Implementing WideResNet from scratch using fast.ai - MNIST

fastai-wideresnet-mnist.ipynb

{
  "cells": [
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## WideResNet - MNIST"
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "https://arxiv.org/pdf/1605.07146.pdf"
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "NOTE: This is a study. I am comprehending to the best of my ability, but there may be mistakes. \n\nI have to first start out by saying that the concept of \"widening\" confused me for quite some time. Even after looking through code implementations, it wasn't clear to me what \"widening\" was or how it was being achieved. After reviewing the paper and different implementations multiple times, I think I'm starting to understand. What follows is a discussion on understanding how to widen a ResNet. Feel free to skip down below for the implementation. \n\nThe first thing to know is that the basic structure of the network relies on our familiar ResBlocks:"
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "![](https://cdn-images-1.medium.com/max/1200/1*ByrVJspW-TefwlH7OLxNkg.png)"
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "https://arxiv.org/pdf/1512.03385.pdf"
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "As depicted above, each ResBlock consists of two convolutional layers (referred to as \"weight layer\" above) , one being passed into the other, and then the output of these two layers is then added to the original input $X$ to form a \"skip\" or \"identity\" connection. \n\nIf our first convolutional layer is $g_{1}(x)$ and our second is $g_{2}(x)$, we can think of a ResBlock as $g_{2}(g_{1}(X)) + X$, setting aside the intermediate ReLU activation for this discussion. Now, a given convolutional layer $g_{n}(x)$ is going to have a set of feature maps, and the number of feature maps in that layer is referred to as the number of \"channels.\" What is a feature map? It is the grid that describes our input. A 64x64 RGB image can be thought of as having 3 different feature maps (for each color channel) of size 64x64. \n\nWhen we pass that image into a typical model, the first convolutional layer $g_{1}(x)$ will usually reduce the size of the feature map by half, from 64x64 to 32x32 (we'll see later that WRNs do not do this for the first layer). In order to retain information from our input whilst reducing the size of the feature map, we correspondingly increase the number of channels; in other words, we typically increase the number of feature maps that we want the model to output whenever we make those feature maps smaller. If we start with an image of (3 channels, 64px, 64px), then our first convolution might output (8 channels, 32px, 32px). In a hypothetical scenario, we may want more than 8 channels of feature maps to work with at the start, in which case we may choose to have the convolution output (16 channels, 32px, 32px) instead. Choosing an appropriate number of starting channels is a topic for a different discussion. \n\nThe main thesis of the WideResNet paper is that increasing both the number of convolutional layers $g(x)$ and the number of channels will improve the performance of a given ResNet model, but that at some point, there is a significant diminishing return when increasing the depth of a model versus increasing its width. The more layers a model has, the deeper it is; the more channels there are within each layer, the wider the model. The authors argue that increasing the number of channels per layer yields better results and is more computationally efficient than simply increasing the number of layers, and they seem to show this from their experiments. \n\nGoing back to our previous example, if we wanted to make our model deeper, we would increase the number $n$ of our layers $g(x)$. If instead we wanted to make our model wider, we would not change the number of layers $n$, but we would increase the number of channels from 8 to 16, or 8 to 24. We can represent how much we want to widen each layer by using the multiplier $k$. If $k = 1$, then we have not widened our model. If $k = 2$, then the output from our first layer would have 16 channels instead of 8. If $k = 3$, then our layer would widen from 8 channels to 24, and so on."
