shuffle.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Tranfer Learning over Shufflnetv2\n",
    "\n",
    "Fine Tuning a Shufflenet model (pre-trained over Imagenet) on the CIFAR-10 Dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from time import time\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "from torchvision import datasets, models, transforms\n",
    "import torchvision\n",
    "import matplotlib.pyplot as plt\n",
    "import torch\n",
    "import torch.optim as optim\n",
    "from torch.utils.data import DataLoader\n",
    "from torchsummary import summary\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Training, Valiation Loop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_test(model, trainload, testload, loss, optim, epochs, device):\n",
    "    start_time = time()\n",
    "    print(f'Start Time: {start_time}')\n",
    "\n",
    "    # Store metrics\n",
    "    train_losses = []\n",
    "    valid_losses = []\n",
    "\n",
    "    for epoch in range(epochs):\n",
    "        model.train()\n",
    "        total_step = len(trainload)\n",
    "        running_loss, running_corrects = 0, 0\n",
    "\n",
    "        for i, (inputs, labels) in enumerate(trainload):\n",
    "            inputs = inputs.to(device)\n",
    "            labels = labels.to(device)\n",
    "\n",
    "            # train\n",
    "            optim.zero_grad()\n",
    "            with torch.set_grad_enabled(True):\n",
    "                outputs = model(inputs) # prediction\n",
    "                _loss = loss(outputs, labels) # loss\n",
    "                _, pred = torch.max(outputs, 1)\n",
    "\n",
    "                # backpropr\n",
    "                _loss.backward()\n",
    "                optim.step()\n",
    "\n",
    "            running_loss += _loss.item() * inputs.size(0)\n",
    "            running_corrects += torch.sum(pred == labels.data)\n",
    "\n",
    "            # display every 100 samples\n",
    "            if (i * inputs.size(0)) % 1000 == 0 and i != 0:\n",
    "                print(f'Epoch {epoch+1}/{epochs}, Step {(i * inputs.size(0))}/{len(trainload.dataset)}, '\n",
    "                      f'Loss: {running_loss/(i * inputs.size(0)):.2f}, Acc: {running_corrects.double()/(i * inputs.size(0)):.2f}')\n",
    "\n",
    "\n",
    "\n",
    "        # valdation\n",
    "        model.eval()\n",
    "        with torch.no_grad():\n",
    "            valid_loss, valid_acc = 0.0, 0.0\n",
    "\n",
    "            for i, (inputs, labels) in enumerate(testload):\n",
    "                inputs = inputs.to(device)\n",
    "                labels = labels.to(device)\n",
    "\n",
    "                # test\n",
    "                with torch.set_grad_enabled(False):\n",
    "                    outputs = model(inputs)\n",
    "                    _, pred = torch.max(outputs, 1)\n",
    "                    _loss = loss(outputs, labels)\n",
    "\n",
    "                valid_loss += _loss.item() * inputs.size(0)\n",
    "                valid_acc += torch.sum(pred == labels.data)\n",
    "\n",
    "            print(f'Epoch #{epoch+1}, '\n",
    "                    f'Validation Loss: {valid_loss/len(testload.dataset):.2f}, '\n",
    "                    f'Validation Acc: {valid_acc.double()/len(testload.dataset):.2f}')\n",
    "\n",
    "        # Save losses for plotting\n",
    "        train_losses.append(running_loss/len(trainload.dataset))\n",
    "        valid_losses.append(valid_loss/len(testload.dataset))\n",
    "\n",
    "    print(f'Training time: {start_time - time()}')\n",
    "    print('Savin model')\n",
    "    torch.save(model.state_dict(), 'my_shufflenet.pt')\n",
    "\n",
    "    return train_losses, valid_losses\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using cache found in /home/local/ASUAD/nkondepu/.cache/torch/hub/pytorch_vision_v0.10.0\n",
      "/home/local/ASUAD/nkondepu/anaconda3/envs/cen598/lib/python3.9/site-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.\n",
      "  warnings.warn(\n",
      "/home/local/ASUAD/nkondepu/anaconda3/envs/cen598/lib/python3.9/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=ShuffleNet_V2_X1_0_Weights.IMAGENET1K_V1`. You can also use `weights=ShuffleNet_V2_X1_0_Weights.DEFAULT` to get the most up-to-date weights.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----------------------------------------------------------------\n",
      "        Layer (type)               Output Shape         Param #\n",
      "================================================================\n",
      "            Conv2d-1         [-1, 24, 112, 112]             648\n",
      "       BatchNorm2d-2         [-1, 24, 112, 112]              48\n",
      "              ReLU-3         [-1, 24, 112, 112]               0\n",
      "         MaxPool2d-4           [-1, 24, 56, 56]               0\n",
      "            Conv2d-5           [-1, 24, 28, 28]             216\n",
      "       BatchNorm2d-6           [-1, 24, 28, 28]              48\n",
      "            Conv2d-7           [-1, 58, 28, 28]           1,392\n",
      "       BatchNorm2d-8           [-1, 58, 28, 28]             116\n",
      "              ReLU-9           [-1, 58, 28, 28]               0\n",
      "           Conv2d-10           [-1, 58, 56, 56]           1,392\n",
      "      BatchNorm2d-11           [-1, 58, 56, 56]             116\n",
      "             ReLU-12           [-1, 58, 56, 56]               0\n",
      "           Conv2d-13           [-1, 58, 28, 28]             522\n",
      "      BatchNorm2d-14           [-1, 58, 28, 28]             116\n",
      "           Conv2d-15           [-1, 58, 28, 28]           3,364\n",
      "      BatchNorm2d-16           [-1, 58, 28, 28]             116\n",
      "             ReLU-17           [-1, 58, 28, 28]               0\n",
      " InvertedResidual-18          [-1, 116, 28, 28]               0\n",
      "           Conv2d-19           [-1, 58, 28, 28]           3,364\n",
      "      BatchNorm2d-20           [-1, 58, 28, 28]             116\n",
      "             ReLU-21           [-1, 58, 28, 28]               0\n",
      "           Conv2d-22           [-1, 58, 28, 28]             522\n",
      "      BatchNorm2d-23           [-1, 58, 28, 28]             116\n",
      "           Conv2d-24           [-1, 58, 28, 28]           3,364\n",
      "      BatchNorm2d-25           [-1, 58, 28, 28]             116\n",
      "             ReLU-26           [-1, 58, 28, 28]               0\n",
