final_project.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "gpuClass": "standard",
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "source": [
        "# установка библиотеки, необходимой для декодирования CTC-loss в разделе с LSTM\n",
        "\n",
        "import os\n",
        "os.chdir('/content/')\n",
        "!git clone --recursive https://github.com/parlance/ctcdecode.git\n",
        "os.chdir('/content/ctcdecode/')\n",
        "!pip install ."
      ],
      "metadata": {
        "id": "UK5q53zUPuPW",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8d5121ce-532f-4f55-9872-71078acda9be"
      },
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Cloning into 'ctcdecode'...\n",
            "remote: Enumerating objects: 1102, done.\u001b[K\n",
            "remote: Counting objects: 100% (39/39), done.\u001b[K\n",
            "remote: Compressing objects: 100% (25/25), done.\u001b[K\n",
            "remote: Total 1102 (delta 16), reused 32 (delta 14), pack-reused 1063\u001b[K\n",
            "Receiving objects: 100% (1102/1102), 782.27 KiB | 15.34 MiB/s, done.\n",
            "Resolving deltas: 100% (529/529), done.\n",
            "Submodule 'third_party/ThreadPool' (https://github.com/progschj/ThreadPool.git) registered for path 'third_party/ThreadPool'\n",
            "Submodule 'third_party/kenlm' (https://github.com/kpu/kenlm.git) registered for path 'third_party/kenlm'\n",
            "Cloning into '/content/ctcdecode/third_party/ThreadPool'...\n",
            "remote: Enumerating objects: 82, done.        \n",
            "remote: Total 82 (delta 0), reused 0 (delta 0), pack-reused 82        \n",
            "Cloning into '/content/ctcdecode/third_party/kenlm'...\n",
            "remote: Enumerating objects: 14147, done.        \n",
            "remote: Counting objects: 100% (460/460), done.        \n",
            "remote: Compressing objects: 100% (320/320), done.        \n",
            "remote: Total 14147 (delta 152), reused 398 (delta 126), pack-reused 13687        \n",
            "Receiving objects: 100% (14147/14147), 5.91 MiB | 15.16 MiB/s, done.\n",
            "Resolving deltas: 100% (8032/8032), done.\n",
            "Submodule path 'third_party/ThreadPool': checked out '9a42ec1329f259a5f4881a291db1dcb8f2ad9040'\n",
            "Submodule path 'third_party/kenlm': checked out '35835f1ac4884126458ac89f9bf6dd9ccad561e0'\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Processing /content/ctcdecode\n",
            "  Preparing metadata (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "Building wheels for collected packages: ctcdecode\n",
            "  Building wheel for ctcdecode (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for ctcdecode: filename=ctcdecode-1.0.3-cp310-cp310-linux_x86_64.whl size=19394951 sha256=f2b1aa2b06540d6fb095490a7f9a6a789303f36b34f81d4549677a3f3c997632\n",
            "  Stored in directory: /tmp/pip-ephem-wheel-cache-b3mm792a/wheels/11/99/36/d6e51fb7936a52f358678eb23a757c096f223ba07f98790e52\n",
            "Successfully built ctcdecode\n",
            "Installing collected packages: ctcdecode\n",
            "Successfully installed ctcdecode-1.0.3\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "os.chdir('/content/')\n",
        "!wget https://github.com/avyuga/captcha_recognition/raw/main/captcha.zip\n"
      ],
      "metadata": {
        "id": "YNVkmtDCVQx5",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b6ce4b1a-faf1-4bc0-ec1a-94b96627766c"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2023-05-26 07:31:04--  https://github.com/avyuga/captcha_recognition/raw/main/captcha.zip\n",
            "Resolving github.com (github.com)... 192.30.255.113\n",
            "Connecting to github.com (github.com)|192.30.255.113|:443... connected.\n",
            "HTTP request sent, awaiting response... 302 Found\n",
            "Location: https://raw.githubusercontent.com/avyuga/captcha_recognition/main/captcha.zip [following]\n",
            "--2023-05-26 07:31:04--  https://raw.githubusercontent.com/avyuga/captcha_recognition/main/captcha.zip\n",
            "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
            "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.108.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 9127102 (8.7M) [application/zip]\n",
            "Saving to: ‘captcha.zip’\n",
            "\n",
            "captcha.zip         100%[===================>]   8.70M  --.-KB/s    in 0.1s    \n",
            "\n",
            "2023-05-26 07:31:05 (79.3 MB/s) - ‘captcha.zip’ saved [9127102/9127102]\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# извлекаем файл\n",
        "import os \n",
        "from zipfile import ZipFile\n",
        "\n",
        "with ZipFile('/content/captcha.zip', 'r') as zipObj:\n",
        "   zipObj.extractall('/content/')\n",
        "\n",
        "os.chdir('/content/captcha/')"
      ],
      "metadata": {
        "id": "-uOSAcfAEYSo"
      },
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "from torch.utils.data import Dataset\n",
        "from torchvision.io import read_image\n",
        "\n",
        "from pathlib import Path\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "import torch.optim as optim\n",
        "from torch.utils.data import DataLoader\n",
        "from torch.utils.data import Subset\n",
        "from torchvision.io import ImageReadMode\n",
        "from torchvision import transforms\n",
        "\n",
        "import cv2\n",
        "from tqdm import tqdm \n",
        "\n",
        "import matplotlib.pyplot as plt"
      ],
      "metadata": {
        "id": "jiH7ah0wYAi3"
      },
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from ctcdecode import CTCBeamDecoder"
      ],
      "metadata": {
        "id": "hQgiRdMwY6E0"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "USE_GPU = True\n",
        "dtype = torch.float32\n",
        "\n",
        "if USE_GPU and torch.cuda.is_available():\n",
        "    device = torch.device('cuda')\n",
        "else:\n",
        "    device = torch.device('cpu')\n",
        "\n",
        "print('using device:', device)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fOh788FgVmnc",
        "outputId": "a24d5ba9-2880-439b-fcfa-e2df37d83eaa"
      },
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "using device: cuda\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# список и словарь из символов, которые будут использоваться далее\n",
        "vocabulary = ['2','3','4','5','6','7','8','b','c','d','e','f','g','m','n','p','w','x','y']\n",
        "char2num = {'2':0,'3':1,'4':2,'5':3,'6':4,'7':5,'8':6,'b':7,'c':8,'d':9,'e':10,'f':11,'g':12,'m':13,'n':14,'p':15,'w':16,'x':17,'y':18}"
      ],
      "metadata": {
        "id": "K6s3-MP4hjwz"
      },
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Подготовка данных"
      ],
      "metadata": {
        "id": "IYypLvP0KfEP"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "cwd = Path.cwd()\n",
        "labels = []\n",
        "for item in cwd.iterdir():\n",
        "    labels += [item.stem]\n",
        "print(f\"Number of objects: {len(labels)}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GnNh2LXKWx0k",
        "outputId": "6c1f86e8-004f-46dd-b4ff-a1f9112e232c"
      },
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Number of objects: 1070\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# так как часть файлов в формате .jpg, то удобно перевести в формат .png для единообразия\n",
        "for item in cwd.iterdir():\n",
        "    item.rename(item.with_suffix('.png'))\n",
        "\n",
        "img = cv2.imread(str(item))\n",
        "print(img.shape)"
      ],
      "metadata": {
        "id": "jht94ADiZTCj",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "64516ff7-4a4b-43c7-d47e-4b56f6ae886a"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(50, 200, 3)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Создаем класс `captcha_dataset`, наследуемый от `torch.utils.data.Dataset`, который загружает изображения в объект и применяет трансформации. Для корректной работы алгоритмов нужно обрезать края, содержащие фон.\n"
      ],
      "metadata": {
        "id": "6LpzoKbnLwT-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "class captcha_dataset(Dataset):\n",
        "    def __init__(self, labels_txt):\n",
        "        self.labels_txt = labels_txt\n",
        "        self.transform = True\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.labels)\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        label = self.labels_txt[idx]\n",
        "        img_path = f\"{label}.png\"\n",
        "        image = read_image(img_path, ImageReadMode.RGB)\n",
        "        label_num = list(map(lambda l: char2num[l], list(label)))\n",
        "        if self.transform:\n",
        "            image = transforms.functional.crop(image, 0, 25, 50, 125)\n",
        "            image = image.type(torch.float32)\n",
        "            image = transforms.functional.resize(image, size=(50,200), antialias=False)\n",
        "            image = transforms.functional.normalize(image, (128, 128, 128), (255, 255, 255))\n",
        "            # image = self.transform(image)\n",
        "        return image, label_num"
      ],
      "metadata": {
        "id": "H9e9zKUdUuqi"
      },
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "full_dataset = captcha_dataset(labels_txt=labels)\n",
        "train_idx_set, test_idx_set = train_test_split(\n",
        "    np.arange(len(labels)), test_size=0.3, random_state=42, shuffle=True)\n",
        "\n",
        "train_dataset = Subset(full_dataset, train_idx_set)\n",
        "test_dataset = Subset(full_dataset, test_idx_set)\n"
      ],
      "metadata": {
        "id": "WqHVMdBSX4S-"
      },
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Сравнение исходного и обрезанного изображения:"
      ],
      "metadata": {
        "id": "GWhPUtU6N7Jr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "figure, axes = plt.subplots(1, 2, figsize=(5,3), constrained_layout=True)\n",
