.636, Validation accuracy: 85.080%\n",
      "Epoch 174/182\n",
      "Training loss: 0.102, Training accuracy: 96.418%\n",
      "Validation loss: 0.477, Validation accuracy: 86.060%\n",
      "Epoch 175/182\n",
      "Training loss: 0.101, Training accuracy: 96.556%\n",
      "Validation loss: 0.493, Validation accuracy: 85.780%\n",
      "Epoch 176/182\n",
      "Training loss: 0.102, Training accuracy: 96.396%\n",
      "Validation loss: 0.480, Validation accuracy: 86.520%\n",
      "Epoch 177/182\n",
      "Training loss: 0.098, Training accuracy: 96.642%\n",
      "Validation loss: 0.521, Validation accuracy: 85.460%\n",
      "Epoch 178/182\n",
      "Training loss: 0.107, Training accuracy: 96.202%\n",
      "Validation loss: 0.571, Validation accuracy: 84.560%\n",
      "Epoch 179/182\n",
      "Training loss: 0.104, Training accuracy: 96.373%\n",
      "Validation loss: 0.541, Validation accuracy: 86.620%\n",
      "Epoch 180/182\n",
      "Training loss: 0.099, Training accuracy: 96.547%\n",
      "Validation loss: 0.587, Validation accuracy: 84.700%\n",
      "Epoch 181/182\n",
      "Training loss: 0.106, Training accuracy: 96.276%\n",
      "Validation loss: 0.435, Validation accuracy: 87.600%\n",
      "Epoch 182/182\n",
      "Training loss: 0.106, Training accuracy: 96.393%\n",
      "Validation loss: 0.526, Validation accuracy: 86.360%\n",
      "Plain Net 56 Finished Training\n",
      "Training time(seconds): 5283.623261213303\n"
     ]
    }
   ],
   "source": [
    "## Train Res Net 56\n",
    "\n",
    "t0 = time.time()\n",
    "\n",
    "for epoch in range(num_epochs):\n",
    "    # Training\n",
    "    res_net56.train()  # Set the model to training mode\n",
    "    running_loss = 0.0\n",
    "    correct_train = 0\n",
    "    total_train = 0\n",
    "\n",
    "    if num_epochs == 92:\n",
    "\n",
    "        optimizer_56 = optim.SGD(res_net56.parameters(), lr = 0.01, momentum = 0.9, weight_decay = 0.0001)\n",
    "\n",
    "\n",
    "    for data in train_loader:                                     #enumerateでバッチに番号を振り、iでそれを引っ張ってくる。dataは一個ずつ引っ張り出してくるのではなく一気にバッチで学習する。\n",
    "        inputs, labels = data                                     #画像データ,正解ラベルのペアになっているのを分解\n",
    "        inputs, labels = inputs.to(device), labels.to(device)     #GPUに載せる\n",
    "        optimizer_56.zero_grad()                                     #勾配を0に設定\n",
    "        outputs = res_net56(inputs)                                     #インスタンス化したモデルに、入力を加えることで主力を得る\n",
    "        loss = criterion(outputs, labels)                         #指定した損失関数から損失を得る。この時点ではテンソルであることに注意。\n",
    "        loss.backward()                                           #得た損失のテンソルから勾配を計算\n",
    "        optimizer_56.step()                                          #得た勾配からパラメータを更新\n",
    "\n",
    "        running_loss += loss.item()                               #各データの損失を、テンソルから普通の数字(フロート)に変換し、epochごとの損失を得るために足し合わせる。\n",
    "        _, predicted = torch.max(outputs.data, 1)                 # 1:データの各クラス間 の最大値を、確率とその値を返す。\n",
    "        total_train += labels.size(0)                             # バッチに含まれているデータの数を継ぎ足す。\n",
    "        correct_train += (predicted == labels).sum().item()       # 正解した分のみを正答数に継ぎ足す。\n",
    "\n",
    "    train_losses_56.append(running_loss / len(train_loader))         # epochごとに、平均損失リストに加えていく。\n",
    "    train_accuracies_56.append(100 * correct_train / total_train)    # epochごとに正答率リストに加えていく。\n",
    "\n",
    "    # Validation\n",
    "    res_net56.eval()  # Set the model to evaluation mode\n",
    "    val_loss = 0.0\n",
    "    correct_val = 0\n",
    "    total_val = 0\n",
    "    with torch.no_grad():\n",
    "        for data in val_loader:\n",
    "            inputs, labels = data\n",
    "            inputs, labels = inputs.to(device), labels.to(device)\n",
    "            outputs = res_net56(inputs)\n",