      " InvertedResidual-27          [-1, 116, 28, 28]               0\n",
      "           Conv2d-28           [-1, 58, 28, 28]           3,364\n",
      "      BatchNorm2d-29           [-1, 58, 28, 28]             116\n",
      "             ReLU-30           [-1, 58, 28, 28]               0\n",
      "           Conv2d-31           [-1, 58, 28, 28]             522\n",
      "      BatchNorm2d-32           [-1, 58, 28, 28]             116\n",
      "           Conv2d-33           [-1, 58, 28, 28]           3,364\n",
      "      BatchNorm2d-34           [-1, 58, 28, 28]             116\n",
      "             ReLU-35           [-1, 58, 28, 28]               0\n",
      " InvertedResidual-36          [-1, 116, 28, 28]               0\n",
      "           Conv2d-37           [-1, 58, 28, 28]           3,364\n",
      "      BatchNorm2d-38           [-1, 58, 28, 28]             116\n",
      "             ReLU-39           [-1, 58, 28, 28]               0\n",
      "           Conv2d-40           [-1, 58, 28, 28]             522\n",
      "      BatchNorm2d-41           [-1, 58, 28, 28]             116\n",
      "           Conv2d-42           [-1, 58, 28, 28]           3,364\n",
      "      BatchNorm2d-43           [-1, 58, 28, 28]             116\n",
      "             ReLU-44           [-1, 58, 28, 28]               0\n",
      " InvertedResidual-45          [-1, 116, 28, 28]               0\n",
      "           Conv2d-46          [-1, 116, 14, 14]           1,044\n",
      "      BatchNorm2d-47          [-1, 116, 14, 14]             232\n",
      "           Conv2d-48          [-1, 116, 14, 14]          13,456\n",
      "      BatchNorm2d-49          [-1, 116, 14, 14]             232\n",
      "             ReLU-50          [-1, 116, 14, 14]               0\n",
      "           Conv2d-51          [-1, 116, 28, 28]          13,456\n",
      "      BatchNorm2d-52          [-1, 116, 28, 28]             232\n",
      "             ReLU-53          [-1, 116, 28, 28]               0\n",
      "           Conv2d-54          [-1, 116, 14, 14]           1,044\n",
      "      BatchNorm2d-55          [-1, 116, 14, 14]             232\n",
      "           Conv2d-56          [-1, 116, 14, 14]          13,456\n",
      "      BatchNorm2d-57          [-1, 116, 14, 14]             232\n",
      "             ReLU-58          [-1, 116, 14, 14]               0\n",
      " InvertedResidual-59          [-1, 232, 14, 14]               0\n",
      "           Conv2d-60          [-1, 116, 14, 14]          13,456\n",
      "      BatchNorm2d-61          [-1, 116, 14, 14]             232\n",
      "             ReLU-62          [-1, 116, 14, 14]               0\n",
      "           Conv2d-63          [-1, 116, 14, 14]           1,044\n",
      "      BatchNorm2d-64          [-1, 116, 14, 14]             232\n",
      "           Conv2d-65          [-1, 116, 14, 14]          13,456\n",
      "      BatchNorm2d-66          [-1, 116, 14, 14]             232\n",
      "             ReLU-67          [-1, 116, 14, 14]               0\n",
      " InvertedResidual-68          [-1, 232, 14, 14]               0\n",
      "           Conv2d-69          [-1, 116, 14, 14]          13,456\n",
      "      BatchNorm2d-70          [-1, 116, 14, 14]             232\n",
      "             ReLU-71          [-1, 116, 14, 14]               0\n",
      "           Conv2d-72          [-1, 116, 14, 14]           1,044\n",
      "      BatchNorm2d-73          [-1, 116, 14, 14]             232\n",
      "           Conv2d-74          [-1, 116, 14, 14]          13,456\n",
      "      BatchNorm2d-75          [-1, 116, 14, 14]             232\n",
      "             ReLU-76          [-1, 116, 14, 14]               0\n",
      " InvertedResidual-77          [-1, 232, 14, 14]               0\n",
      "           Conv2d-78          [-1, 116, 14, 14]          13,456\n",
      "      BatchNorm2d-79          [-1, 116, 14, 14]             232\n",
      "             ReLU-80          [-1, 116, 14, 14]               0\n",
      "           Conv2d-81          [-1, 116, 14, 14]           1,044\n",
      "      BatchNorm2d-82          [-1, 116, 14, 14]             232\n",
      "           Conv2d-83          [-1, 116, 14, 14]          13,456\n",
      "      BatchNorm2d-84          [-1, 116, 14, 14]             232\n",
      "             ReLU-85          [-1, 116, 14, 14]               0\n",
      " InvertedResidual-86          [-1, 232, 14, 14]               0\n",
      "           Conv2d-87          [-1, 116, 14, 14]          13,456\n",
      "      BatchNorm2d-88          [-1, 116, 14, 14]             232\n",
      "             ReLU-89          [-1, 116, 14, 14]               0\n",
      "           Conv2d-90          [-1, 116, 14, 14]           1,044\n",
      "      BatchNorm2d-91          [-1, 116, 14, 14]             232\n",
      "           Conv2d-92          [-1, 116, 14, 14]          13,456\n",
      "      BatchNorm2d-93          [-1, 116, 14, 14]             232\n",
      "             ReLU-94          [-1, 116, 14, 14]               0\n",
      " InvertedResidual-95          [-1, 232, 14, 14]               0\n",
      "           Conv2d-96          [-1, 116, 14, 14]          13,456\n",
      "      BatchNorm2d-97          [-1, 116, 14, 14]             232\n",
      "             ReLU-98          [-1, 116, 14, 14]               0\n",
      "           Conv2d-99          [-1, 116, 14, 14]           1,044\n",
      "     BatchNorm2d-100          [-1, 116, 14, 14]             232\n",
      "          Conv2d-101          [-1, 116, 14, 14]          13,456\n",
      "     BatchNorm2d-102          [-1, 116, 14, 14]             232\n",
      "            ReLU-103          [-1, 116, 14, 14]               0\n",
      "InvertedResidual-104          [-1, 232, 14, 14]               0\n",
      "          Conv2d-105          [-1, 116, 14, 14]          13,456\n",
      "     BatchNorm2d-106          [-1, 116, 14, 14]             232\n",
      "            ReLU-107          [-1, 116, 14, 14]               0\n",
      "          Conv2d-108          [-1, 116, 14, 14]           1,044\n",
      "     BatchNorm2d-109          [-1, 116, 14, 14]             232\n",
      "          Conv2d-110          [-1, 116, 14, 14]          13,456\n",
      "     BatchNorm2d-111          [-1, 116, 14, 14]             232\n",
      "            ReLU-112          [-1, 116, 14, 14]               0\n",
      "InvertedResidual-113          [-1, 232, 14, 14]               0\n",
      "          Conv2d-114          [-1, 116, 14, 14]          13,456\n",