        "sample_idx = np.random.randint(len(labels), size=1)\n",
        "img, label_num = full_dataset[int(sample_idx)]\n",
        "label_list = list(map(lambda l: vocabulary[l], list(label_num)))\n",
        "label = \"\".join(label_list) \n",
        "img_not_cropped = cv2.imread(f\"{label}.png\", cv2.IMREAD_UNCHANGED)\n",
        "\n",
        "axes[0].imshow(img_not_cropped, cmap='gray')\n",
        "axes[1].imshow((img.permute([1, 2, 0])+0.5).numpy().clip(0,1), cmap=\"gray\")\n",
        "\n",
        "axes[0].axis(\"off\")\n",
        "axes[1].axis(\"off\")\n",
        "\n",
        "axes[0].set_title(\"Not cropped\")\n",
        "axes[1].set_title(\"Cropped\")\n",
        "figure.suptitle(label, fontsize=16, y=0.75)\n",
        "\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 142
        },
        "id": "yUGgfoKLBJ0p",
        "outputId": "c9dba168-76ea-4020-b5b9-0de35a7280ad"
      },
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 500x300 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Больше примеров изображений:"
      ],
      "metadata": {
        "id": "ZVWS1mPaOAEK"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "figure, axes = plt.subplots(3, 3, figsize=(5,3))\n",
        "for i in range(0, 3 * 3):\n",
        "    sample_idx = torch.randint(len(train_dataset), size=(1,)).item()\n",
        "    img, label_num = train_dataset[sample_idx]\n",
        "    label_list = list(map(lambda l: vocabulary[l], list(label_num)))\n",
        "    label = \"\".join(label_list) \n",
        "    axes[i//3, i%3].imshow((img.permute([1, 2, 0])+0.5).numpy().clip(0,1), cmap=\"gray\")\n",
        "    axes[i//3, i%3].set_title(str(label))\n",
        "    axes[i//3, i%3].axis(\"off\")\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 250
        },
        "id": "zZNQip_3AEZa",
        "outputId": "6a9a6722-46a8-4893-9e8a-0639b5d1abf2"
      },
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 500x300 with 9 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Вспомогательные классы и функции\n",
        "Необходимы для удобного подсчета средней потери и точности за батч, а также для логгирования"
      ],
      "metadata": {
        "id": "i83MBNiEFrZ1"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "class ValueMeter(object):\n",
        "  \"\"\"\n",
        "  Вспомогательный класс, чтобы отслеживать loss и метрику\n",
        "  \"\"\"\n",
        "  def __init__(self):\n",
        "      self.sum = 0\n",
        "      self.total = 0\n",
        "\n",
        "  def add(self, value, n):\n",
        "      self.sum += value*n\n",
        "      self.total += n\n",
        "\n",
        "  def value(self):\n",
        "      return self.sum/self.total"
      ],
      "metadata": {
        "id": "5DtVvpBvK1Mo"
      },
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def log(mode, epoch, loss_meter=None, accuracy_meter=None):\n",
        "    line = f\"[{mode}] Epoch: {(epoch+1):d}\"\n",
        "    if loss_meter:\n",
        "        loss_line = f\" ==== Loss: {loss_meter.value():.5f}\"\n",
        "        line += loss_line\n",
        "    if accuracy_meter:\n",
        "        acc_line = f\" ==== Accuracy: {accuracy_meter.value():.5f}\"\n",
        "        line += acc_line\n",
        "    print(line)\n",
        "    "
      ],
      "metadata": {
        "id": "lw3kaOMVLT1X"
      },
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## CNN"
      ],
      "metadata": {
        "id": "R2vd_5JfFxUA"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Первый способ решения поставленной задачи - сверточная нейронная сеть. Для распознавания каптчи достаточно простейшей архитектуры, состоящей из двух сверточных блоков, которые могут также иметь нормализацию или Dropout, а затем линейный блок. У линейного блока вывод многомерный, так как для каждой из 5 букв 19 вариантов.\n",
        "\n",
        "В качестве потери используется `CrossEntropyLoss`, а точность считается как среднее число верно предсказанных букв на батч."
      ],
      "metadata": {
        "id": "m3Nz5odCOo6y"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def accuracy_cnn(outputs, labels):\n",
        "    labels = torch.argmax(labels, dim=2)\n",
        "    preds = torch.argmax(outputs, dim=2)\n",
        "    return torch.tensor(torch.sum(preds == labels).item() / (preds.size()[0]*preds.size()[1]))\n"
      ],
      "metadata": {
        "id": "qD7n7ZRQI8oc"
      },
      "execution_count": 21,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "class CNN_model(nn.Module):\n",
        "    def __init__(self, params):\n",
        "        super().__init__()\n",
        "\n",
        "        self.conv1 = nn.Conv2d(3, 32, kernel_size=(3,3), padding=1)\n",
        "        nn.init.xavier_normal_(self.conv1.weight)\n",
        "        self.conv2 = nn.Conv2d(32, 64, kernel_size=(3,3), padding=1)\n",
        "        nn.init.xavier_normal_(self.conv2.weight)\n",
        "\n",
        "        self.pool = nn.MaxPool2d(kernel_size=(2, 2))\n",
        "\n",
        "        self.batch1 = nn.BatchNorm2d(32)\n",
        "        self.batch2 = nn.BatchNorm2d(64)\n",
        "\n",
        "        self.linear1 = nn.Linear(64*12*10, 64, bias=False)\n",
        "        nn.init.xavier_normal_(self.linear1.weight)\n",
        "        self.linear2 = nn.Linear(64, 19, bias=False)\n",
        "        nn.init.xavier_normal_(self.linear2.weight)        \n",
        "\n",
        "        if 'is_batchnorm' in params.keys():\n",
        "            self.is_batchnorm = params['is_batchnorm']\n",
        "        if 'is_dropout' in params.keys():\n",
        "            self.is_dropout = params['is_dropout']\n",
        "            self.dropout_coeffs = [0.5, 0.5]\n",
        "            if 'dropout_coeffs' in params.keys():\n",
        "                self.dropout_coeffs = params['dropout_coeffs']\n",
        "   \n",
        "    def forward(self, x):\n",
        "        x = self.conv1(x)\n",
        "        x = F.relu(x)\n",
        "        if self.is_batchnorm:\n",
        "            x = self.batch1(x)\n",
        "        x = self.pool(x)\n",
        "        if self.is_dropout:\n",
        "            x = nn.Dropout(p=self.dropout_coeffs[0])(x)\n",
        "\n",
        "\n",
        "        x = self.conv2(x)\n",
        "        x = F.relu(x)\n",
        "        if self.is_batchnorm:\n",
        "            x = self.batch2(x)\n",
        "        x = self.pool(x)\n",
        "        if self.is_dropout:\n",
        "            x = nn.Dropout(p=self.dropout_coeffs[1])(x)\n",
        "        \n",
        "        N = x.shape[0]\n",
        "        x = x.permute(0, 3, 2, 1)\n",
        "        x = x.reshape(N, 5, -1)\n",
        "\n",
        "        x = self.linear1(x)\n",
        "        x = F.relu(x)\n",
        "        scores = self.linear2(x)\n",
        "\n",
        "        return scores\n"
      ],
      "metadata": {
        "id": "Rp5fFCvBI0pR"
      },
      "execution_count": 22,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def train_test_cnn(model, loaders, optimizer, criterion, epochs=1):\n",
        "    model = model.to(device)\n",
        "\n",
        "    loss_meter = {'train': ValueMeter(), 'test': ValueMeter()}\n",
        "    accuracy_meter = {'train': ValueMeter(), 'test': ValueMeter()}\n",
        "\n",
        "    loss_track = {'train': [], 'test': []}\n",
        "    accuracy_track = {'train': [], 'test': []}\n",
        "\n",
        "    for epoch in range(epochs):\n",
        "        for mode in ['train', 'test']:\n",
        "            with torch.set_grad_enabled(mode == 'train'):\n",
        "                model.train() if mode == 'train' else model.eval()\n",
        "                for x, y in tqdm(loaders[mode]):\n",
        "                    x = x.to(device, dtype=dtype)  # move to device, e.g. GPU\n",
        "                \n",
        "                    y = torch.stack(y, dim=1).to(device, dtype=torch.long)\n",
        "                    scores = model(x)\n",
        "                    groundtruth_probs = torch.zeros(scores.size())\n",
        "                    groundtruth_probs[torch.arange(y.size()[0])[:, None], torch.arange(y.size()[1])[None, :], y] = 1\n",
        "                    groundtruth_probs = groundtruth_probs.to(device, dtype=torch.float)\n",
        "                    \n",
        "                    loss = criterion(scores.permute([0, 2, 1]), y)\n",
        "\n",
        "                    acc = accuracy_cnn(scores, groundtruth_probs)\n",
        "\n",
        "                    if mode == 'train':\n",
        "                        optimizer.zero_grad()\n",
        "                        loss.backward()\n",
        "                        optimizer.step()\n",
        "\n",
        "                    batch_size = y.shape[0]\n",
        "\n",
        "                    loss_meter[mode].add(loss.item(), batch_size)\n",
        "                    accuracy_meter[mode].add(acc.item(), batch_size)\n",
        "\n",
        "            log(mode, epoch, loss_meter=loss_meter[mode], accuracy_meter=accuracy_meter[mode])\n",
        "\n",
        "            loss_track[mode].append(loss_meter[mode].value())\n",
        "            accuracy_track[mode].append(accuracy_meter[mode].value())\n",
        "\n",
        "    return {'loss': loss_track, 'accuracy': accuracy_track}"
      ],
      "metadata": {
        "id": "xXF388utggXB"
      },
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "loader_train = DataLoader(train_dataset, batch_size=8, shuffle=True)\n",
        "loader_test = DataLoader(test_dataset, batch_size=8, shuffle=False)\n",
        "\n",
        "loaders = {'train': loader_train, 'test': loader_test}"
      ],
      "metadata": {
        "id": "PL74A6Tf3bFa"
      },
      "execution_count": 24,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Сравним два варианта сверточной сети: первый - без применения нормализации и слоев Dropout, второй - с ними."