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "Now that we understand how to \"widen\" our model, we can proceed to build a WideResNet from scratch. Let's first prepare our data. "
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "%reload_ext autoreload\n%autoreload 2\n%matplotlib inline",
      "execution_count": 1,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "from fastai.vision import *",
      "execution_count": 2,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "path = untar_data(URLs.MNIST)",
      "execution_count": 3,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "path.ls()",
      "execution_count": 4,
      "outputs": [
        {
          "data": {
            "text/plain": "[PosixPath('/home/jupyter/.fastai/data/mnist_png/testing'),\n PosixPath('/home/jupyter/.fastai/data/mnist_png/training'),\n PosixPath('/home/jupyter/.fastai/data/mnist_png/models')]"
          },
          "output_type": "execute_result",
          "metadata": {},
          "execution_count": 4
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "il = ImageList.from_folder(path, convert_mode='L')",
      "execution_count": 5,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "defaults.cmap='binary'",
      "execution_count": 6,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "il[0].show()",
      "execution_count": 7,
      "outputs": [
        {
          "data": {
            "text/plain": "<Figure size 216x216 with 1 Axes>",
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAMUAAADDCAYAAAAyYdXtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAABHRJREFUeJzt3b9LlW0cx3F9FIKWhpxscgocGgNBJGgIIZpyKneHQPBvaHUR4Sw2CxI4tERTv4YzNvQHHJRARBAHB5Go8Xn4nO+B+/QcT0fP6zV+uTleQ28uvLq8z+SvX78mgH/987cXAKNGFBBEAUEUEEQBQRQQRAFhesg/z3+KMEomq6GdAoIoIIgCgiggiAKCKCCIAoIoIIgCgiggiAKCKCCIAoIoIIgCgiggiAKCKCCIAoIoIIgCwrDf5sGIe/v2bTlfWVkp52tra+W81WoNbE3DZqeAIAoIooAgCgiigDA55O+88y7ZEbK3t9c1e/HiRfnsz58/y/ns7Gw5//LlSzmfm5truLqh8C5ZaEIUEEQBQRQQRAHB6dMYODk5Kef379/vmp2enpbPPnjwoJxvbm6W88ePHzdc3V/l9AmaEAUEUUAQBQRRQPCXd2Nge3u7nFcnTdPT9T+J169fl/NrcsrUFzsFBFFAEAUEUUAQBQR3n26Qjx8/lvPl5eVyfnFx0TV7+PBh+Wy73f7jdY0wd5+gCVFAEAUEUUAQBQR3n26QT58+lfPqlKmXp0+fDmo515adAoIoIIgCgiggiAKCu0/XUKfTKeePHj0q5wcHB+V8amqqa3Z2dlY+e/v27UZru2bcfYImRAFBFBBEAcE1j2tofX29nPf6hbqX/f39rtkN/YW6L3YKCKKAIAoIooAgCghOn0bYjx8/yvm3b9/6+pw7d+6U84WFhb7XNA7sFBBEAUEUEEQBQRQQnD6NgF5f6L66ulrODw8P+/r8V69elfO7d+/29Tnjwk4BQRQQRAFBFBBEAcErbkbA0dFROZ+dne3rcxYXF8v5+/fvy7m/svOKG2hEFBBEAUEUEEQBwd2nEbCzszOQz1laWirnTpn6Y6eAIAoIooAgCgiigODu0xB9/fq1nC8vL5fz8/Pzcn7v3r1y/vnz53I+NzfXYHVjyd0naEIUEEQBQRQQRAHB3ach+vDhQznvdcpUffn7xMTERKvVKudOmQbDTgFBFBBEAUEUEFzzuCLv3r3rmj1//rx89vLyspzPz8+X8+/fv//5wvgv1zygCVFAEAUEUUAQBQTXPK5I9ZVavU6Zenn27NmglkMf7BQQRAFBFBBEAUEUEJw+jbB2u13OT05OyvnMzMxVLmds2CkgiAKCKCCIAoIoIDh9+p+2trbK+fHxcePPuHXrVjnvdZrU6XT6ep7+2CkgiAKCKCCIAoIoIHjvU0O9vjrryZMn5fzi4qJr1uuFybu7u+W813uiGBjvfYImRAFBFBBEAUEUENx9amhhYaGcv3z5spy/efOma7axsVE+65RptNgpIIgCgiggiAKCKCC4+8Q4c/cJmhAFBFFAEAUEUUAQBQRRQBAFBFFAEAUEUUAQBQRRQBAFBFFAEAUEUUAQBQRRQBAFBFFAEAWEYb9guXylCIwSOwUEUUAQBQRRQBAFBFFAEAUEUUAQBQRRQBAFBFFAEAUEUUAQBQRRQBAFBFFAEAUEUUAQBQRRQBAFhN98HLeB0FHJmgAAAABJRU5ErkJggg==\n"
          },
          "output_type": "display_data",
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "# split the data into training and validation sets\nsd = il.split_by_folder(train='training', valid='testing')",
      "execution_count": 8,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "sd",
      "execution_count": 9,
      "outputs": [
        {
          "data": {
            "text/plain": "ItemLists;\n\nTrain: ImageList (60000 items)\nImage (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28)\nPath: /home/jupyter/.fastai/data/mnist_png;\n\nValid: ImageList (10000 items)\nImage (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28)\nPath: /home/jupyter/.fastai/data/mnist_png;\n\nTest: None"
          },
          "output_type": "execute_result",
          "metadata": {},
          "execution_count": 9
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "(path/'training').ls()",
      "execution_count": 10,
      "outputs": [
        {
          "data": {