      "     BatchNorm2d-115          [-1, 116, 14, 14]             232\n",
      "            ReLU-116          [-1, 116, 14, 14]               0\n",
      "          Conv2d-117          [-1, 116, 14, 14]           1,044\n",
      "     BatchNorm2d-118          [-1, 116, 14, 14]             232\n",
      "          Conv2d-119          [-1, 116, 14, 14]          13,456\n",
      "     BatchNorm2d-120          [-1, 116, 14, 14]             232\n",
      "            ReLU-121          [-1, 116, 14, 14]               0\n",
      "InvertedResidual-122          [-1, 232, 14, 14]               0\n",
      "          Conv2d-123            [-1, 232, 7, 7]           2,088\n",
      "     BatchNorm2d-124            [-1, 232, 7, 7]             464\n",
      "          Conv2d-125            [-1, 232, 7, 7]          53,824\n",
      "     BatchNorm2d-126            [-1, 232, 7, 7]             464\n",
      "            ReLU-127            [-1, 232, 7, 7]               0\n",
      "          Conv2d-128          [-1, 232, 14, 14]          53,824\n",
      "     BatchNorm2d-129          [-1, 232, 14, 14]             464\n",
      "            ReLU-130          [-1, 232, 14, 14]               0\n",
      "          Conv2d-131            [-1, 232, 7, 7]           2,088\n",
      "     BatchNorm2d-132            [-1, 232, 7, 7]             464\n",
      "          Conv2d-133            [-1, 232, 7, 7]          53,824\n",
      "     BatchNorm2d-134            [-1, 232, 7, 7]             464\n",
      "            ReLU-135            [-1, 232, 7, 7]               0\n",
      "InvertedResidual-136            [-1, 464, 7, 7]               0\n",
      "          Conv2d-137            [-1, 232, 7, 7]          53,824\n",
      "     BatchNorm2d-138            [-1, 232, 7, 7]             464\n",
      "            ReLU-139            [-1, 232, 7, 7]               0\n",
      "          Conv2d-140            [-1, 232, 7, 7]           2,088\n",
      "     BatchNorm2d-141            [-1, 232, 7, 7]             464\n",
      "          Conv2d-142            [-1, 232, 7, 7]          53,824\n",
      "     BatchNorm2d-143            [-1, 232, 7, 7]             464\n",
      "            ReLU-144            [-1, 232, 7, 7]               0\n",
      "InvertedResidual-145            [-1, 464, 7, 7]               0\n",
      "          Conv2d-146            [-1, 232, 7, 7]          53,824\n",
      "     BatchNorm2d-147            [-1, 232, 7, 7]             464\n",
      "            ReLU-148            [-1, 232, 7, 7]               0\n",
      "          Conv2d-149            [-1, 232, 7, 7]           2,088\n",
      "     BatchNorm2d-150            [-1, 232, 7, 7]             464\n",
      "          Conv2d-151            [-1, 232, 7, 7]          53,824\n",
      "     BatchNorm2d-152            [-1, 232, 7, 7]             464\n",
      "            ReLU-153            [-1, 232, 7, 7]               0\n",
      "InvertedResidual-154            [-1, 464, 7, 7]               0\n",
      "          Conv2d-155            [-1, 232, 7, 7]          53,824\n",
      "     BatchNorm2d-156            [-1, 232, 7, 7]             464\n",
      "            ReLU-157            [-1, 232, 7, 7]               0\n",
      "          Conv2d-158            [-1, 232, 7, 7]           2,088\n",
      "     BatchNorm2d-159            [-1, 232, 7, 7]             464\n",
      "          Conv2d-160            [-1, 232, 7, 7]          53,824\n",
      "     BatchNorm2d-161            [-1, 232, 7, 7]             464\n",
      "            ReLU-162            [-1, 232, 7, 7]               0\n",
      "InvertedResidual-163            [-1, 464, 7, 7]               0\n",
      "          Conv2d-164           [-1, 1024, 7, 7]         475,136\n",
      "     BatchNorm2d-165           [-1, 1024, 7, 7]           2,048\n",
      "            ReLU-166           [-1, 1024, 7, 7]               0\n",
      "          Linear-167                 [-1, 1000]       1,025,000\n",
      "================================================================\n",
      "Total params: 2,278,604\n",
      "Trainable params: 2,278,604\n",
      "Non-trainable params: 0\n",
      "----------------------------------------------------------------\n",
      "Input size (MB): 0.57\n",
      "Forward/backward pass size (MB): 47.94\n",
      "Params size (MB): 8.69\n",
      "Estimated Total Size (MB): 57.20\n",
      "----------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "model = torch.hub.load('pytorch/vision:v0.10.0', 'shufflenet_v2_x1_0', pretrained=True)\n",
    "model.eval()\n",
    "model.cuda()\n",
    "summary(model, (3, 224, 224))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Freeze network params\n",
    "- Freeze all params in the network\n",
    "- Replace the final classification layer with a new one to train."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# freeze all model params\n",
    "for params in model.parameters():\n",
    "    params.requires_grad = False\n",
    "\n",
    "numftrs = model.fc.in_features\n",
    "model.fc = nn.Linear(numftrs, 10)\n",
    "\n",
    "model = model.to(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Default transformations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_transforms = transforms.Compose([\n",
    "    transforms.Resize(256),\n",
    "    transforms.CenterCrop(224),\n",
    "    transforms.ToTensor(),\n",
    "    transforms.Normalize(\n",
    "        (0.4914, 0.4822, 0.4465),\n",
    "        (0.247, 0.243, 0.261)\n",
    "    )\n",
    "])\n",
    "\n",
    "test_transforms = transforms.Compose([\n",
    "    transforms.Resize(256),\n",
    "    transforms.CenterCrop(224),\n",
    "    transforms.ToTensor(),\n",
    "    transforms.Normalize(\n",
    "        (0.4914, 0.4822, 0.4465),\n",
    "        (0.247, 0.243, 0.261)\n",
    "    )\n",
    "])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Dataloaders"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to ./data/cifar-10-python.tar.gz\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100.0%\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./data/cifar-10-python.tar.gz to ./data\n",
      "Files already downloaded and verified\n"
     ]
    }
   ],
   "source": [
    "batch_size = 32\n",
    "\n",
    "trainset = torchvision.datasets.CIFAR10(root='./data', train=True,\n",
    "                                        download=True, transform=train_transforms)\n",
    "trainloader = DataLoader(trainset, batch_size=batch_size,shuffle=True)\n",
    "\n",
    "testset = torchvision.datasets.CIFAR10(root='./data', train=False,\n",