      ],
      "metadata": {
        "id": "mgKhz3KBPWfo"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "params = {'is_batchnorm': False, 'is_dropout': False}\n",
        "model = CNN_model(params)\n",
        "print(model)\n",
        "\n",
        "criterion = nn.CrossEntropyLoss(reduction='mean')\n",
        "optimizer = optim.Adam(model.parameters(), lr=1e-3)\n",
        "epochs = 50\n",
        "\n",
        "tracks_simple = train_test_cnn(model, loaders, optimizer, criterion, epochs=epochs)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VbVS4-UhPhaG",
        "outputId": "00ee6a89-3f33-476b-c9d1-fe02a8b3cb40"
      },
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "CNN_model(\n",
            "  (conv1): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "  (conv2): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "  (pool): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=0, dilation=1, ceil_mode=False)\n",
            "  (batch1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "  (batch2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "  (linear1): Linear(in_features=7680, out_features=64, bias=False)\n",
            "  (linear2): Linear(in_features=64, out_features=19, bias=False)\n",
            ")\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.45it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 1 ==== Loss: 1.71232 ==== Accuracy: 0.47690\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 85.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 1 ==== Loss: 0.73787 ==== Accuracy: 0.78816\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 66.38it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 2 ==== Loss: 1.13832 ==== Accuracy: 0.65474\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 97.81it/s] \n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 2 ==== Loss: 0.60533 ==== Accuracy: 0.82710\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 86.57it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 3 ==== Loss: 0.87924 ==== Accuracy: 0.73280\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 119.78it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 3 ==== Loss: 0.53590 ==== Accuracy: 0.84341\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 87.39it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 4 ==== Loss: 0.71597 ==== Accuracy: 0.78231\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 123.17it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 4 ==== Loss: 0.47962 ==== Accuracy: 0.85701\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 85.16it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 5 ==== Loss: 0.60599 ==== Accuracy: 0.81597\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 87.91it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 5 ==== Loss: 0.45326 ==== Accuracy: 0.86555\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 83.09it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 6 ==== Loss: 0.52279 ==== Accuracy: 0.84085\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 120.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 6 ==== Loss: 0.44526 ==== Accuracy: 0.86906\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 85.38it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 7 ==== Loss: 0.45943 ==== Accuracy: 0.86000\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 117.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 7 ==== Loss: 0.43808 ==== Accuracy: 0.87138\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 85.04it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 8 ==== Loss: 0.40983 ==== Accuracy: 0.87503\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.91it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 8 ==== Loss: 0.42809 ==== Accuracy: 0.87430\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.82it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 9 ==== Loss: 0.36901 ==== Accuracy: 0.88744\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 84.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 9 ==== Loss: 0.41923 ==== Accuracy: 0.87747\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 66.58it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 10 ==== Loss: 0.33749 ==== Accuracy: 0.89688\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 84.66it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 10 ==== Loss: 0.42209 ==== Accuracy: 0.87826\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 63.11it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 11 ==== Loss: 0.31038 ==== Accuracy: 0.90513\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.48it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 11 ==== Loss: 0.41366 ==== Accuracy: 0.88202\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 84.86it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 12 ==== Loss: 0.28641 ==== Accuracy: 0.91242\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 114.24it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 12 ==== Loss: 0.40914 ==== Accuracy: 0.88479\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 83.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 13 ==== Loss: 0.26546 ==== Accuracy: 0.91876\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.64it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 13 ==== Loss: 0.40470 ==== Accuracy: 0.88732\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 85.80it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 14 ==== Loss: 0.24710 ==== Accuracy: 0.92439\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.41it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 14 ==== Loss: 0.40206 ==== Accuracy: 0.88945\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 84.84it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 15 ==== Loss: 0.23104 ==== Accuracy: 0.92933\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.89it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 15 ==== Loss: 0.39889 ==== Accuracy: 0.89188\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 82.71it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 16 ==== Loss: 0.21755 ==== Accuracy: 0.93349\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 119.03it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 16 ==== Loss: 0.39922 ==== Accuracy: 0.89346\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 84.86it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 17 ==== Loss: 0.20555 ==== Accuracy: 0.93719\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 85.11it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 17 ==== Loss: 0.39736 ==== Accuracy: 0.89514\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 68.09it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 18 ==== Loss: 0.19472 ==== Accuracy: 0.94047\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 85.70it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 18 ==== Loss: 0.39922 ==== Accuracy: 0.89571\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 76.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 19 ==== Loss: 0.18589 ==== Accuracy: 0.94317\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 117.75it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 19 ==== Loss: 0.40686 ==== Accuracy: 0.89549\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 86.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 20 ==== Loss: 0.18074 ==== Accuracy: 0.94458\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.31it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 20 ==== Loss: 0.40840 ==== Accuracy: 0.89598\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 77.05it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 21 ==== Loss: 0.17449 ==== Accuracy: 0.94634\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 86.69it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 21 ==== Loss: 0.41066 ==== Accuracy: 0.89637\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 67.37it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 22 ==== Loss: 0.16825 ==== Accuracy: 0.94829\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 85.84it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 22 ==== Loss: 0.41032 ==== Accuracy: 0.89734\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.50it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 23 ==== Loss: 0.16137 ==== Accuracy: 0.95044\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.75it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 23 ==== Loss: 0.40858 ==== Accuracy: 0.89855\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 84.27it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 24 ==== Loss: 0.15483 ==== Accuracy: 0.95245\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.90it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 24 ==== Loss: 0.40772 ==== Accuracy: 0.89964\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 79.42it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 25 ==== Loss: 0.14871 ==== Accuracy: 0.95435\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 85.03it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 25 ==== Loss: 0.40710 ==== Accuracy: 0.90071\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 66.45it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 26 ==== Loss: 0.14301 ==== Accuracy: 0.95611\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 84.11it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 26 ==== Loss: 0.40599 ==== Accuracy: 0.90175\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 76.29it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 27 ==== Loss: 0.13773 ==== Accuracy: 0.95773\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 117.75it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 27 ==== Loss: 0.40511 ==== Accuracy: 0.90271\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 83.66it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 28 ==== Loss: 0.13282 ==== Accuracy: 0.95924\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 28 ==== Loss: 0.40442 ==== Accuracy: 0.90360\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 84.75it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 29 ==== Loss: 0.12825 ==== Accuracy: 0.96065\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 114.94it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 29 ==== Loss: 0.40395 ==== Accuracy: 0.90444\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 71.30it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 30 ==== Loss: 0.12398 ==== Accuracy: 0.96196\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 72.83it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 30 ==== Loss: 0.40357 ==== Accuracy: 0.90521\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 71.57it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 31 ==== Loss: 0.11999 ==== Accuracy: 0.96319\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 118.61it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 31 ==== Loss: 0.40331 ==== Accuracy: 0.90598\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 84.50it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 32 ==== Loss: 0.11624 ==== Accuracy: 0.96434\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.54it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 32 ==== Loss: 0.40318 ==== Accuracy: 0.90668\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 77.61it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 33 ==== Loss: 0.11272 ==== Accuracy: 0.96542\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 82.25it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 33 ==== Loss: 0.40324 ==== Accuracy: 0.90734\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 67.47it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 34 ==== Loss: 0.10941 ==== Accuracy: 0.96643\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 86.86it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 34 ==== Loss: 0.40332 ==== Accuracy: 0.90795\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.84it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 35 ==== Loss: 0.10629 ==== Accuracy: 0.96739\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 117.33it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 35 ==== Loss: 0.40346 ==== Accuracy: 0.90857\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 83.84it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 36 ==== Loss: 0.10334 ==== Accuracy: 0.96830\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 121.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 36 ==== Loss: 0.40365 ==== Accuracy: 0.90916\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 84.87it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 37 ==== Loss: 0.10055 ==== Accuracy: 0.96916\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 117.58it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 37 ==== Loss: 0.40393 ==== Accuracy: 0.90969\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 85.18it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 38 ==== Loss: 0.09790 ==== Accuracy: 0.96997\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.45it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 38 ==== Loss: 0.40426 ==== Accuracy: 0.91020\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 84.32it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 39 ==== Loss: 0.09540 ==== Accuracy: 0.97074\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 118.27it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 39 ==== Loss: 0.40461 ==== Accuracy: 0.91071\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 83.63it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 40 ==== Loss: 0.09301 ==== Accuracy: 0.97147\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.98it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 40 ==== Loss: 0.40502 ==== Accuracy: 0.91120\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 83.