{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "-"
    }
   },
   "source": [
    "### Uczenie maszynowe — zastosowania\n",
    "# Sieci neuronowe (PyTorch)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Przykład implementacji sieci neuronowej do rozpoznawania cyfr ze [zbioru MNIST](https://en.wikipedia.org/wiki/MNIST_database), według https://github.com/pytorch/examples/tree/master/mnist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "from torchvision import datasets, transforms\n",
    "from torch.optim.lr_scheduler import StepLR\n",
    "\n",
    "\n",
    "# Ta sieć ma taką samą architekturę, jak sieć z pliku Sieci_neuronowe_Keras.ipynb\n",
    "\n",
    "class Net(nn.Module):\n",
    "    \"\"\"W PyTorchu tworzenie sieci neuronowej\n",
    "    polega na zdefiniowaniu klasy, która dziedziczy z nn.Module.\n",
    "    \"\"\"\n",
    "    \n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        \n",
    "        # Warstwy splotowe\n",
    "        self.conv1 = nn.Conv2d(1, 32, 3, 1)\n",
    "        self.conv2 = nn.Conv2d(32, 64, 3, 1)\n",
    "        \n",
    "        # Warstwa dropout\n",
    "        self.dropout = nn.Dropout(0.5)\n",
    "        \n",
    "        # Warstwa liniowa (gęsta)\n",
    "        self.dense = nn.Linear(9216, 10)\n",
    "\n",
    "    def forward(self, x):\n",
    "        \"\"\"Definiujemy przechodzenie \"do przodu\" jako kolejne przekształcenia wejścia x\"\"\"\n",
    "        x = self.conv1(x)\n",
    "        x = F.relu(x)\n",
    "        x = F.max_pool2d(x, 2)\n",
    "        x = self.conv2(x)\n",
    "        x = F.relu(x)\n",
    "        x = F.max_pool2d(x, 2)\n",
    "        x = torch.flatten(x, 1)\n",
    "        x = self.dropout(x)\n",
    "        x = self.dense(x)\n",
    "        output = F.log_softmax(x, dim=1)\n",
    "        return output\n",
    "\n",
    "\n",
    "def train(model, device, train_loader, optimizer, epoch, log_interval, dry_run):\n",
    "    \"\"\"Uczenie modelu\"\"\"\n",
    "    model.train()\n",
    "    for batch_idx, (data, target) in enumerate(train_loader):\n",
    "        data, target = data.to(device), target.to(device)  # wrzucenie danych na kartę graficzną (jeśli dotyczy)\n",
    "        optimizer.zero_grad()  # wyzerowanie gradientu\n",
    "        output = model(data)  # przejście \"do przodu\"\n",
    "        loss = F.nll_loss(output, target)  # obliczenie funkcji kosztu\n",
    "        loss.backward()  # propagacja wsteczna\n",
    "        optimizer.step()  # krok optymalizatora\n",
    "        \n",
    "        # Wypisanie wartości funkcji kosztu na ekran\n",
    "        if batch_idx % log_interval == 0:\n",
    "            print('Train Epoch: {} [{}/{} ({:.0f}%)]\\tLoss: {:.6f}'.format(\n",
    "                epoch, batch_idx * len(data), len(train_loader.dataset),\n",
    "                100. * batch_idx / len(train_loader), loss.item()))\n",
    "            if dry_run:\n",
    "                break\n",
    "\n",
    "\n",
    "def test(model, device, test_loader):\n",
    "    \"\"\"Testowanie modelu\"\"\"\n",
    "    model.eval()\n",
    "    test_loss = 0\n",
    "    correct = 0\n",
    "    with torch.no_grad():  # nie musimy przechowywać gradientów, bo nie wykonujemy propagacji wstecznej\n",
    "        for data, target in test_loader:\n",
    "            data, target = data.to(device), target.to(device)  # wrzucenie danych na kartę graficzną (jeśli dotyczy)\n",
    "            output = model(data)  # przejście \"do przodu\"\n",
    "            test_loss += F.nll_loss(output, target, reduction='sum').item()  # suma kosztów z każdego batcha\n",
    "            pred = output.argmax(dim=1, keepdim=True)  # predykcja na podstawie maks. logarytmu prawdopodobieństwa\n",
    "            correct += pred.eq(target.view_as(pred)).sum().item()\n",
    "\n",
    "    test_loss /= len(test_loader.dataset)  # obliczenie kosztu na zbiorze testowym\n",
    "\n",
    "    print('\\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\\n'.format(\n",
    "        test_loss, correct, len(test_loader.dataset),\n",
    "        100. * correct / len(test_loader.dataset)))\n",
    "\n",
    "\n",
    "def run(\n",
    "        batch_size=64,\n",
    "        test_batch_size=1000,\n",
    "        epochs=14,\n",
    "        lr=1.0,\n",
    "        gamma=0.7,\n",
    "        no_cuda=False,\n",
    "        dry_run=False,\n",
    "        seed=1,\n",
    "        log_interval=10,\n",
    "        save_model=False,\n",
    "    ):\n",
    "    \"\"\"Główna funkcja służąca do uruchamiania przykładu.\n",
    "    \n",
    "    Argumenty:\n",
    "        batch_size - wielkość batcha podczas uczenia (default: 64),\n",
    "        test_batch_size - wielkość batcha podczas testowania (default: 1000)\n",
    "        epochs - liczba epok uczenia (default: 14)\n",
    "        lr - współczynnik uczenia (learning rate) (default: 1.0)\n",
    "        gamma - współczynnik gamma (dla optymalizatora) (default: 0.7)\n",
    "        no_cuda - wyłącza uczenie na karcie graficznej (default: False)\n",
    "        dry_run - szybko (\"na sucho\") sprawdza pojedyncze przejście (default: False)\n",
    "        seed - ziarno generatora liczb pseudolosowych (default: 1)\n",
    "        log_interval - interwał logowania stanu uczenia (default: 10)\n",
    "        save_model - zapisuje bieżący model (default: False)\n",
    "    \"\"\"\n",
    "    # Ustawienie ziarna generatora liczb pseudolosowych\n",
    "    torch.manual_seed(seed)\n",
    "\n",
    "    # Ustawienie, czy uczenie ma odbywać się z wykorzystaniem karty graficznej\n",