    "            loss = criterion(outputs, labels)\n",
    "            val_loss += loss.item()\n",
    "            _, predicted = torch.max(outputs.data, 1)\n",
    "            total_val += labels.size(0)\n",
    "            correct_val += (predicted == labels).sum().item()\n",
    "\n",
    "    val_losses_56.append(val_loss / len(val_loader))\n",
    "    val_accuracies_56.append(100 * correct_val / total_val)\n",
    "\n",
    "    print(f'Epoch {epoch+1}/{num_epochs}')\n",
    "    print(f'Training loss: {train_losses_56[-1]:.3f}, Training accuracy: {train_accuracies_56[-1]:.3f}%')\n",
    "    print(f'Validation loss: {val_losses_56[-1]:.3f}, Validation accuracy: {val_accuracies_56[-1]:.3f}%')\n",
    "\n",
    "print('Plain Net 56 Finished Training')\n",
    "print(\"Training time(seconds):\", time.time()-t0)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b5c32f80",
   "metadata": {
    "papermill": {
     "duration": 0.036091,
     "end_time": "2023-12-04T12:33:11.572812",
     "exception": false,
     "start_time": "2023-12-04T12:33:11.536721",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h1>結果のプロット</h1>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "bd3d1089",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-12-04T12:33:11.647442Z",
     "iopub.status.busy": "2023-12-04T12:33:11.646706Z",
     "iopub.status.idle": "2023-12-04T12:33:12.280026Z",
     "shell.execute_reply": "2023-12-04T12:33:12.279178Z"
    },
    "papermill": {
     "duration": 0.672909,
     "end_time": "2023-12-04T12:33:12.282692",
     "exception": false,
     "start_time": "2023-12-04T12:33:11.609783",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1200x400 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "# Plot the training and validation loss\n",
    "plt.figure(figsize=(12, 4))                                              # 横12インチ、縦4インチのグラフを作成\n",
    "plt.subplot(1, 2, 1)                                                     # 1行 2列にグラフを表示、最初に左からグラフをプロットする。\n",
    "plt.plot(range(1, num_epochs+1), train_losses_20, label='20 Training Loss', color = 'blue', linestyle = 'dashed')\n",
    "plt.plot(range(1, num_epochs+1), train_losses_56,label='56 Training Loss' , color = 'magenta', linestyle = 'dashed')\n",
    "plt.plot(range(1, num_epochs+1), val_losses_20,label='20 Validation Loss' , color = 'blue')\n",
    "plt.plot(range(1, num_epochs+1), val_losses_56 ,label='56 Validation Loss', color = 'magenta')\n",
    "\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Loss over Epochs')\n",
    "plt.legend()\n",
    "\n",
    "# Plot the training and validation accuracy\n",
    "plt.subplot(1, 2, 2)\n",
    "plt.plot(range(1, num_epochs+1), train_accuracies_20, label='20 Training Accuracy', color = 'blue', linestyle = 'dashed')\n",
    "plt.plot(range(1, num_epochs+1), train_accuracies_56,  label='56 Training Accuracy', color = 'magenta', linestyle = 'dashed')\n",
    "plt.plot(range(1, num_epochs+1), val_accuracies_20, label='20 Validation Accuracy', color = 'blue')\n",
    "plt.plot(range(1, num_epochs+1), val_accuracies_56, label='56 Training Accuracy', color = 'magenta')\n",
    "\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.title('Accuracy over Epochs')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kaggle": {
   "accelerator": "nvidiaTeslaT4",
   "dataSources": [],
   "dockerImageVersionId": 30588,
   "isGpuEnabled": true,
   "isInternetEnabled": true,
   "language": "python",
   "sourceType": "notebook"
  },
  "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": "3.10.12"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 7813.173268,
   "end_time": "2023-12-04T12:33:14.995539",
   "environment_variables": {},
   "exception": null,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2023-12-04T10:23:01.822271",
   "version": "2.4.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}