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 41 ==== Loss: 0.09075 ==== Accuracy: 0.97216\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 94.40it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 41 ==== Loss: 0.40550 ==== Accuracy: 0.91165\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 67.19it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 42 ==== Loss: 0.08859 ==== Accuracy: 0.97283\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 87.05it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 42 ==== Loss: 0.40599 ==== Accuracy: 0.91211\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 72.17it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 43 ==== Loss: 0.08653 ==== Accuracy: 0.97346\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 107.28it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 43 ==== Loss: 0.40654 ==== Accuracy: 0.91256\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 83.29it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 44 ==== Loss: 0.08456 ==== Accuracy: 0.97406\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 114.90it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 44 ==== Loss: 0.40712 ==== Accuracy: 0.91294\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 84.13it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 45 ==== Loss: 0.08269 ==== Accuracy: 0.97464\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.61it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 45 ==== Loss: 0.40773 ==== Accuracy: 0.91335\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 83.41it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 46 ==== Loss: 0.08089 ==== Accuracy: 0.97519\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 113.55it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 46 ==== Loss: 0.40838 ==== Accuracy: 0.91373\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 85.23it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 47 ==== Loss: 0.07917 ==== Accuracy: 0.97572\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 113.32it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 47 ==== Loss: 0.40904 ==== Accuracy: 0.91411\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 84.42it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 48 ==== Loss: 0.07752 ==== Accuracy: 0.97622\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 117.60it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 48 ==== Loss: 0.40972 ==== Accuracy: 0.91447\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 85.15it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 49 ==== Loss: 0.07594 ==== Accuracy: 0.97671\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 117.27it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 49 ==== Loss: 0.41043 ==== Accuracy: 0.91481\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 69.76it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 50 ==== Loss: 0.07442 ==== Accuracy: 0.97717\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 87.72it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 50 ==== Loss: 0.41115 ==== Accuracy: 0.91514\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "params = {'is_batchnorm': True, 'is_dropout': True, 'dropout_coeffs':[0.1, 0.1]}\n",
        "model = CNN_model(params)\n",
        "print(model)\n",
        "\n",
        "criterion = nn.CrossEntropyLoss(reduction='mean')\n",
        "optimizer = optim.Adam(model.parameters(), lr=1e-3)\n",
        "epochs = 50\n",
        "\n",
        "tracks = train_test_cnn(model, loaders, optimizer, criterion, epochs=epochs)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YbivZqguf_3V",
        "outputId": "10e07ecc-7618-474d-843a-9fb4e9395752"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "CNN_model(\n",
            "  (conv1): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "  (conv2): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "  (pool): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=0, dilation=1, ceil_mode=False)\n",
            "  (batch1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "  (batch2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "  (linear1): Linear(in_features=7680, out_features=64, bias=False)\n",
            "  (linear2): Linear(in_features=64, out_features=19, bias=False)\n",
            ")\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 54.84it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 1 ==== Loss: 2.25295 ==== Accuracy: 0.31375\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.46it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 1 ==== Loss: 1.42286 ==== Accuracy: 0.54642\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.95it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 2 ==== Loss: 1.56165 ==== Accuracy: 0.51963\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.57it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 2 ==== Loss: 1.02322 ==== Accuracy: 0.68287\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.59it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 3 ==== Loss: 1.18555 ==== Accuracy: 0.63498\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.32it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 3 ==== Loss: 0.85067 ==== Accuracy: 0.74102\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.04it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 4 ==== Loss: 0.95938 ==== Accuracy: 0.70421\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.36it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 4 ==== Loss: 0.72422 ==== Accuracy: 0.77897\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.54it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 5 ==== Loss: 0.80244 ==== Accuracy: 0.75258\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 117.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 5 ==== Loss: 0.64360 ==== Accuracy: 0.80349\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.39it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 6 ==== Loss: 0.68938 ==== Accuracy: 0.78781\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 117.26it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 6 ==== Loss: 0.59278 ==== Accuracy: 0.81994\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.91it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 7 ==== Loss: 0.60480 ==== Accuracy: 0.81320\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.80it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 7 ==== Loss: 0.55493 ==== Accuracy: 0.83124\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 64.28it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 8 ==== Loss: 0.53776 ==== Accuracy: 0.83411\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 85.43it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 8 ==== Loss: 0.52159 ==== Accuracy: 0.84252\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 69.12it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 9 ==== Loss: 0.48459 ==== Accuracy: 0.85047\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 117.43it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 9 ==== Loss: 0.49857 ==== Accuracy: 0.85061\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.58it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 10 ==== Loss: 0.44053 ==== Accuracy: 0.86411\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.97it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 10 ==== Loss: 0.48065 ==== Accuracy: 0.85670\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.84it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 11 ==== Loss: 0.40323 ==== Accuracy: 0.87576\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 113.15it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 11 ==== Loss: 0.46518 ==== Accuracy: 0.86191\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.56it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 12 ==== Loss: 0.37258 ==== Accuracy: 0.88540\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.36it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 12 ==== Loss: 0.45288 ==== Accuracy: 0.86677\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.38it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 13 ==== Loss: 0.34521 ==== Accuracy: 0.89403\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 113.54it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 13 ==== Loss: 0.44156 ==== Accuracy: 0.87160\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.37it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 14 ==== Loss: 0.32179 ==== Accuracy: 0.90134\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.03it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 14 ==== Loss: 0.43001 ==== Accuracy: 0.87552\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.19it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 15 ==== Loss: 0.30100 ==== Accuracy: 0.90781\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 112.63it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 15 ==== Loss: 0.42178 ==== Accuracy: 0.87867\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 64.34it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 16 ==== Loss: 0.28269 ==== Accuracy: 0.91350\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 86.20it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 16 ==== Loss: 0.41331 ==== Accuracy: 0.88193\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 67.84it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 17 ==== Loss: 0.26668 ==== Accuracy: 0.91851\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.68it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 17 ==== Loss: 0.40746 ==== Accuracy: 0.88463\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.07it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 18 ==== Loss: 0.25218 ==== Accuracy: 0.92301\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 113.23it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 18 ==== Loss: 0.40338 ==== Accuracy: 0.88664\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.26it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 19 ==== Loss: 0.23949 ==== Accuracy: 0.92692\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 114.07it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 19 ==== Loss: 0.40117 ==== Accuracy: 0.88805\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 20 ==== Loss: 0.22923 ==== Accuracy: 0.93009\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 114.83it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 20 ==== Loss: 0.42143 ==== Accuracy: 0.88551\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.09it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 21 ==== Loss: 0.22448 ==== Accuracy: 0.93126\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.26it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 21 ==== Loss: 0.42820 ==== Accuracy: 0.88527\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 79.98it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 22 ==== Loss: 0.22058 ==== Accuracy: 0.93256\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 114.60it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 22 ==== Loss: 0.42871 ==== Accuracy: 0.88595\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 79.81it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 23 ==== Loss: 0.21437 ==== Accuracy: 0.93450\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 113.52it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 23 ==== Loss: 0.43642 ==== Accuracy: 0.88620\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 63.98it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 24 ==== Loss: 0.20728 ==== Accuracy: 0.93659\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 83.25it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 24 ==== Loss: 0.43592 ==== Accuracy: 0.88762\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 66.89it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 25 ==== Loss: 0.20053 ==== Accuracy: 0.93868\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 112.76it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 25 ==== Loss: 0.43305 ==== Accuracy: 0.88890\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 79.64it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 26 ==== Loss: 0.19337 ==== Accuracy: 0.94090\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.35it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 26 ==== Loss: 0.43049 ==== Accuracy: 0.89049\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 78.97it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 27 ==== Loss: 0.18657 ==== Accuracy: 0.94297\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 113.94it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 27 ==== Loss: 0.42813 ==== Accuracy: 0.89200\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.92it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 28 ==== Loss: 0.18006 ==== Accuracy: 0.94496\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 114.88it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 28 ==== Loss: 0.42480 ==== Accuracy: 0.89348\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.70it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 29 ==== Loss: 0.17395 ==== Accuracy: 0.94685\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 114.16it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 29 ==== Loss: 0.42253 ==== Accuracy: 0.89503\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 78.71it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 30 ==== Loss: 0.16819 ==== Accuracy: 0.94862\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 114.83it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 30 ==== Loss: 0.41986 ==== Accuracy: 0.89651\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 79.20it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 31 ==== Loss: 0.16282 ==== Accuracy: 0.95026\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 106.33it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 31 ==== Loss: 0.41758 ==== Accuracy: 0.89774\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 63.80it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 32 ==== Loss: 0.15778 ==== Accuracy: 0.95182\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 83.26it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 32 ==== Loss: 0.41577 ==== Accuracy: 0.89903\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 69.