    "                                    download=True, transform=test_transforms)\n",
    "testloader = DataLoader(testset, batch_size=batch_size,shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start Time: 1698541743.9947479\n",
      "Epoch 1/35, Step 4000/50000, Loss: 2.24, Acc: 0.28\n",
      "Epoch 1/35, Step 8000/50000, Loss: 2.15, Acc: 0.43\n",
      "Epoch 1/35, Step 12000/50000, Loss: 2.07, Acc: 0.51\n",
      "Epoch 1/35, Step 16000/50000, Loss: 2.01, Acc: 0.55\n",
      "Epoch 1/35, Step 20000/50000, Loss: 1.95, Acc: 0.57\n",
      "Epoch 1/35, Step 24000/50000, Loss: 1.89, Acc: 0.59\n",
      "Epoch 1/35, Step 28000/50000, Loss: 1.84, Acc: 0.61\n",
      "Epoch 1/35, Step 32000/50000, Loss: 1.79, Acc: 0.62\n",
      "Epoch 1/35, Step 36000/50000, Loss: 1.75, Acc: 0.63\n",
      "Epoch 1/35, Step 40000/50000, Loss: 1.71, Acc: 0.63\n",
      "Epoch 1/35, Step 44000/50000, Loss: 1.67, Acc: 0.64\n",
      "Epoch 1/35, Step 48000/50000, Loss: 1.64, Acc: 0.65\n",
      "Epoch #1, Validation Loss: 1.18, Validation Acc: 0.73\n",
      "Epoch 2/35, Step 4000/50000, Loss: 1.22, Acc: 0.71\n",
      "Epoch 2/35, Step 8000/50000, Loss: 1.20, Acc: 0.72\n",
      "Epoch 2/35, Step 12000/50000, Loss: 1.18, Acc: 0.72\n",
      "Epoch 2/35, Step 16000/50000, Loss: 1.16, Acc: 0.72\n",
      "Epoch 2/35, Step 20000/50000, Loss: 1.15, Acc: 0.72\n",
      "Epoch 2/35, Step 24000/50000, Loss: 1.14, Acc: 0.72\n",
      "Epoch 2/35, Step 28000/50000, Loss: 1.13, Acc: 0.72\n",
      "Epoch 2/35, Step 32000/50000, Loss: 1.11, Acc: 0.72\n",
      "Epoch 2/35, Step 36000/50000, Loss: 1.10, Acc: 0.72\n",
      "Epoch 2/35, Step 40000/50000, Loss: 1.09, Acc: 0.72\n",
      "Epoch 2/35, Step 44000/50000, Loss: 1.08, Acc: 0.72\n",
      "Epoch 2/35, Step 48000/50000, Loss: 1.07, Acc: 0.73\n",
      "Epoch #2, Validation Loss: 0.91, Validation Acc: 0.76\n",
      "Epoch 3/35, Step 4000/50000, Loss: 0.95, Acc: 0.74\n",
      "Epoch 3/35, Step 8000/50000, Loss: 0.95, Acc: 0.73\n",
      "Epoch 3/35, Step 12000/50000, Loss: 0.94, Acc: 0.73\n",
      "Epoch 3/35, Step 16000/50000, Loss: 0.93, Acc: 0.74\n",
      "Epoch 3/35, Step 20000/50000, Loss: 0.92, Acc: 0.74\n",
      "Epoch 3/35, Step 24000/50000, Loss: 0.92, Acc: 0.74\n",
      "Epoch 3/35, Step 28000/50000, Loss: 0.91, Acc: 0.74\n",
      "Epoch 3/35, Step 32000/50000, Loss: 0.91, Acc: 0.74\n",
      "Epoch 3/35, Step 36000/50000, Loss: 0.91, Acc: 0.74\n",
      "Epoch 3/35, Step 40000/50000, Loss: 0.90, Acc: 0.74\n",
      "Epoch 3/35, Step 44000/50000, Loss: 0.90, Acc: 0.74\n",
      "Epoch 3/35, Step 48000/50000, Loss: 0.89, Acc: 0.74\n",
      "Epoch #3, Validation Loss: 0.80, Validation Acc: 0.77\n",
      "Epoch 4/35, Step 4000/50000, Loss: 0.85, Acc: 0.75\n",
      "Epoch 4/35, Step 8000/50000, Loss: 0.84, Acc: 0.76\n",
      "Epoch 4/35, Step 12000/50000, Loss: 0.83, Acc: 0.76\n",
      "Epoch 4/35, Step 16000/50000, Loss: 0.83, Acc: 0.76\n",
      "Epoch 4/35, Step 20000/50000, Loss: 0.82, Acc: 0.76\n",
      "Epoch 4/35, Step 24000/50000, Loss: 0.82, Acc: 0.76\n",
      "Epoch 4/35, Step 28000/50000, Loss: 0.82, Acc: 0.76\n",
      "Epoch 4/35, Step 32000/50000, Loss: 0.82, Acc: 0.76\n",
      "Epoch 4/35, Step 36000/50000, Loss: 0.82, Acc: 0.76\n",
      "Epoch 4/35, Step 40000/50000, Loss: 0.81, Acc: 0.76\n",
      "Epoch 4/35, Step 44000/50000, Loss: 0.81, Acc: 0.76\n",
      "Epoch 4/35, Step 48000/50000, Loss: 0.81, Acc: 0.76\n",
      "Epoch #4, Validation Loss: 0.73, Validation Acc: 0.78\n",
      "Epoch 5/35, Step 4000/50000, Loss: 0.77, Acc: 0.78\n",
      "Epoch 5/35, Step 8000/50000, Loss: 0.77, Acc: 0.77\n",
      "Epoch 5/35, Step 12000/50000, Loss: 0.77, Acc: 0.77\n",
      "Epoch 5/35, Step 16000/50000, Loss: 0.77, Acc: 0.77\n",
      "Epoch 5/35, Step 20000/50000, Loss: 0.77, Acc: 0.77\n",
      "Epoch 5/35, Step 24000/50000, Loss: 0.76, Acc: 0.77\n",
      "Epoch 5/35, Step 28000/50000, Loss: 0.76, Acc: 0.77\n",
      "Epoch 5/35, Step 32000/50000, Loss: 0.76, Acc: 0.77\n",
      "Epoch 5/35, Step 36000/50000, Loss: 0.76, Acc: 0.77\n",
      "Epoch 5/35, Step 40000/50000, Loss: 0.75, Acc: 0.77\n",
      "Epoch 5/35, Step 44000/50000, Loss: 0.75, Acc: 0.77\n",
      "Epoch 5/35, Step 48000/50000, Loss: 0.75, Acc: 0.77\n",
      "Epoch #5, Validation Loss: 0.69, Validation Acc: 0.78\n",
      "Epoch 6/35, Step 4000/50000, Loss: 0.74, Acc: 0.77\n",
      "Epoch 6/35, Step 8000/50000, Loss: 0.73, Acc: 0.77\n",
      "Epoch 6/35, Step 12000/50000, Loss: 0.73, Acc: 0.77\n",
      "Epoch 6/35, Step 16000/50000, Loss: 0.73, Acc: 0.77\n",
      "Epoch 6/35, Step 20000/50000, Loss: 0.73, Acc: 0.77\n",
      "Epoch 6/35, Step 24000/50000, Loss: 0.73, Acc: 0.77\n",
      "Epoch 6/35, Step 28000/50000, Loss: 0.73, Acc: 0.77\n",
      "Epoch 6/35, Step 32000/50000, Loss: 0.72, Acc: 0.77\n",
      "Epoch 6/35, Step 36000/50000, Loss: 0.72, Acc: 0.77\n",
      "Epoch 6/35, Step 40000/50000, Loss: 0.72, Acc: 0.77\n",
      "Epoch 6/35, Step 44000/50000, Loss: 0.72, Acc: 0.77\n",
      "Epoch 6/35, Step 48000/50000, Loss: 0.72, Acc: 0.77\n",
      "Epoch #6, Validation Loss: 0.66, Validation Acc: 0.79\n",
      "Epoch 7/35, Step 4000/50000, Loss: 0.71, Acc: 0.78\n",
      "Epoch 7/35, Step 8000/50000, Loss: 0.70, Acc: 0.78\n",
      "Epoch 7/35, Step 12000/50000, Loss: 0.71, Acc: 0.78\n",
      "Epoch 7/35, Step 16000/50000, Loss: 0.70, Acc: 0.78\n",
      "Epoch 7/35, Step 20000/50000, Loss: 0.70, Acc: 0.78\n",
      "Epoch 7/35, Step 24000/50000, Loss: 0.70, Acc: 0.78\n",
      "Epoch 7/35, Step 28000/50000, Loss: 0.70, Acc: 0.78\n",
      "Epoch 7/35, Step 32000/50000, Loss: 0.70, Acc: 0.78\n",
      "Epoch 7/35, Step 36000/50000, Loss: 0.70, Acc: 0.78\n",
      "Epoch 7/35, Step 40000/50000, Loss: 0.69, Acc: 0.78\n",
      "Epoch 7/35, Step 44000/50000, Loss: 0.69, Acc: 0.78\n",
      "Epoch 7/35, Step 48000/50000, Loss: 0.69, Acc: 0.78\n",
      "Epoch #7, Validation Loss: 0.64, Validation Acc: 0.80\n",
      "Epoch 8/35, Step 4000/50000, Loss: 0.67, Acc: 0.80\n",
      "Epoch 8/35, Step 8000/50000, Loss: 0.67, Acc: 0.79\n",
      "Epoch 8/35, Step 12000/50000, Loss: 0.67, Acc: 0.79\n",
      "Epoch 8/35, Step 16000/50000, Loss: 0.67, Acc: 0.79\n",
      "Epoch 8/35, Step 20000/50000, Loss: 0.68, Acc: 0.78\n",
      "Epoch 8/35, Step 24000/50000, Loss: 0.68, Acc: 0.78\n",
      "Epoch 8/35, Step 28000/50000, Loss: 0.67, Acc: 0.78\n",
      "Epoch 8/35, Step 32000/50000, Loss: 0.67, Acc: 0.78\n",
      "Epoch 8/35, Step 36000/50000, Loss: 0.67, Acc: 0.78\n",