            "text/plain": "[PosixPath('/home/jupyter/.fastai/data/mnist_png/training/1'),\n PosixPath('/home/jupyter/.fastai/data/mnist_png/training/6'),\n PosixPath('/home/jupyter/.fastai/data/mnist_png/training/8'),\n PosixPath('/home/jupyter/.fastai/data/mnist_png/training/2'),\n PosixPath('/home/jupyter/.fastai/data/mnist_png/training/5'),\n PosixPath('/home/jupyter/.fastai/data/mnist_png/training/9'),\n PosixPath('/home/jupyter/.fastai/data/mnist_png/training/0'),\n PosixPath('/home/jupyter/.fastai/data/mnist_png/training/3'),\n PosixPath('/home/jupyter/.fastai/data/mnist_png/training/7'),\n PosixPath('/home/jupyter/.fastai/data/mnist_png/training/4')]"
          },
          "output_type": "execute_result",
          "metadata": {},
          "execution_count": 10
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "# label the data according to folders, as above\nll = sd.label_from_folder()",
      "execution_count": 11,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "ll",
      "execution_count": 12,
      "outputs": [
        {
          "data": {
            "text/plain": "LabelLists;\n\nTrain: LabelList (60000 items)\nx: ImageList\nImage (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28)\ny: CategoryList\n1,1,1,1,1\nPath: /home/jupyter/.fastai/data/mnist_png;\n\nValid: LabelList (10000 items)\nx: ImageList\nImage (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28),Image (1, 28, 28)\ny: CategoryList\n1,1,1,1,1\nPath: /home/jupyter/.fastai/data/mnist_png;\n\nTest: None"
          },
          "output_type": "execute_result",
          "metadata": {},
          "execution_count": 12
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "x,y = ll.train[0]  # input and output (label)",
      "execution_count": 13,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "x.show()\nprint(y,x.shape)",
      "execution_count": 14,
      "outputs": [
        {
          "name": "stdout",
          "text": "1 torch.Size([1, 28, 28])\n",
          "output_type": "stream"
        },
        {
          "data": {
            "text/plain": "<Figure size 216x216 with 1 Axes>",
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAMUAAADDCAYAAAAyYdXtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAABAtJREFUeJzt3e9JI0EAxuGNXhnWEdQ2RINWoZIqFBS78E8hfrALYxdBvY/HvZmR3Zisa/I8H4eFG5Cfw42zs6PPz88G+GfnpycAQyMKCKKAIAoIooAgCgiigPCn53/PH0UYklFp0EoBQRQQRAFBFBBEAUEUEEQBQRQQRAFBFBBEAUEUEEQBQRQQRAFBFBBEAUEUEEQBQRQQRAGh79s8GJDZbLYwNplMis8+Pz8Xx6fTaXH8+vp6+Yn9MCsFBFFAEAUEUUAQBYRRz9+8c5fsgBweHi6Mvby8FJ99f38vju/u7hbH5/P58hPrj7tkoQ1RQBAFBFFAEAUEZ5+2QO3cUml8NCpuyDR7e3vF8cfHx+UnNlBWCgiigCAKCKKAIAoIdp82SG2X6fT0tDhe2mmqnWUqnZNqmqbZ399vObvfw0oBQRQQRAFBFBBEAcHuU0u1nZ2ag4ODNc2k7u3trTj++vpaHC+9dVl7w248Hi8/sV/GSgFBFBBEAUEUEEQBwe5TS3d3d52e/4ndp5raeabSTlPt2cvLy5XOacisFBBEAUEUEEQBQRQQ7D61dH9/Xxy/ubnpeSb1c1gnJyfF8S53OW3iPU5dWSkgiAKCKCCIAoLPew1Y1ytrai8T1Y5uHB0dLYzVNhQ2lM97QRuigCAKCKKAIAoIjnkM2CqurGma+rU1Pe88/hpWCgiigCAKCKKAIAoIdp8G7Pb2tjje5cqar56vvXy07awUEEQBQRQQRAFBFBC8eTcAtTfsah90r+0a1X6Wtec/Pj5azG6jefMO2hAFBFFAEAUEUUBw9mkAap8Oq+0adT37NJ1Ol5vYlrJSQBAFBFFAEAUEUUBw9qlHT09PxfHJZFIc73qW6fj4uDj+8PDQYnZbydknaEMUEEQBQRQQRAHB2acBcI/TsFgpIIgCgiggiAKC/2j3qHaVTdfPb9WeH4/Hy02M/1gpIIgCgiggiAKCKCB4yahHOzvl30GrOuYxn8+Xm9j28pIRtCEKCKKAIAoIooDg7NM3zWaz4njp2pquZ5lqz5+fnxfHa1fo1NSuxNl2VgoIooAgCgiigCAKCM4+fVNtx+fs7GxhrOtZplU9X/u819XVVXF8izj7BG2IAoIoIIgCgiggOPu0JqWdoFWdfao9XzvLZJepGysFBFFAEAUEUUAQBQS7T2tSOp+07rNPFxcXLWfHV6wUEEQBQRQQRAFBFBC8ecc28+YdtCEKCKKAIAoIooAgCgiigCAKCKKAIAoIooAgCgiigCAKCKKAIAoIooAgCgiigCAKCKKAIAoIfV+wXLxSBIbESgFBFBBEAUEUEEQBQRQQRAFBFBBEAUEUEEQBQRQQRAFBFBBEAUEUEEQBQRQQRAFBFBBEAUEUEEQB4S+1C9//ZZvC9gAAAABJRU5ErkJggg==\n"
          },
          "output_type": "display_data",
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "tfms = ([*rand_pad(padding=3, size=28, mode='zeros')], [])  # only minimal tfms",
      "execution_count": 15,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "ll = ll.transform(tfms)  # append the tfms",
      "execution_count": 16,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "bs = 128",
      "execution_count": 17,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "data = ll.databunch(bs=bs).normalize()  # normalize accomplishes the mean/std normalization mentioned in the paper",
      "execution_count": 18,
      "outputs": []
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "Data augmentation occurs on-the-fly when the object itself is called; thus, every time the `data.train_ds[0][0]` object below is called, it will produce a new version with tfms applied. `plot_multi` will then call that object multiple times, showing our tfms. "
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "def _plot(i,j,ax): data.train_ds[0][0].show(ax, cmap='gray')  # i is row, j is col, ax is the figsize\nplot_multi(_plot, 3, 3, figsize=(8,8))",