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 33 ==== Loss: 0.15303 ==== Accuracy: 0.95327\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.06it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 33 ==== Loss: 0.41328 ==== Accuracy: 0.90048\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.90it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 34 ==== Loss: 0.14855 ==== Accuracy: 0.95465\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 114.02it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 34 ==== Loss: 0.41129 ==== Accuracy: 0.90161\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.57it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 35 ==== Loss: 0.14433 ==== Accuracy: 0.95594\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 116.14it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 35 ==== Loss: 0.40884 ==== Accuracy: 0.90282\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.45it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 36 ==== Loss: 0.14033 ==== Accuracy: 0.95717\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 118.25it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 36 ==== Loss: 0.40743 ==== Accuracy: 0.90393\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.35it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 37 ==== Loss: 0.13655 ==== Accuracy: 0.95832\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 109.12it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 37 ==== Loss: 0.40531 ==== Accuracy: 0.90496\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 38 ==== Loss: 0.13297 ==== Accuracy: 0.95942\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 119.70it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 38 ==== Loss: 0.40357 ==== Accuracy: 0.90589\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 39 ==== Loss: 0.12957 ==== Accuracy: 0.96046\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 102.91it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 39 ==== Loss: 0.40202 ==== Accuracy: 0.90677\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 62.27it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 40 ==== Loss: 0.12634 ==== Accuracy: 0.96145\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 83.17it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 40 ==== Loss: 0.40057 ==== Accuracy: 0.90766\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 70.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 41 ==== Loss: 0.12326 ==== Accuracy: 0.96239\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.52it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 41 ==== Loss: 0.39876 ==== Accuracy: 0.90853\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 80.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 42 ==== Loss: 0.12034 ==== Accuracy: 0.96328\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 108.21it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 42 ==== Loss: 0.39776 ==== Accuracy: 0.90929\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.70it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 43 ==== Loss: 0.11754 ==== Accuracy: 0.96414\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 113.80it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 43 ==== Loss: 0.39609 ==== Accuracy: 0.91016\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.35it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 44 ==== Loss: 0.11527 ==== Accuracy: 0.96483\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 109.67it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 44 ==== Loss: 0.40003 ==== Accuracy: 0.91024\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 81.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 45 ==== Loss: 0.11678 ==== Accuracy: 0.96457\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 118.83it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 45 ==== Loss: 0.40671 ==== Accuracy: 0.90944\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 82.66it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 46 ==== Loss: 0.11719 ==== Accuracy: 0.96447\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 113.52it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 46 ==== Loss: 0.41256 ==== Accuracy: 0.90912\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 82.52it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 47 ==== Loss: 0.11630 ==== Accuracy: 0.96474\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 115.18it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 47 ==== Loss: 0.41384 ==== Accuracy: 0.90923\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 64.29it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 48 ==== Loss: 0.11452 ==== Accuracy: 0.96529\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 84.14it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 48 ==== Loss: 0.41488 ==== Accuracy: 0.90937\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 68.49it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 49 ==== Loss: 0.11253 ==== Accuracy: 0.96589\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 113.92it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 49 ==== Loss: 0.41496 ==== Accuracy: 0.90972\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:01<00:00, 78.63it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 50 ==== Loss: 0.11041 ==== Accuracy: 0.96654\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:00<00:00, 114.44it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 50 ==== Loss: 0.41476 ==== Accuracy: 0.91014\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "loss_track_simple, accuracy_track_simple = tracks_simple['loss'], tracks_simple['accuracy']\n",
        "loss_track, accuracy_track = tracks['loss'], tracks['accuracy']\n",
        "fig, axes = plt.subplots(1, 2, constrained_layout=True, figsize=(9, 3))\n",
        "\n",
        "axes[0].plot(loss_track_simple['train'], 'b', label='train w/o')\n",
        "axes[0].plot(loss_track['train'], 'b--', label='train')\n",
        "axes[0].plot(loss_track_simple['test'], 'r', label='test w/o')\n",
        "axes[0].plot(loss_track['test'], 'r--', label='test')\n",
        "axes[0].set_title('Model loss')\n",
        "axes[0].legend()\n",
        "\n",
        "axes[1].plot(accuracy_track_simple['train'], 'b', label='train w/o')\n",
        "axes[1].plot(accuracy_track['train'], 'b--', label='train')\n",
        "axes[1].plot(accuracy_track_simple['test'], 'r', label='test w/o')\n",
        "axes[1].plot(accuracy_track['test'], 'r--', label='test')\n",
        "\n",
        "axes[1].set_title('Model accuracy')\n",
        "axes[1].legend()\n",
        "plt.show()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 328
        },
        "id": "-X84CrEZCe90",
        "outputId": "1d51c521-f301-4315-a645-8c39d2108052"
      },
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 900x300 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## LSTM"
      ],
      "metadata": {
        "id": "GZUL0nSJF3FG"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Модель LSTM состоит из двух частей: первая часть - сверточная, аналогичная рассмотренной выше; вторая - рекуррентная. В качестве рекуррентного слоя используется двунаправленный LSTM, после него - линейный слой (в отличие от архитектуры CNN, одномерный). Для использования рекуррентного слоя необходимо изменить форму тензора, что и отображено в функции `forward()`. "
      ],
      "metadata": {
        "id": "LuJXRcO4XUxK"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# initial size (N, 3, 50, 200)\n",
        "class LSTM_model(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(LSTM_model, self).__init__()\n",
        "\n",
        "        self.conv1 = nn.Conv2d(3, 32, kernel_size=(3,3), padding=1)\n",
        "        nn.init.xavier_uniform_(self.conv1.weight)\n",
        "\n",
        "        self.conv2 = nn.Conv2d(32, 64, kernel_size=(3,3), padding=1)\n",
        "        nn.init.xavier_uniform_(self.conv2.weight)\n",
        "\n",
        "        self.pool = nn.MaxPool2d(kernel_size=(2,2))\n",
        "        self.lstm = nn.LSTM(768, 64, bidirectional=True, batch_first=True)\n",
        "\n",
        "        self.final = nn.Linear(128, 20)\n",
        "        nn.init.xavier_uniform_(self.final.weight)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.conv1(x)\n",
        "        x = F.relu(x)\n",
        "        x = self.pool(x)\n",
        "\n",
        "        x = self.conv2(x)\n",
        "        x = F.relu(x)\n",
        "        x = self.pool(x)\n",
        "\n",
        "        N, C, w, h = x.size()\n",
        "        x = x.view(N, -1, h)\n",
        "        x = x.permute(2, 0, 1)\n",
        "        x, _ = self.lstm(x)\n",
        "        return self.final(x)\n",
        "    \n"
      ],
      "metadata": {
        "id": "ODdrCcrbF46Q"
      },
      "execution_count": 28,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Подсчет точности модели LSTM также выполняется сравнением предсказанных букв с оригинальными. В данной ситуации, если длина слова не равна 5, то точность его предсказания считается равной 0. Для того, чтобы из вывода модели получить буквы, воспользуемся специальным декодером. Библиотека PyTorch встроенным декодером не обладает, поэтому используется сторонняя библиотека.\n",
        "\n",
        "Также при тренировки модели будет изменяться `learning_rate` оптимизатора для избежания попадания в локальный минимум. Поэтому будет также отображен график изменения `lr` в зависимости от эпохи."
      ],
      "metadata": {
        "id": "j5WNcK4PVqHp"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def accuracy_lstm(out, labels, decoder):\n",
        "    out = nn.functional.log_softmax(out, dim=2)\n",
        "    out = out.permute(1, 0, 2)\n",
        "\n",
        "    beam_results, beam_scores, timesteps, out_len = decoder.decode(out)\n",
        "    N = len(labels)\n",
        "    acc = 0\n",
        "    for i in range(N):\n",
        "        preds = beam_results[i][0][:out_len[i][0]].numpy()\n",
        "        if len(preds) == 5:\n",
        "            acc += np.sum(preds == labels[i]) / 5\n",
        "    return acc / N"
      ],
      "metadata": {
        "id": "sJZUQc3FjRzB"
      },
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def train_test_LSTM(model, loaders, optimizer, scheduler, decoder, epochs=1):\n",
        "    model = model.to(device)\n",
        "    loss_meter = {'train': ValueMeter(), 'test': ValueMeter()}\n",
        "    accuracy_meter = {'train': ValueMeter(), 'test': ValueMeter()}\n",
        "\n",
        "    loss_track = {'train': [], 'test': []}\n",
        "    accuracy_track = {'train': [], 'test': []}\n",
        "    lr = []\n",
        "\n",
        "    for epoch in range(epochs):\n",
        "        for mode in ['train', 'test']:\n",
        "            with torch.set_grad_enabled(mode == 'train'):\n",
        "                model.train() if mode == 'train' else model.eval()\n",
        "                for x, y in tqdm(loaders[mode]):\n",
        "                    \n",
        "                    x = x.to(device, dtype=dtype)  # move to device, e.g. GPU\n",
        "                    y = torch.stack(y, dim=1).to(device, dtype=torch.long)\n",
        "                    \n",
        "                    out = model(x)\n",
        "                    T, N = out.size(0), out.size(1)\n",
        "                \n",
        "                    scores = nn.functional.log_softmax(out, dim=2)\n",
        "                    input_lengths = torch.full(size=(N,), fill_value=T, dtype=torch.int32)\n",
        "                    target_lengths = torch.full(size=(N,), fill_value=5, dtype=torch.int32)\n",
        "                    \n",
        "                    loss = criterion(scores, y, input_lengths=input_lengths, target_lengths=target_lengths)\n",
        "                    acc = accuracy_lstm(out, y.cpu().numpy(), decoder)\n",
        "\n",
        "                    if mode == 'train':\n",
        "                        optimizer.zero_grad()\n",
        "                        loss.backward()\n",
        "                        optimizer.step()\n",
        "                        scheduler.step()\n",
        "\n",
        "                    batch_size = y.shape[0]\n",
        "\n",
        "                    loss_meter[mode].add(loss.item(), batch_size)\n",
        "                    accuracy_meter[mode].add(acc, batch_size)\n",
        "                    \n",
        "            log(mode, epoch, loss_meter[mode], accuracy_meter[mode])\n",
        "            loss_track[mode].append(loss_meter[mode].value())\n",
        "            accuracy_track[mode].append(accuracy_meter[mode].value())\n",
        "        lr += [scheduler.get_last_lr()]\n",
        "        print(f\"Epoch: {epoch+1} ==== learning rate: {scheduler.get_last_lr()}\")\n",
        "\n",
        "    return {'loss': loss_track, 'accuracy': accuracy_track, 'learning_rate': lr}"
      ],
      "metadata": {