      "Epoch 8/35, Step 40000/50000, Loss: 0.67, Acc: 0.78\n",
      "Epoch 8/35, Step 44000/50000, Loss: 0.67, Acc: 0.78\n",
      "Epoch 8/35, Step 48000/50000, Loss: 0.67, Acc: 0.78\n",
      "Epoch #8, Validation Loss: 0.62, Validation Acc: 0.80\n",
      "Epoch 9/35, Step 4000/50000, Loss: 0.65, Acc: 0.79\n",
      "Epoch 9/35, Step 8000/50000, Loss: 0.65, Acc: 0.79\n",
      "Epoch 9/35, Step 12000/50000, Loss: 0.66, Acc: 0.79\n",
      "Epoch 9/35, Step 16000/50000, Loss: 0.65, Acc: 0.79\n",
      "Epoch 9/35, Step 20000/50000, Loss: 0.65, Acc: 0.79\n",
      "Epoch 9/35, Step 24000/50000, Loss: 0.65, Acc: 0.79\n",
      "Epoch 9/35, Step 28000/50000, Loss: 0.65, Acc: 0.79\n",
      "Epoch 9/35, Step 32000/50000, Loss: 0.65, Acc: 0.79\n",
      "Epoch 9/35, Step 36000/50000, Loss: 0.65, Acc: 0.79\n",
      "Epoch 9/35, Step 40000/50000, Loss: 0.65, Acc: 0.79\n",
      "Epoch 9/35, Step 44000/50000, Loss: 0.65, Acc: 0.79\n",
      "Epoch 9/35, Step 48000/50000, Loss: 0.65, Acc: 0.79\n",
      "Epoch #9, Validation Loss: 0.61, Validation Acc: 0.80\n",
      "Epoch 10/35, Step 4000/50000, Loss: 0.65, Acc: 0.80\n",
      "Epoch 10/35, Step 8000/50000, Loss: 0.64, Acc: 0.79\n",
      "Epoch 10/35, Step 12000/50000, Loss: 0.64, Acc: 0.79\n",
      "Epoch 10/35, Step 16000/50000, Loss: 0.64, Acc: 0.79\n",
      "Epoch 10/35, Step 20000/50000, Loss: 0.64, Acc: 0.79\n",
      "Epoch 10/35, Step 24000/50000, Loss: 0.64, Acc: 0.79\n",
      "Epoch 10/35, Step 28000/50000, Loss: 0.64, Acc: 0.79\n",
      "Epoch 10/35, Step 32000/50000, Loss: 0.64, Acc: 0.79\n",
      "Epoch 10/35, Step 36000/50000, Loss: 0.64, Acc: 0.79\n",
      "Epoch 10/35, Step 40000/50000, Loss: 0.64, Acc: 0.79\n",
      "Epoch 10/35, Step 44000/50000, Loss: 0.64, Acc: 0.79\n",
      "Epoch 10/35, Step 48000/50000, Loss: 0.64, Acc: 0.79\n",
      "Epoch #10, Validation Loss: 0.60, Validation Acc: 0.80\n",
      "Epoch 11/35, Step 4000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 11/35, Step 8000/50000, Loss: 0.63, Acc: 0.79\n",
      "Epoch 11/35, Step 12000/50000, Loss: 0.63, Acc: 0.79\n",
      "Epoch 11/35, Step 16000/50000, Loss: 0.63, Acc: 0.79\n",
      "Epoch 11/35, Step 20000/50000, Loss: 0.63, Acc: 0.79\n",
      "Epoch 11/35, Step 24000/50000, Loss: 0.63, Acc: 0.79\n",
      "Epoch 11/35, Step 28000/50000, Loss: 0.63, Acc: 0.79\n",
      "Epoch 11/35, Step 32000/50000, Loss: 0.63, Acc: 0.79\n",
      "Epoch 11/35, Step 36000/50000, Loss: 0.63, Acc: 0.79\n",
      "Epoch 11/35, Step 40000/50000, Loss: 0.63, Acc: 0.79\n",
      "Epoch 11/35, Step 44000/50000, Loss: 0.63, Acc: 0.79\n",
      "Epoch 11/35, Step 48000/50000, Loss: 0.63, Acc: 0.79\n",
      "Epoch #11, Validation Loss: 0.58, Validation Acc: 0.81\n",
      "Epoch 12/35, Step 4000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch 12/35, Step 8000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 12/35, Step 12000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 12/35, Step 16000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 12/35, Step 20000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 12/35, Step 24000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 12/35, Step 28000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 12/35, Step 32000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 12/35, Step 36000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 12/35, Step 40000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch 12/35, Step 44000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch 12/35, Step 48000/50000, Loss: 0.62, Acc: 0.79\n",
      "Epoch #12, Validation Loss: 0.58, Validation Acc: 0.80\n",
      "Epoch 13/35, Step 4000/50000, Loss: 0.63, Acc: 0.80\n",
      "Epoch 13/35, Step 8000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 13/35, Step 12000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 13/35, Step 16000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 13/35, Step 20000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 13/35, Step 24000/50000, Loss: 0.62, Acc: 0.80\n",
      "Epoch 13/35, Step 28000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch 13/35, Step 32000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch 13/35, Step 36000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch 13/35, Step 40000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch 13/35, Step 44000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch 13/35, Step 48000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch #13, Validation Loss: 0.57, Validation Acc: 0.81\n",
      "Epoch 14/35, Step 4000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch 14/35, Step 8000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch 14/35, Step 12000/50000, Loss: 0.61, Acc: 0.80\n",
      "Epoch 14/35, Step 16000/50000, Loss: 0.60, Acc: 0.80\n",
      "Epoch 14/35, Step 20000/50000, Loss: 0.60, Acc: 0.80\n",
      "Epoch 14/35, Step 24000/50000, Loss: 0.60, Acc: 0.80\n",
      "Epoch 14/35, Step 28000/50000, Loss: 0.60, Acc: 0.80\n",
      "Epoch 14/35, Step 32000/50000, Loss: 0.60, Acc: 0.80\n",
      "Epoch 14/35, Step 36000/50000, Loss: 0.60, Acc: 0.80\n",
      "Epoch 14/35, Step 40000/50000, Loss: 0.60, Acc: 0.80\n",
      "Epoch 14/35, Step 44000/50000, Loss: 0.60, Acc: 0.80\n",
      "Epoch 14/35, Step 48000/50000, Loss: 0.60, Acc: 0.80\n",
      "Epoch #14, Validation Loss: 0.56, Validation Acc: 0.81\n",
      "Epoch 15/35, Step 4000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 15/35, Step 8000/50000, Loss: 0.60, Acc: 0.81\n",
      "Epoch 15/35, Step 12000/50000, Loss: 0.59, Acc: 0.80\n",
      "Epoch 15/35, Step 16000/50000, Loss: 0.59, Acc: 0.80\n",
      "Epoch 15/35, Step 20000/50000, Loss: 0.59, Acc: 0.80\n",
      "Epoch 15/35, Step 24000/50000, Loss: 0.59, Acc: 0.80\n",
      "Epoch 15/35, Step 28000/50000, Loss: 0.59, Acc: 0.80\n",
      "Epoch 15/35, Step 32000/50000, Loss: 0.59, Acc: 0.80\n",
      "Epoch 15/35, Step 36000/50000, Loss: 0.59, Acc: 0.80\n",
      "Epoch 15/35, Step 40000/50000, Loss: 0.59, Acc: 0.80\n",
      "Epoch 15/35, Step 44000/50000, Loss: 0.59, Acc: 0.80\n",
      "Epoch 15/35, Step 48000/50000, Loss: 0.59, Acc: 0.80\n",
      "Epoch #15, Validation Loss: 0.56, Validation Acc: 0.81\n",