      "execution_count": 19,
      "outputs": [
        {
          "data": {
            "text/plain": "<Figure size 576x576 with 9 Axes>",
            "image/png": "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\n"
          },
          "output_type": "display_data",
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "data.show_batch(rows=3, figsize=(5,5))",
      "execution_count": 24,
      "outputs": [
        {
          "data": {
            "text/plain": "<Figure size 360x360 with 9 Axes>",
            "image/png": "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\n"
          },
          "output_type": "display_data",
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "---"
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## Implementing WideResNet"
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "https://github.com/szagoruyko/wide-residual-networks/blob/master/models/wide-resnet.lua\n\nThere are two different WRNs from the paper that seem to offer the best performance and speed: WRN-28-10 and WRN-50-2-bottleneck, the former on CIFAR and the latter on ImageNet. For our purposes, let's start with WRN-28-10. \n\nThe experimenters point out that having 3x3 kernels throughout the network as well as having only two convolutional layers in each ResBlock results in the best performance. Further, the researchers indicate that placing a dropout layer between each of the convolutional layers in a ResBlock results in better performance. We will build our network using this information. \n\nAbout the WRN-$n$-$k$ notation: $n$ is the number of convolutions occurring in our network in total. If we have a WRN-28-10, what exactly does $n = 28$ total layers mean? Let's count our convolutions: A ResBlock has 2 convolutions occurring inside it. Remember from the paper that there are three \"conv groups\" that follow the initial convolution, each with a count of $N$ ResBlocks. If $N = 4$, each conv group will have 4 ResBlocks, or 4 x 2 convolutions, for a total of 8 per group. That gives us 24 convolutions; where are the other 4? At this point, I can only surmise from the authors' code that they are counting the \"identity\" convolutions in each conv group as 1 convolution per group (even though, technically, each ResBlock has its own identity convolution occurring...), so 24 + 3 gives us 27 convolutions. Finally, we add our initial convolution (\"conv1\") to the total for a depth of $n = 28$. That's the most sense I can make of it. "
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "# WRN-28-10, 28 layers with widening factor 10\n# N is (depth - 4) / 6, so (28 - 4) / 6 = 4\n\ndef wresgroup(ni, nf, k=10, N=4, stride=1, dropout=0.):\n    layers = []\n    nfk = nf * k\n    for i in range(N):\n        if i == 0:\n            layers.append(WResBlock(ni, nfk, stride=stride, dropout=dropout))\n        else:\n            layers.append(WResBlock(nfk, nfk, dropout=dropout))\n    return layers\n\nclass WResBlock(nn.Module):\n    def __init__(self, ni, nfk, stride=1, dropout=0.):\n        super().__init__()\n        layers = [\n            nn.BatchNorm2d(ni),\n            nn.ReLU(inplace=True),\n            nn.Conv2d(ni, nfk, 3, stride=stride, padding=1, bias=False),\n            nn.BatchNorm2d(nfk),\n            nn.ReLU(inplace=True),\n            nn.Dropout(dropout),\n            nn.Conv2d(nfk, nfk, 3, stride=1, padding=1, bias=False)\n        ]\n        self.conv_block = nn.Sequential(*layers)\n        self.identity = nn.Sequential(nn.Conv2d(ni, nfk, 1, stride=stride, padding=0, bias=False))\n\n    def forward(self, x): return self.identity(x) + self.conv_block(x)",
      "execution_count": 20,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "model = nn.Sequential(\n    nn.Conv2d(1, 8, 3, padding=1, bias=False),   # conv1 -- we'll start with 8 features because the images are so small\n    *wresgroup(8, 8, dropout=0.3),               # conv2\n    *wresgroup(80, 16, stride=2, dropout=0.3),   # conv3 -- more features, half the grid size\n    *wresgroup(160, 32, stride=2, dropout=0.3),  # conv4\n    nn.BatchNorm2d(320),\n    nn.ReLU(inplace=True),\n    nn.AvgPool2d(7),                             # the output of the last wresgroup is 7x7\n    Flatten(),\n    nn.Linear(320, 10)\n)",