        "id": "SIbBpA6_H3e8"
      },
      "execution_count": 30,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "loader_train = DataLoader(train_dataset, batch_size=8, shuffle=True)\n",
        "loader_test = DataLoader(test_dataset, batch_size=8, shuffle=False)\n",
        "\n",
        "loaders = {'train': loader_train, 'test': loader_test}"
      ],
      "metadata": {
        "id": "pWSzGiWHWSnM"
      },
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model = LSTM_model()\n",
        "epochs = 100\n",
        "optimizer = optim.Adam(model.parameters(), lr=1e-3)\n",
        "\n",
        "scheduler = optim.lr_scheduler.OneCycleLR(optimizer, max_lr=1e-2, epochs=epochs, steps_per_epoch=94, div_factor=10, pct_start=0.2)\n",
        "\n",
        "criterion = torch.nn.CTCLoss(blank=19, zero_infinity=False)\n",
        "\n",
        "labels_list = list(char2num.keys()) + list('_')\n",
        "decoder = CTCBeamDecoder(labels_list, blank_id=19, log_probs_input=True, alpha=1)"
      ],
      "metadata": {
        "id": "PsSXZuoDK9Cz"
      },
      "execution_count": 32,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "tracks = train_test_LSTM(model, loaders, optimizer, scheduler, decoder, epochs=epochs)"
      ],
      "metadata": {
        "id": "qVsIUw6BIHpt",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "cdcb8ef5-a911-4bbf-b47d-2cddddc7111b"
      },
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:12<00:00,  7.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 1 ==== Loss: 4.40764 ==== Accuracy: 0.00027\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:02<00:00, 14.32it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 1 ==== Loss: 3.31272 ==== Accuracy: 0.00000\n",
            "Epoch: 1 ==== learning rate: [0.0010554613317121271]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 12.58it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 2 ==== Loss: 3.84962 ==== Accuracy: 0.00013\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 11.48it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 2 ==== Loss: 3.29458 ==== Accuracy: 0.00000\n",
            "Epoch: 2 ==== learning rate: [0.001220478233819499]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:06<00:00, 13.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 3 ==== Loss: 3.64960 ==== Accuracy: 0.00009\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 3 ==== Loss: 3.26635 ==== Accuracy: 0.00000\n",
            "Epoch: 3 ==== learning rate: [0.0014909831253683888]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 12.32it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 4 ==== Loss: 3.50948 ==== Accuracy: 0.00007\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:02<00:00, 14.33it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 4 ==== Loss: 3.17725 ==== Accuracy: 0.00000\n",
            "Epoch: 4 ==== learning rate: [0.0018603082012388153]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.06it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 5 ==== Loss: 3.31538 ==== Accuracy: 0.00005\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 13.34it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 5 ==== Loss: 2.94988 ==== Accuracy: 0.00062\n",
            "Epoch: 5 ==== learning rate: [0.0023193497900785816]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.48it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 6 ==== Loss: 3.01078 ==== Accuracy: 0.00872\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 13.07it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 6 ==== Loss: 2.62748 ==== Accuracy: 0.02150\n",
            "Epoch: 6 ==== learning rate: [0.0028567927550751035]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00,  9.78it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 7 ==== Loss: 2.69219 ==== Accuracy: 0.05226\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.61it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 7 ==== Loss: 2.34397 ==== Accuracy: 0.08278\n",
            "Epoch: 7 ==== learning rate: [0.003459389406202596]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.46it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 8 ==== Loss: 2.42516 ==== Accuracy: 0.09963\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.11it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 8 ==== Loss: 2.11131 ==== Accuracy: 0.13294\n",
            "Epoch: 8 ==== learning rate: [0.004112286048902481]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.06it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 9 ==== Loss: 2.20633 ==== Accuracy: 0.15054\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00, 10.15it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 9 ==== Loss: 1.92899 ==== Accuracy: 0.18609\n",
            "Epoch: 9 ==== learning rate: [0.0047993891199416385]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.94it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 10 ==== Loss: 2.02311 ==== Accuracy: 0.20005\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  8.89it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 10 ==== Loss: 1.77441 ==== Accuracy: 0.23283\n",
            "Epoch: 10 ==== learning rate: [0.005503761885389706]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.65it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 11 ==== Loss: 1.86717 ==== Accuracy: 0.24743\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:05<00:00,  8.13it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 11 ==== Loss: 1.64828 ==== Accuracy: 0.26661\n",
            "Epoch: 11 ==== learning rate: [0.006208041922316235]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.42it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 12 ==== Loss: 1.73677 ==== Accuracy: 0.28652\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.47it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 12 ==== Loss: 1.53742 ==== Accuracy: 0.29964\n",
            "Epoch: 12 ==== learning rate: [0.006894869093500686]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.20it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 13 ==== Loss: 1.62295 ==== Accuracy: 0.32316\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.48it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 13 ==== Loss: 1.44260 ==== Accuracy: 0.33276\n",
            "Epoch: 13 ==== learning rate: [0.007547313465800855]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.24it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 14 ==== Loss: 1.52552 ==== Accuracy: 0.35409\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 14 ==== Loss: 1.36752 ==== Accuracy: 0.35260\n",
            "Epoch: 14 ==== learning rate: [0.008149292624214138]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.19it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 15 ==== Loss: 1.44012 ==== Accuracy: 0.38287\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.61it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 15 ==== Loss: 1.29414 ==== Accuracy: 0.37499\n",
            "Epoch: 15 ==== learning rate: [0.008685968095056739]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.27it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 16 ==== Loss: 1.36332 ==== Accuracy: 0.40975\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.66it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 16 ==== Loss: 1.23026 ==== Accuracy: 0.39576\n",
            "Epoch: 16 ==== learning rate: [0.009144111106635251]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.18it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 17 ==== Loss: 1.29483 ==== Accuracy: 0.43357\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.50it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 17 ==== Loss: 1.17251 ==== Accuracy: 0.41646\n",
            "Epoch: 17 ==== learning rate: [0.009512428671600006]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.73it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 18 ==== Loss: 1.23404 ==== Accuracy: 0.45419\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 11.02it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 18 ==== Loss: 1.12395 ==== Accuracy: 0.43236\n",
            "Epoch: 18 ==== learning rate: [0.009781841953219489]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.78it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 19 ==== Loss: 1.18057 ==== Accuracy: 0.47135\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.26it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 19 ==== Loss: 1.07960 ==== Accuracy: 0.44729\n",
            "Epoch: 19 ==== learning rate: [0.009945710053991676]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.97it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 20 ==== Loss: 1.13142 ==== Accuracy: 0.48842\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.19it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 20 ==== Loss: 1.03743 ==== Accuracy: 0.46243\n",
            "Epoch: 20 ==== learning rate: [0.009999999563685422]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.70it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 21 ==== Loss: 1.08653 ==== Accuracy: 0.50390\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 10.92it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 21 ==== Loss: 1.00054 ==== Accuracy: 0.47168\n",
            "Epoch: 21 ==== learning rate: [0.00999606277771523]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.67it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 22 ==== Loss: 1.04709 ==== Accuracy: 0.51634\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.64it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 22 ==== Loss: 0.96929 ==== Accuracy: 0.47958\n",
            "Epoch: 22 ==== learning rate: [0.009984422502160032]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.25it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 23 ==== Loss: 1.01115 ==== Accuracy: 0.52795\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.55it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 23 ==== Loss: 0.93711 ==== Accuracy: 0.49194\n",
            "Epoch: 23 ==== learning rate: [0.009965096685481028]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.22it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 24 ==== Loss: 0.97707 ==== Accuracy: 0.54052\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.59it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 24 ==== Loss: 0.90702 ==== Accuracy: 0.50382\n",
            "Epoch: 24 ==== learning rate: [0.009938115126686775]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.23it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 25 ==== Loss: 0.94499 ==== Accuracy: 0.55315\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.46it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 25 ==== Loss: 0.88105 ==== Accuracy: 0.51237\n",
            "Epoch: 25 ==== learning rate: [0.009903519429385272]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.27it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 26 ==== Loss: 0.91630 ==== Accuracy: 0.56268\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.60it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 26 ==== Loss: 0.85599 ==== Accuracy: 0.52257\n",
            "Epoch: 26 ==== learning rate: [0.00986136293763421]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.24it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 27 ==== Loss: 0.88859 ==== Accuracy: 0.57315\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.59it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 27 ==== Loss: 0.83225 ==== Accuracy: 0.53163\n",
            "Epoch: 27 ==== learning rate: [0.009811710653688306]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.54it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 28 ==== Loss: 0.86391 ==== Accuracy: 0.58178\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 11.32it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 28 ==== Loss: 0.80940 ==== Accuracy: 0.54014\n",
            "Epoch: 28 ==== learning rate: [0.009754639137770537]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.52it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 29 ==== Loss: 0.83917 ==== Accuracy: 0.59196\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.45it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 29 ==== Loss: 0.78838 ==== Accuracy: 0.54998\n",
            "Epoch: 29 ==== learning rate: [0.00969023639002182]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.85it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 30 ==== Loss: 0.81614 ==== Accuracy: 0.60128\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.13it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 30 ==== Loss: 0.76897 ==== Accuracy: 0.55772\n",
            "Epoch: 30 ==== learning rate: [0.009618601714811182]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.82it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 31 ==== Loss: 0.79638 ==== Accuracy: 0.60775\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 10.68it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 31 ==== Loss: 0.75295 ==== Accuracy: 0.56256\n",
            "Epoch: 31 ==== learning rate: [0.00953984556761561]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.40it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 32 ==== Loss: 0.77681 ==== Accuracy: 0.61456\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  8.58it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 32 ==== Loss: 0.73644 ==== Accuracy: 0.56947\n",
            "Epoch: 32 ==== learning rate: [0.009454089384705721]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.84it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 33 ==== Loss: 0.75872 ==== Accuracy: 0.62097\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.58it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 33 ==== Loss: 0.72239 ==== Accuracy: 0.57411\n",