    "    use_cuda = no_cuda and torch.cuda.is_available()\n",
    "    device = torch.device(\"cuda\" if use_cuda else \"cpu\")\n",
    "\n",
    "    # Ustawienie parametrów wsadu\n",
    "    train_kwargs = {'batch_size': batch_size}\n",
    "    test_kwargs = {'batch_size': test_batch_size}\n",
    "    if use_cuda:\n",
    "        cuda_kwargs = {'num_workers': 1,\n",
    "                       'pin_memory': True,\n",
    "                       'shuffle': True}\n",
    "        train_kwargs.update(cuda_kwargs)\n",
    "        test_kwargs.update(cuda_kwargs)\n",
    "\n",
    "    transform=transforms.Compose([\n",
    "        transforms.ToTensor(),\n",
    "        transforms.Normalize((0.1307,), (0.3081,))\n",
    "        ])\n",
    "    \n",
    "    # Załadowanie danych\n",
    "    dataset1 = datasets.MNIST('../data', train=True, download=True,\n",
    "                       transform=transform)\n",
    "    dataset2 = datasets.MNIST('../data', train=False,\n",
    "                       transform=transform)\n",
    "    \n",
    "    # Klasa DataLoader ułatwia zarządzanie danymi uczącymi\n",
    "    train_loader = torch.utils.data.DataLoader(dataset1,**train_kwargs)\n",
    "    test_loader = torch.utils.data.DataLoader(dataset2, **test_kwargs)\n",
    "\n",
    "    # Stworzenie modelu - klasa Net to zdefiniowana przez nas sieć neuronowa\n",
    "    model = Net().to(device)\n",
    "    \n",
    "    # Wybór metody optymalizacji (tutaj: Adadelta)\n",
    "    optimizer = optim.Adadelta(model.parameters(), lr=lr)\n",
    "\n",
    "    # Scheduler - kwestie techniczne\n",
    "    scheduler = StepLR(optimizer, step_size=1, gamma=gamma)\n",
    "    \n",
    "    # Pętla główna uczenia i testowania - kolejne epoki\n",
    "    for epoch in range(1, epochs + 1):\n",
    "        train(model, device, train_loader, optimizer, epoch, log_interval, dry_run)\n",
    "        test(model, device, test_loader)\n",
    "        scheduler.step()\n",
    "\n",
    "    # Opcjonalne zapisanie wytrenowanego modelu do pliku\n",
    "    if save_model:\n",
    "        torch.save(model.state_dict(), \"mnist_cnn.pt\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Uwaga**: uruchomienie tego przykładu długo trwa. Żeby trwało krócej, można zmniejszyć liczbę epok."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to ../data\\MNIST\\raw\\train-images-idx3-ubyte.gz\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "02f11492b1ff4fdfa05d7dd086c5989b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=1.0, bar_style='info', max=1.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "ename": "HTTPError",
     "evalue": "HTTP Error 503: Service Unavailable",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mHTTPError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-3-c80278f8b80c>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mrun\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mepochs\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m5\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdry_run\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32m<ipython-input-2-55aa66586ee8>\u001b[0m in \u001b[0;36mrun\u001b[1;34m(batch_size, test_batch_size, epochs, lr, gamma, no_cuda, dry_run, seed, log_interval, save_model)\u001b[0m\n\u001b[0;32m    125\u001b[0m         \u001b[0mtransforms\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mNormalize\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m0.1307\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m(\u001b[0m\u001b[1;36m0.3081\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    126\u001b[0m         ])\n\u001b[1;32m--> 127\u001b[1;33m     dataset1 = datasets.MNIST('../data', train=True, download=True,\n\u001b[0m\u001b[0;32m    128\u001b[0m                        transform=transform)\n\u001b[0;32m    129\u001b[0m     dataset2 = datasets.MNIST('../data', train=False,\n",
      "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\torchvision\\datasets\\mnist.py\u001b[0m in \u001b[0;36m__init__\u001b[1;34m(self, root, train, transform, target_transform, download)\u001b[0m\n\u001b[0;32m     77\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     78\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mdownload\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 79\u001b[1;33m             \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdownload\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     80\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     81\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_check_exists\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\torchvision\\datasets\\mnist.py\u001b[0m in \u001b[0;36mdownload\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    144\u001b[0m         \u001b[1;32mfor\u001b[0m \u001b[0murl\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmd5\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mresources\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    145\u001b[0m             \u001b[0mfilename\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0murl\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mrpartition\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'/'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m2\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 146\u001b[1;33m             \u001b[0mdownload_and_extract_archive\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0murl\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdownload_root\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mraw_folder\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfilename\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mfilename\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmd5\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mmd5\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    147\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    148\u001b[0m         \u001b[1;31m# process and save as torch files\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\torchvision\\datasets\\utils.py\u001b[0m in \u001b[0;36mdownload_and_extract_archive\u001b[1;34m(url, download_root, extract_root, filename, md5, remove_finished)\u001b[0m\n\u001b[0;32m    254\u001b[0m         \u001b[0mfilename\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mos\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbasename\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0murl\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    255\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 256\u001b[1;33m     \u001b[0mdownload_url\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0murl\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdownload_root\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfilename\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmd5\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    257\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    258\u001b[0m     \u001b[0marchive\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mos\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mjoin\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdownload_root\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfilename\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\torchvision\\datasets\\utils.py\u001b[0m in \u001b[0;36mdownload_url\u001b[1;34m(url, root, filename, md5)\u001b[0m\n\u001b[0;32m     82\u001b[0m                 )\n\u001b[0;32m     83\u001b[0m             \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 84\u001b[1;33m                 \u001b[1;32mraise\u001b[0m \u001b[0me\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     85\u001b[0m         \u001b[1;31m# check integrity of downloaded file\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     86\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0mcheck_integrity\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mfpath\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmd5\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\torchvision\\datasets\\utils.py\u001b[0m in \u001b[0;36mdownload_url\u001b[1;34m(url, root, filename, md5)\u001b[0m\n\u001b[0;32m     68\u001b[0m         \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     69\u001b[0m             \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'Downloading '\u001b[0m \u001b[1;33m+\u001b[0m \u001b[0murl\u001b[0m \u001b[1;33m+\u001b[0m \u001b[1;34m' to '\u001b[0m \u001b[1;33m+\u001b[0m \u001b[0mfpath\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 70\u001b[1;33m             urllib.request.urlretrieve(\n\u001b[0m\u001b[0;32m     71\u001b[0m                 \u001b[0murl\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfpath\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     72\u001b[0m                 \u001b[0mreporthook\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mgen_bar_updater\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\urllib\\request.py\u001b[0m in \u001b[0;36murlretrieve\u001b[1;34m(url, filename, reporthook, data)\u001b[0m\n\u001b[0;32m    245\u001b[0m     \u001b[0murl_type\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mpath\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0m_splittype\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0murl\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    246\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 247\u001b[1;33m     \u001b[1;32mwith\u001b[0m \u001b[0mcontextlib\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mclosing\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0murlopen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0murl\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdata\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0mfp\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    248\u001b[0m         \u001b[0mheaders\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mfp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0minfo\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    249\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\urllib\\request.py\u001b[0m in \u001b[0;36murlopen\u001b[1;34m(url, data, timeout, cafile, capath, cadefault, context)\u001b[0m\n\u001b[0;32m    220\u001b[0m     \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    221\u001b[0m         \u001b[0mopener\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0m_opener\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 222\u001b[1;33m     \u001b[1;32mreturn\u001b[0m \u001b[0mopener\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mopen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0murl\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdata\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtimeout\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    223\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    224\u001b[0m \u001b[1;32mdef\u001b[0m \u001b[0minstall_opener\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mopener\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\urllib\\request.py\u001b[0m in \u001b[0;36mopen\u001b[1;34m(self, fullurl, data, timeout)\u001b[0m\n\u001b[0;32m    