      "execution_count": 21,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "learn = Learner(data, model, loss_func = nn.CrossEntropyLoss(), metrics=accuracy)",
      "execution_count": 22,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "print(learn.summary())",
      "execution_count": 23,
      "outputs": [
        {
          "name": "stdout",
          "text": "======================================================================\nLayer (type)         Output Shape         Param #    Trainable \n======================================================================\nConv2d               [1, 8, 28, 28]       72         True      \n______________________________________________________________________\nBatchNorm2d          [1, 8, 28, 28]       16         True      \n______________________________________________________________________\nReLU                 [1, 8, 28, 28]       0          False     \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      5,760      True      \n______________________________________________________________________\nBatchNorm2d          [1, 80, 28, 28]      160        True      \n______________________________________________________________________\nReLU                 [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nDropout              [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      57,600     True      \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      640        True      \n______________________________________________________________________\nBatchNorm2d          [1, 80, 28, 28]      160        True      \n______________________________________________________________________\nReLU                 [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      57,600     True      \n______________________________________________________________________\nBatchNorm2d          [1, 80, 28, 28]      160        True      \n______________________________________________________________________\nReLU                 [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nDropout              [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      57,600     True      \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      6,400      True      \n______________________________________________________________________\nBatchNorm2d          [1, 80, 28, 28]      160        True      \n______________________________________________________________________\nReLU                 [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      57,600     True      \n______________________________________________________________________\nBatchNorm2d          [1, 80, 28, 28]      160        True      \n______________________________________________________________________\nReLU                 [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nDropout              [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      57,600     True      \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      6,400      True      \n______________________________________________________________________\nBatchNorm2d          [1, 80, 28, 28]      160        True      \n______________________________________________________________________\nReLU                 [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      57,600     True      \n______________________________________________________________________\nBatchNorm2d          [1, 80, 28, 28]      160        True      \n______________________________________________________________________\nReLU                 [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nDropout              [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      57,600     True      \n______________________________________________________________________\nConv2d               [1, 80, 28, 28]      6,400      True      \n______________________________________________________________________\nBatchNorm2d          [1, 80, 28, 28]      160        True      \n______________________________________________________________________\nReLU                 [1, 80, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     115,200    True      \n______________________________________________________________________\nBatchNorm2d          [1, 160, 14, 14]     320        True      \n______________________________________________________________________\nReLU                 [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nDropout              [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     230,400    True      \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     12,800     True      \n______________________________________________________________________\nBatchNorm2d          [1, 160, 14, 14]     320        True      \n______________________________________________________________________\nReLU                 [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     230,400    True      \n______________________________________________________________________\nBatchNorm2d          [1, 160, 14, 14]     320        True      \n______________________________________________________________________\nReLU                 [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nDropout              [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     230,400    True      \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     25,600     True      \n______________________________________________________________________\nBatchNorm2d          [1, 160, 14, 14]     320        True      \n______________________________________________________________________\nReLU                 [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     230,400    True      \n______________________________________________________________________\nBatchNorm2d          [1, 160, 14, 14]     320        True      \n______________________________________________________________________\nReLU                 [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nDropout              [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     230,400    True      \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     25,600     True      \n______________________________________________________________________\nBatchNorm2d          [1, 160, 14, 14]     320        True      \n______________________________________________________________________\nReLU                 [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     230,400    True      \n______________________________________________________________________\nBatchNorm2d          [1, 160, 14, 14]     320        True      \n______________________________________________________________________\nReLU                 [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nDropout              [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     230,400    True      \n______________________________________________________________________\nConv2d               [1, 160, 14, 14]     25,600     True      \n______________________________________________________________________\nBatchNorm2d          [1, 160, 14, 14]     320        True      \n______________________________________________________________________\nReLU                 [1, 160, 14, 14]     0          False     \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       460,800    True      \n______________________________________________________________________\nBatchNorm2d          [1, 320, 7, 7]       640        True      \n______________________________________________________________________\nReLU                 [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nDropout              [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       921,600    True      \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       51,200     True      \n______________________________________________________________________\nBatchNorm2d          [1, 320, 7, 7]       640        True      \n______________________________________________________________________\nReLU                 [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       921,600    True      \n______________________________________________________________________\nBatchNorm2d          [1, 320, 7, 7]       640        True      \n______________________________________________________________________\nReLU                 [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nDropout              [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       921,600    True      \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       102,400    True      \n______________________________________________________________________\nBatchNorm2d          [1, 320, 7, 7]       640        True      \n______________________________________________________________________\nReLU                 [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       