            "Epoch: 33 ==== learning rate: [0.009361465395899831]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.23it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 34 ==== Loss: 0.74114 ==== Accuracy: 0.62772\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 34 ==== Loss: 0.70889 ==== Accuracy: 0.57783\n",
            "Epoch: 34 ==== learning rate: [0.009262116420675159]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.15it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 35 ==== Loss: 0.72423 ==== Accuracy: 0.63431\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 35 ==== Loss: 0.69460 ==== Accuracy: 0.58409\n",
            "Epoch: 35 ==== learning rate: [0.00915619564795054]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.24it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 36 ==== Loss: 0.70877 ==== Accuracy: 0.63936\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.59it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 36 ==== Loss: 0.68189 ==== Accuracy: 0.58842\n",
            "Epoch: 36 ==== learning rate: [0.009043866399880227]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.13it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 37 ==== Loss: 0.69388 ==== Accuracy: 0.64523\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.39it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 37 ==== Loss: 0.66868 ==== Accuracy: 0.59365\n",
            "Epoch: 37 ==== learning rate: [0.008925301880022959]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.20it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 38 ==== Loss: 0.67956 ==== Accuracy: 0.65125\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 38 ==== Loss: 0.65600 ==== Accuracy: 0.59882\n",
            "Epoch: 38 ==== learning rate: [0.008800684906274628]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.80it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 39 ==== Loss: 0.66577 ==== Accuracy: 0.65664\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 10.93it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 39 ==== Loss: 0.64456 ==== Accuracy: 0.60343\n",
            "Epoch: 39 ==== learning rate: [0.008670207628976334]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.46it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 40 ==== Loss: 0.65270 ==== Accuracy: 0.66208\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.47it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 40 ==== Loss: 0.63341 ==== Accuracy: 0.60888\n",
            "Epoch: 40 ==== learning rate: [0.00853407123463249]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.76it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 41 ==== Loss: 0.63996 ==== Accuracy: 0.66782\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  8.89it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 41 ==== Loss: 0.62314 ==== Accuracy: 0.61280\n",
            "Epoch: 41 ==== learning rate: [0.008392485635695812]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.84it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 42 ==== Loss: 0.62802 ==== Accuracy: 0.67265\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00, 10.10it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 42 ==== Loss: 0.61406 ==== Accuracy: 0.61675\n",
            "Epoch: 42 ==== learning rate: [0.00824566914689753]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.12it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 43 ==== Loss: 0.61676 ==== Accuracy: 0.67708\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.25it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 43 ==== Loss: 0.60451 ==== Accuracy: 0.62101\n",
            "Epoch: 43 ==== learning rate: [0.008093848148621902]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.39it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 44 ==== Loss: 0.60567 ==== Accuracy: 0.68194\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 44 ==== Loss: 0.59624 ==== Accuracy: 0.62417\n",
            "Epoch: 44 ==== learning rate: [0.00793725673784406]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.20it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 45 ==== Loss: 0.59526 ==== Accuracy: 0.68647\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.67it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 45 ==== Loss: 0.58761 ==== Accuracy: 0.62780\n",
            "Epoch: 45 ==== learning rate: [0.007776136367169456]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.27it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 46 ==== Loss: 0.58608 ==== Accuracy: 0.68994\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.76it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 46 ==== Loss: 0.57961 ==== Accuracy: 0.63148\n",
            "Epoch: 46 ==== learning rate: [0.007610735472531429]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.16it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 47 ==== Loss: 0.57642 ==== Accuracy: 0.69429\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.43it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 47 ==== Loss: 0.57164 ==== Accuracy: 0.63530\n",
            "Epoch: 47 ==== learning rate: [0.0074413090901210175]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.13it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 48 ==== Loss: 0.56694 ==== Accuracy: 0.69876\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.73it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 48 ==== Loss: 0.56414 ==== Accuracy: 0.63923\n",
            "Epoch: 48 ==== learning rate: [0.007268118463139627]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.78it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 49 ==== Loss: 0.55769 ==== Accuracy: 0.70311\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 10.94it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 49 ==== Loss: 0.55647 ==== Accuracy: 0.64341\n",
            "Epoch: 49 ==== learning rate: [0.007091430638980957]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.76it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 50 ==== Loss: 0.54875 ==== Accuracy: 0.70728\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.11it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 50 ==== Loss: 0.54919 ==== Accuracy: 0.64703\n",
            "Epoch: 50 ==== learning rate: [0.0069115180574632696]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.69it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 51 ==== Loss: 0.54017 ==== Accuracy: 0.71128\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.51it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 51 ==== Loss: 0.54229 ==== Accuracy: 0.65027\n",
            "Epoch: 51 ==== learning rate: [0.006728658130746944]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.66it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 52 ==== Loss: 0.53182 ==== Accuracy: 0.71515\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 11.29it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 52 ==== Loss: 0.53554 ==== Accuracy: 0.65392\n",
            "Epoch: 52 ==== learning rate: [0.006543132815585023]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00,  9.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 53 ==== Loss: 0.52359 ==== Accuracy: 0.71903\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00, 10.23it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 53 ==== Loss: 0.52948 ==== Accuracy: 0.65673\n",
            "Epoch: 53 ==== learning rate: [0.006355228178566343]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.64it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 54 ==== Loss: 0.51575 ==== Accuracy: 0.72280\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.53it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 54 ==== Loss: 0.52355 ==== Accuracy: 0.65994\n",
            "Epoch: 54 ==== learning rate: [0.006165233955021588]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.21it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 55 ==== Loss: 0.50806 ==== Accuracy: 0.72666\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.45it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 55 ==== Loss: 0.51785 ==== Accuracy: 0.66272\n",
            "Epoch: 55 ==== learning rate: [0.005973443102272408]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.29it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 56 ==== Loss: 0.50052 ==== Accuracy: 0.73055\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.60it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 56 ==== Loss: 0.51226 ==== Accuracy: 0.66530\n",
            "Epoch: 56 ==== learning rate: [0.005780151347912462]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.17it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 57 ==== Loss: 0.49314 ==== Accuracy: 0.73432\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.70it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 57 ==== Loss: 0.50662 ==== Accuracy: 0.66825\n",
            "Epoch: 57 ==== learning rate: [0.005585656733816909]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.27it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 58 ==== Loss: 0.48597 ==== Accuracy: 0.73815\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 58 ==== Loss: 0.50128 ==== Accuracy: 0.67085\n",
            "Epoch: 58 ==== learning rate: [0.005390259156583433]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.40it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 59 ==== Loss: 0.47907 ==== Accuracy: 0.74184\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 11.95it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 59 ==== Loss: 0.49590 ==== Accuracy: 0.67378\n",
            "Epoch: 59 ==== learning rate: [0.005194259905113429]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.21it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 60 ==== Loss: 0.47230 ==== Accuracy: 0.74545\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.93it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 60 ==== Loss: 0.49097 ==== Accuracy: 0.67679\n",
            "Epoch: 60 ==== learning rate: [0.004997961196046356]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.79it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 61 ==== Loss: 0.46561 ==== Accuracy: 0.74913\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  8.98it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 61 ==== Loss: 0.48615 ==== Accuracy: 0.67966\n",
            "Epoch: 61 ==== learning rate: [0.00480166570776358]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.69it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 62 ==== Loss: 0.45927 ==== Accuracy: 0.75239\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.86it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 62 ==== Loss: 0.48123 ==== Accuracy: 0.68253\n",
            "Epoch: 62 ==== learning rate: [0.004605676113680251]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.32it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 63 ==== Loss: 0.45305 ==== Accuracy: 0.75567\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 11.93it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 63 ==== Loss: 0.47681 ==== Accuracy: 0.68503\n",
            "Epoch: 63 ==== learning rate: [0.004410294615544839]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.51it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 64 ==== Loss: 0.44696 ==== Accuracy: 0.75888\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.78it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 64 ==== Loss: 0.47242 ==== Accuracy: 0.68761\n",
            "Epoch: 64 ==== learning rate: [0.0042158224774659315]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.22it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 65 ==== Loss: 0.44101 ==== Accuracy: 0.76199\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.78it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 65 ==== Loss: 0.46823 ==== Accuracy: 0.68988\n",
            "Epoch: 65 ==== learning rate: [0.004022559561384829]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.19it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 66 ==== Loss: 0.43511 ==== Accuracy: 0.76515\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.59it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 66 ==== Loss: 0.46411 ==== Accuracy: 0.69246\n",
            "Epoch: 66 ==== learning rate: [0.0038308038647101497]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.23it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 67 ==== Loss: 0.42935 ==== Accuracy: 0.76828\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.59it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 67 ==== Loss: 0.46030 ==== Accuracy: 0.69490\n",
            "Epoch: 67 ==== learning rate: [0.0036408510608274373]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.19it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 68 ==== Loss: 0.42373 ==== Accuracy: 0.77134\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.60it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 68 ==== Loss: 0.45651 ==== Accuracy: 0.69744\n",
            "Epoch: 68 ==== learning rate: [0.0034529940431922254]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 69 ==== Loss: 0.41829 ==== Accuracy: 0.77423\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 11.07it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 69 ==== Loss: 0.45281 ==== Accuracy: 0.69991\n",
            "Epoch: 69 ==== learning rate: [0.0032675224737095565]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.49it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 70 ==== Loss: 0.41295 ==== Accuracy: 0.77715\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.52it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 70 ==== Loss: 0.44912 ==== Accuracy: 0.70231\n",