      "Epoch 16/35, Step 4000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 16/35, Step 8000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 16/35, Step 12000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 16/35, Step 16000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 16/35, Step 20000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 16/35, Step 24000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 16/35, Step 28000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 16/35, Step 32000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 16/35, Step 36000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 16/35, Step 40000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 16/35, Step 44000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 16/35, Step 48000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch #16, Validation Loss: 0.55, Validation Acc: 0.82\n",
      "Epoch 17/35, Step 4000/50000, Loss: 0.60, Acc: 0.81\n",
      "Epoch 17/35, Step 8000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 17/35, Step 12000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 17/35, Step 16000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 17/35, Step 20000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 17/35, Step 24000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 17/35, Step 28000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 17/35, Step 32000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 17/35, Step 36000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 17/35, Step 40000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 17/35, Step 44000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 17/35, Step 48000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch #17, Validation Loss: 0.55, Validation Acc: 0.81\n",
      "Epoch 18/35, Step 4000/50000, Loss: 0.58, Acc: 0.82\n",
      "Epoch 18/35, Step 8000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 18/35, Step 12000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 18/35, Step 16000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 18/35, Step 20000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 18/35, Step 24000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 18/35, Step 28000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 18/35, Step 32000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 18/35, Step 36000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 18/35, Step 40000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 18/35, Step 44000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 18/35, Step 48000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch #18, Validation Loss: 0.54, Validation Acc: 0.82\n",
      "Epoch 19/35, Step 4000/50000, Loss: 0.59, Acc: 0.81\n",
      "Epoch 19/35, Step 8000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 19/35, Step 12000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 19/35, Step 16000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 19/35, Step 20000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 19/35, Step 24000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 19/35, Step 28000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 19/35, Step 32000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 19/35, Step 36000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 19/35, Step 40000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 19/35, Step 44000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 19/35, Step 48000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch #19, Validation Loss: 0.54, Validation Acc: 0.82\n",
      "Epoch 20/35, Step 4000/50000, Loss: 0.58, Acc: 0.82\n",
      "Epoch 20/35, Step 8000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 20/35, Step 12000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 20/35, Step 16000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 20/35, Step 20000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 20/35, Step 24000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 20/35, Step 28000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 20/35, Step 32000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 20/35, Step 36000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 20/35, Step 40000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 20/35, Step 44000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 20/35, Step 48000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch #20, Validation Loss: 0.54, Validation Acc: 0.82\n",
      "Epoch 21/35, Step 4000/50000, Loss: 0.59, Acc: 0.82\n",
      "Epoch 21/35, Step 8000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 21/35, Step 12000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 21/35, Step 16000/50000, Loss: 0.58, Acc: 0.81\n",
      "Epoch 21/35, Step 20000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 21/35, Step 24000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 21/35, Step 28000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 21/35, Step 32000/50000, Loss: 0.57, Acc: 0.81\n",
      "Epoch 21/35, Step 36000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 21/35, Step 40000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 21/35, Step 44000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 21/35, Step 48000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch #21, Validation Loss: 0.53, Validation Acc: 0.82\n",
      "Epoch 22/35, Step 4000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 22/35, Step 8000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 22/35, Step 12000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 22/35, Step 16000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 22/35, Step 20000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 22/35, Step 24000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 22/35, Step 28000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 22/35, Step 32000/50000, Loss: 0.56, Acc: 0.82\n",
      "Epoch 22/35, Step 36000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 22/35, Step 40000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 22/35, Step 44000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 22/35, Step 48000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch #22, Validation Loss: 0.53, Validation Acc: 0.82\n",
      "Epoch 23/35, Step 4000/50000, Loss: 0.54, Acc: 0.83\n",
      "Epoch 23/35, Step 8000/50000, Loss: 0.56, Acc: 0.82\n",
      "Epoch 23/35, Step 12000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 23/35, Step 16000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 23/35, Step 20000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 23/35, Step 24000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 23/35, Step 28000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 23/35, Step 32000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 23/35, Step 36000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 23/35, Step 40000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 23/35, Step 44000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 23/35, Step 48000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch #23, Validation Loss: 0.53, Validation Acc: 0.82\n",