529\u001b[0m         \u001b[1;32mfor\u001b[0m \u001b[0mprocessor\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mprocess_response\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mprotocol\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m[\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    530\u001b[0m             \u001b[0mmeth\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mgetattr\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mprocessor\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmeth_name\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 531\u001b[1;33m             \u001b[0mresponse\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmeth\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mreq\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mresponse\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    532\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    533\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0mresponse\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\urllib\\request.py\u001b[0m in \u001b[0;36mhttp_response\u001b[1;34m(self, request, response)\u001b[0m\n\u001b[0;32m    638\u001b[0m         \u001b[1;31m# request was successfully received, understood, and accepted.\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    639\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[1;33m(\u001b[0m\u001b[1;36m200\u001b[0m \u001b[1;33m<=\u001b[0m \u001b[0mcode\u001b[0m \u001b[1;33m<\u001b[0m \u001b[1;36m300\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 640\u001b[1;33m             response = self.parent.error(\n\u001b[0m\u001b[0;32m    641\u001b[0m                 'http', request, response, code, msg, hdrs)\n\u001b[0;32m    642\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\urllib\\request.py\u001b[0m in \u001b[0;36merror\u001b[1;34m(self, proto, *args)\u001b[0m\n\u001b[0;32m    567\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mhttp_err\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    568\u001b[0m             \u001b[0margs\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m(\u001b[0m\u001b[0mdict\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m'default'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m'http_error_default'\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m+\u001b[0m \u001b[0morig_args\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 569\u001b[1;33m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_call_chain\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    570\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    571\u001b[0m \u001b[1;31m# XXX probably also want an abstract factory that knows when it makes\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\urllib\\request.py\u001b[0m in \u001b[0;36m_call_chain\u001b[1;34m(self, chain, kind, meth_name, *args)\u001b[0m\n\u001b[0;32m    500\u001b[0m         \u001b[1;32mfor\u001b[0m \u001b[0mhandler\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mhandlers\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    501\u001b[0m             \u001b[0mfunc\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mgetattr\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mhandler\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmeth_name\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 502\u001b[1;33m             \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mfunc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    503\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mresult\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    504\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0mresult\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\urllib\\request.py\u001b[0m in \u001b[0;36mhttp_error_default\u001b[1;34m(self, req, fp, code, msg, hdrs)\u001b[0m\n\u001b[0;32m    647\u001b[0m \u001b[1;32mclass\u001b[0m \u001b[0mHTTPDefaultErrorHandler\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mBaseHandler\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    648\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mhttp_error_default\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mreq\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfp\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcode\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmsg\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mhdrs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 649\u001b[1;33m         \u001b[1;32mraise\u001b[0m \u001b[0mHTTPError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mreq\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfull_url\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcode\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmsg\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mhdrs\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfp\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    650\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    651\u001b[0m \u001b[1;32mclass\u001b[0m \u001b[0mHTTPRedirectHandler\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mBaseHandler\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mHTTPError\u001b[0m: HTTP Error 503: Service Unavailable"
     ]
    }
   ],
   "source": [
    "run(epochs=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "celltoolbar": "Slideshow",
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.3"
  },
  "livereveal": {
   "start_slideshow_at": "selected",
   "theme": "amu"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}