            "Epoch: 70 ==== learning rate: [0.003084722336096329]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.88it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 71 ==== Loss: 0.40771 ==== Accuracy: 0.77999\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.06it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 71 ==== Loss: 0.44560 ==== Accuracy: 0.70469\n",
            "Epoch: 71 ==== learning rate: [0.002904875494915121]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.75it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 72 ==== Loss: 0.40259 ==== Accuracy: 0.78282\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 10.68it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 72 ==== Loss: 0.44220 ==== Accuracy: 0.70693\n",
            "Epoch: 72 ==== learning rate: [0.0027282592609594787]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.94it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 73 ==== Loss: 0.39755 ==== Accuracy: 0.78554\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.51it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 73 ==== Loss: 0.43889 ==== Accuracy: 0.70903\n",
            "Epoch: 73 ==== learning rate: [0.0025551459636607592]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:10<00:00,  9.06it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 74 ==== Loss: 0.39262 ==== Accuracy: 0.78828\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 11.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 74 ==== Loss: 0.43577 ==== Accuracy: 0.71094\n",
            "Epoch: 74 ==== learning rate: [0.0023858025311758934]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.22it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 75 ==== Loss: 0.38779 ==== Accuracy: 0.79093\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.78it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 75 ==== Loss: 0.43280 ==== Accuracy: 0.71273\n",
            "Epoch: 75 ==== learning rate: [0.0022204900788035196]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.18it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 76 ==== Loss: 0.38307 ==== Accuracy: 0.79354\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.77it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 76 ==== Loss: 0.42986 ==== Accuracy: 0.71460\n",
            "Epoch: 76 ==== learning rate: [0.002059463506363107]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.11it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 77 ==== Loss: 0.37844 ==== Accuracy: 0.79609\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.44it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 77 ==== Loss: 0.42690 ==== Accuracy: 0.71654\n",
            "Epoch: 77 ==== learning rate: [0.0019029711051579243]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.16it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 78 ==== Loss: 0.37390 ==== Accuracy: 0.79859\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.64it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 78 ==== Loss: 0.42409 ==== Accuracy: 0.71845\n",
            "Epoch: 78 ==== learning rate: [0.0017512541751278292]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.12it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 79 ==== Loss: 0.36945 ==== Accuracy: 0.80102\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.50it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 79 ==== Loss: 0.42136 ==== Accuracy: 0.72037\n",
            "Epoch: 79 ==== learning rate: [0.0016045466527822684]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.51it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 80 ==== Loss: 0.36511 ==== Accuracy: 0.80340\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 10.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 80 ==== Loss: 0.41869 ==== Accuracy: 0.72213\n",
            "Epoch: 80 ==== learning rate: [0.001463074750487127]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.32it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 81 ==== Loss: 0.36085 ==== Accuracy: 0.80575\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 81 ==== Loss: 0.41608 ==== Accuracy: 0.72388\n",
            "Epoch: 81 ==== learning rate: [0.001327056607661653]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 82 ==== Loss: 0.35670 ==== Accuracy: 0.80803\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  8.91it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 82 ==== Loss: 0.41354 ==== Accuracy: 0.72559\n",
            "Epoch: 82 ==== learning rate: [0.0011967019544232944]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.88it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 83 ==== Loss: 0.35262 ==== Accuracy: 0.81028\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 10.39it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 83 ==== Loss: 0.41105 ==== Accuracy: 0.72737\n",
            "Epoch: 83 ==== learning rate: [0.0010722117881990496]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 84 ==== Loss: 0.34863 ==== Accuracy: 0.81245\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.38it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 84 ==== Loss: 0.40864 ==== Accuracy: 0.72905\n",
            "Epoch: 84 ==== learning rate: [0.0009537780638020075]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.25it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 85 ==== Loss: 0.34472 ==== Accuracy: 0.81457\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.71it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 85 ==== Loss: 0.40633 ==== Accuracy: 0.73070\n",
            "Epoch: 85 ==== learning rate: [0.0008415833974509442]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.18it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 86 ==== Loss: 0.34091 ==== Accuracy: 0.81665\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.41it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 86 ==== Loss: 0.40404 ==== Accuracy: 0.73236\n",
            "Epoch: 86 ==== learning rate: [0.0007358007851893379]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.02it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 87 ==== Loss: 0.33717 ==== Accuracy: 0.81869\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.62it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 87 ==== Loss: 0.40182 ==== Accuracy: 0.73392\n",
            "Epoch: 87 ==== learning rate: [0.0006365933361380223]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.18it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 88 ==== Loss: 0.33351 ==== Accuracy: 0.82069\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.59it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 88 ==== Loss: 0.39966 ==== Accuracy: 0.73547\n",
            "Epoch: 88 ==== learning rate: [0.0005441140209927326]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.23it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 89 ==== Loss: 0.32993 ==== Accuracy: 0.82266\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.45it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 89 ==== Loss: 0.39756 ==== Accuracy: 0.73707\n",
            "Epoch: 89 ==== learning rate: [0.00045850543615439017]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 10.46it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 90 ==== Loss: 0.32642 ==== Accuracy: 0.82457\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 11.65it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 90 ==== Loss: 0.39551 ==== Accuracy: 0.73862\n",
            "Epoch: 90 ==== learning rate: [0.00037989958385578187]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.29it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 91 ==== Loss: 0.32299 ==== Accuracy: 0.82644\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 91 ==== Loss: 0.39351 ==== Accuracy: 0.74007\n",
            "Epoch: 91 ==== learning rate: [0.0003084176686236848]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.82it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 92 ==== Loss: 0.31963 ==== Accuracy: 0.82826\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  8.97it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 92 ==== Loss: 0.39157 ==== Accuracy: 0.74148\n",
            "Epoch: 92 ==== learning rate: [0.00024416991039027545]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:07<00:00, 11.78it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 93 ==== Loss: 0.31634 ==== Accuracy: 0.83006\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:04<00:00,  9.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 93 ==== Loss: 0.38966 ==== Accuracy: 0.74293\n",
            "Epoch: 93 ==== learning rate: [0.00018725537454197938]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:08<00:00, 11.08it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 94 ==== Loss: 0.31313 ==== Accuracy: 0.83182\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.18it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 94 ==== Loss: 0.38781 ==== Accuracy: 0.74428\n",
            "Epoch: 94 ==== learning rate: [0.00013776181916782626]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:10<00:00,  9.02it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 95 ==== Loss: 0.30997 ==== Accuracy: 0.83354\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.23it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 95 ==== Loss: 0.38599 ==== Accuracy: 0.74560\n",
            "Epoch: 95 ==== learning rate: [9.576555974283876e-05]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.28it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 96 ==== Loss: 0.30689 ==== Accuracy: 0.83523\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.68it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 96 ==== Loss: 0.38421 ==== Accuracy: 0.74693\n",
            "Epoch: 96 ==== learning rate: [6.133135145509998e-05]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.15it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 97 ==== Loss: 0.30387 ==== Accuracy: 0.83687\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.60it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 97 ==== Loss: 0.38246 ==== Accuracy: 0.74826\n",
            "Epoch: 97 ==== learning rate: [3.4512289357949965e-05]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.23it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 98 ==== Loss: 0.30091 ==== Accuracy: 0.83847\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.70it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 98 ==== Loss: 0.38075 ==== Accuracy: 0.74956\n",
            "Epoch: 98 ==== learning rate: [1.5349726501256596e-05]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.14it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 99 ==== Loss: 0.29801 ==== Accuracy: 0.84005\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.51it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 99 ==== Loss: 0.37907 ==== Accuracy: 0.75084\n",
            "Epoch: 99 ==== learning rate: [3.87321016801582e-06]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 94/94 [00:09<00:00, 10.16it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[train] Epoch: 100 ==== Loss: 0.29516 ==== Accuracy: 0.84159\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 41/41 [00:03<00:00, 12.57it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[test] Epoch: 100 ==== Loss: 0.37743 ==== Accuracy: 0.75209\n",
            "Epoch: 100 ==== learning rate: [1.0043631457778787e-07]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "loss_track, acc_track = tracks['loss'], tracks['accuracy']\n",
        "\n",
        "fig, axes = plt.subplots(1, 3, constrained_layout=True, figsize=(12, 3))\n",
        "axes[0].plot(loss_track['train'], label='train')\n",
        "axes[0].plot(loss_track['test'], label='test')\n",
        "axes[0].set_title('Model loss')\n",
        "axes[0].legend()\n",
        "\n",
        "axes[1].plot(accuracy_track['train'], label='train')\n",
        "axes[1].plot(accuracy_track['test'], label='test')\n",
        "axes[1].set_title('Model accuracy')\n",
        "axes[1].legend()\n",
        "\n",
        "axes[2].plot(tracks['learning_rate'])\n",
        "axes[2].set_title('Learning rate')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 314
        },
        "id": "K2ZiSFKl_gX1",
        "outputId": "32c86d50-f3ae-46ab-f100-8b7aff3bad63"
      },
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1200x300 with 3 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "По полученным графикам потерь видно, что к 100-ой эпохе потери на тестовом датасете практически не изменяются, а на тренировочном медленно убывают. Это означает, что если продолжать тренировку, то модель может переобучиться на тренировочном датасете и потери на тестовом начнут расти. Аналогичная ситуация и с точностью. Точность на тестовом датасете установилась на уровне 73%, на тренировочном еще растет. При установке других оптимизаторов и дольшей тренировке возможно получить настолько же точные результаты, как и у сверточной нейросети."
      ],
      "metadata": {
        "id": "cUI5XQ6vWgPa"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Выводы\n",
        "\n",
        "Обе архитектуры нейронных сетей справляются с поставленной задачей. В случае со сверточной нейронной сетью точность достигает 91% на тренировочном датасете, с LSTM - порядка 75%. При этом если в сверточную нейросеть добавлять нормализацию и dropout, то скорость сходимости уменьшается. При параметре p=0.1 к 50-ой эпохе модели сходятся к одинаковым значениям loss и accuracy."
      ],
      "metadata": {
        "id": "NINDJgWBEa_k"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "JH-9NCzZ_ltZ"
      },
      "execution_count": 34,
      "outputs": []
    }
  ]
}