      "Epoch 24/35, Step 4000/50000, Loss: 0.57, Acc: 0.82\n",
      "Epoch 24/35, Step 8000/50000, Loss: 0.56, Acc: 0.82\n",
      "Epoch 24/35, Step 12000/50000, Loss: 0.56, Acc: 0.82\n",
      "Epoch 24/35, Step 16000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 24/35, Step 20000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 24/35, Step 24000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 24/35, Step 28000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 24/35, Step 32000/50000, Loss: 0.56, Acc: 0.81\n",
      "Epoch 24/35, Step 36000/50000, Loss: 0.55, Acc: 0.81\n",
      "Epoch 24/35, Step 40000/50000, Loss: 0.55, Acc: 0.81\n",
      "Epoch 24/35, Step 44000/50000, Loss: 0.55, Acc: 0.81\n",
      "Epoch 24/35, Step 48000/50000, Loss: 0.55, Acc: 0.81\n",
      "Epoch #24, Validation Loss: 0.52, Validation Acc: 0.82\n",
      "Epoch 25/35, Step 4000/50000, Loss: 0.54, Acc: 0.83\n",
      "Epoch 25/35, Step 8000/50000, Loss: 0.54, Acc: 0.83\n",
      "Epoch 25/35, Step 12000/50000, Loss: 0.54, Acc: 0.83\n",
      "Epoch 25/35, Step 16000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 25/35, Step 20000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 25/35, Step 24000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 25/35, Step 28000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 25/35, Step 32000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 25/35, Step 36000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 25/35, Step 40000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 25/35, Step 44000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 25/35, Step 48000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch #25, Validation Loss: 0.52, Validation Acc: 0.83\n",
      "Epoch 26/35, Step 4000/50000, Loss: 0.56, Acc: 0.82\n",
      "Epoch 26/35, Step 8000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 26/35, Step 12000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 26/35, Step 16000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 26/35, Step 20000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 26/35, Step 24000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 26/35, Step 28000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 26/35, Step 32000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 26/35, Step 36000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 26/35, Step 40000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 26/35, Step 44000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 26/35, Step 48000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch #26, Validation Loss: 0.52, Validation Acc: 0.83\n",
      "Epoch 27/35, Step 4000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 27/35, Step 8000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 27/35, Step 12000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 27/35, Step 16000/50000, Loss: 0.55, Acc: 0.81\n",
      "Epoch 27/35, Step 20000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 27/35, Step 24000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 27/35, Step 28000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 27/35, Step 32000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 27/35, Step 36000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 27/35, Step 40000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 27/35, Step 44000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 27/35, Step 48000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch #27, Validation Loss: 0.52, Validation Acc: 0.82\n",
      "Epoch 28/35, Step 4000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 28/35, Step 8000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 28/35, Step 12000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 28/35, Step 16000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 28/35, Step 20000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 28/35, Step 24000/50000, Loss: 0.55, Acc: 0.82\n",
      "Epoch 28/35, Step 28000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 28/35, Step 32000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 28/35, Step 36000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 28/35, Step 40000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 28/35, Step 44000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 28/35, Step 48000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch #28, Validation Loss: 0.51, Validation Acc: 0.83\n",
      "Epoch 29/35, Step 4000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 29/35, Step 8000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 29/35, Step 12000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 29/35, Step 16000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 29/35, Step 20000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 29/35, Step 24000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 29/35, Step 28000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 29/35, Step 32000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 29/35, Step 36000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 29/35, Step 40000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 29/35, Step 44000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 29/35, Step 48000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch #29, Validation Loss: 0.52, Validation Acc: 0.82\n",
      "Epoch 30/35, Step 4000/50000, Loss: 0.52, Acc: 0.84\n",
      "Epoch 30/35, Step 8000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 30/35, Step 12000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 30/35, Step 16000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 30/35, Step 20000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 30/35, Step 24000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 30/35, Step 28000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 30/35, Step 32000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 30/35, Step 36000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 30/35, Step 40000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 30/35, Step 44000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 30/35, Step 48000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch #30, Validation Loss: 0.51, Validation Acc: 0.83\n",
      "Epoch 31/35, Step 4000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 31/35, Step 8000/50000, Loss: 0.54, Acc: 0.83\n",
      "Epoch 31/35, Step 12000/50000, Loss: 0.54, Acc: 0.82\n",
      "Epoch 31/35, Step 16000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 31/35, Step 20000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 31/35, Step 24000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 31/35, Step 28000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 31/35, Step 32000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 31/35, Step 36000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 31/35, Step 40000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 31/35, Step 44000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 31/35, Step 48000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch #31, Validation Loss: 0.51, Validation Acc: 0.83\n",
      "Epoch 32/35, Step 4000/50000, Loss: 0.53, Acc: 0.83\n",
      "Epoch 32/35, Step 8000/50000, Loss: 0.53, Acc: 0.83\n",
      "Epoch 32/35, Step 12000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 32/35, Step 16000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 32/35, Step 20000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 32/35, Step 24000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 32/35, Step 28000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 32/35, Step 32000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 32/35, Step 36000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 32/35, Step 40000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 32/35, Step 44000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 32/35, Step 48000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch #32, Validation Loss: 0.50, Validation Acc: 0.83\n",
      "Epoch 33/35, Step 4000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 33/35, Step 8000/50000, Loss: 0.53, Acc: 0.83\n",
      "Epoch 33/35, Step 12000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 33/35, Step 16000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 33/35, Step 20000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 33/35, Step 24000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 33/35, Step 28000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 33/35, Step 32000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 33/35, Step 36000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 33/35, Step 40000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 33/35, Step 44000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 33/35, Step 48000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch #33, Validation Loss: 0.50, Validation Acc: 0.83\n",
      "Epoch 34/35, Step 4000/50000, Loss: 0.51, Acc: 0.83\n",
      "Epoch 34/35, Step 8000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 34/35, Step 12000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 34/35, Step 16000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 34/35, Step 20000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 34/35, Step 24000/50000, Loss: 0.52, Acc: 0.82\n",
      "Epoch 34/35, Step 28000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 34/35, Step 32000/50000, Loss: 0.52, Acc: 0.82\n",
      "Epoch 34/35, Step 36000/50000, Loss: 0.52, Acc: 0.82\n",
      "Epoch 34/35, Step 40000/50000, Loss: 0.52, Acc: 0.82\n",
      "Epoch 34/35, Step 44000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 34/35, Step 48000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch #34, Validation Loss: 0.50, Validation Acc: 0.83\n",
      "Epoch 35/35, Step 4000/50000, Loss: 0.54, Acc: 0.83\n",
      "Epoch 35/35, Step 8000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 35/35, Step 12000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 35/35, Step 16000/50000, Loss: 0.53, Acc: 0.83\n",
      "Epoch 35/35, Step 20000/50000, Loss: 0.53, Acc: 0.82\n",
      "Epoch 35/35, Step 24000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 35/35, Step 28000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 35/35, Step 32000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 35/35, Step 36000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 35/35, Step 40000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 35/35, Step 44000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch 35/35, Step 48000/50000, Loss: 0.52, Acc: 0.83\n",
      "Epoch #35, Validation Loss: 0.50, Validation Acc: 0.83\n",
      "Training time: -2785.6960401535034\n",
      "Savin model\n",
      "==END==\n"
     ]
    }
   ],
   "source": [
    "tloss, vloss = train_test(\n",
    "        model=model,\n",
    "        trainload=trainloader,\n",
    "        testload=testloader,\n",
    "        loss=nn.CrossEntropyLoss(),\n",
    "        optim=optim.Adam(model.parameters(), lr=1e-3),\n",
    "        epochs=35,\n",
    "        device=device\n",
    "    )\n",
    "print('==END==')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting training and validation loss\n",
    "plt.plot(tloss, label='Training loss')\n",
    "plt.plot(vloss, label='Validation loss')\n",
    "plt.title('Loss vs. Epochs')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Confusion matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay\n",
    "import matplotlib.pyplot as plt\n",
    "import torch\n",
    "\n",
    "# Assuming model is your trained PyTorch model, and test_loader is your DataLoader for the test set\n",
    "model.cpu().eval()  # Set the model to evaluation mode\n",
    "\n",
    "# Collect all the predictions and true labels\n",
    "all_preds = []\n",
    "all_labels = []\n",
    "\n",
    "with torch.no_grad():  # No gradients needed\n",
    "    for data, labels in testloader:\n",
    "        outputs = model(data)\n",
    "        _, preds = torch.max(outputs, 1)\n",
    "        all_preds.extend(preds.cpu().numpy())\n",
    "        all_labels.extend(labels.cpu().numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Calculate the confusion matrix\n",
    "cm = confusion_matrix(all_labels, all_preds)\n",
    "\n",
    "# Plot the confusion matrix\n",
    "disp = ConfusionMatrixDisplay(confusion_matrix=cm)\n",
    "disp.plot()\n",
    "# disp.plot(cmap=plt.cm.Blues)\n",
    "\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "cen598",
   "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": "3.9.18"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}