921,600    True      \n______________________________________________________________________\nBatchNorm2d          [1, 320, 7, 7]       640        True      \n______________________________________________________________________\nReLU                 [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nDropout              [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       921,600    True      \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       102,400    True      \n______________________________________________________________________\nBatchNorm2d          [1, 320, 7, 7]       640        True      \n______________________________________________________________________\nReLU                 [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       921,600    True      \n______________________________________________________________________\nBatchNorm2d          [1, 320, 7, 7]       640        True      \n______________________________________________________________________\nReLU                 [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nDropout              [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       921,600    True      \n______________________________________________________________________\nConv2d               [1, 320, 7, 7]       102,400    True      \n______________________________________________________________________\nBatchNorm2d          [1, 320, 7, 7]       640        True      \n______________________________________________________________________\nReLU                 [1, 320, 7, 7]       0          False     \n______________________________________________________________________\nAvgPool2d            [1, 320, 1, 1]       0          False     \n______________________________________________________________________\nFlatten              [1, 320]             0          False     \n______________________________________________________________________\nLinear               [1, 10]              3,210      True      \n______________________________________________________________________\n\nTotal params: 9,529,058\nTotal trainable params: 9,529,058\nTotal non-trainable params: 0\n\n",
          "output_type": "stream"
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "learn.lr_find()\nlearn.recorder.plot()",
      "execution_count": 24,
      "outputs": [
        {
          "data": {
            "text/html": "",
            "text/plain": "<IPython.core.display.HTML object>"
          },
          "output_type": "display_data",
          "metadata": {}
        },
        {
          "name": "stdout",
          "text": "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n",
          "output_type": "stream"
        },
        {
          "data": {
            "text/plain": "<Figure size 432x288 with 1 Axes>",
            "image/png": "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\n"
          },
          "output_type": "display_data",
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "learn.fit_one_cycle(5, 5e-3)",
      "execution_count": 25,
      "outputs": [
        {
          "data": {
            "text/html": "Total time: 05:30 <p><table style='width:375px; margin-bottom:10px'>\n  <tr>\n    <th>epoch</th>\n    <th>train_loss</th>\n    <th>valid_loss</th>\n    <th>accuracy</th>\n    <th>time</th>\n  </tr>\n  <tr>\n    <th>0</th>\n    <th>0.145045</th>\n    <th>1.070395</th>\n    <th>0.736500</th>\n    <th>01:06</th>\n  </tr>\n  <tr>\n    <th>1</th>\n    <th>0.100319</th>\n    <th>0.155969</th>\n    <th>0.943800</th>\n    <th>01:06</th>\n  </tr>\n  <tr>\n    <th>2</th>\n    <th>0.061960</th>\n    <th>0.049387</th>\n    <th>0.982200</th>\n    <th>01:06</th>\n  </tr>\n  <tr>\n    <th>3</th>\n    <th>0.032360</th>\n    <th>0.024816</th>\n    <th>0.992700</th>\n    <th>01:06</th>\n  </tr>\n  <tr>\n    <th>4</th>\n    <th>0.016214</th>\n    <th>0.010952</th>\n    <th>0.995900</th>\n    <th>01:06</th>\n  </tr>\n</table>\n",
            "text/plain": "<IPython.core.display.HTML object>"
          },
          "output_type": "display_data",
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "terp = learn.interpret()\nterp.plot_top_losses(9, figsize=(7,7))",
      "execution_count": 26,
      "outputs": [
        {
          "data": {
            "text/plain": "<Figure size 504x504 with 9 Axes>",
            "image/png": "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\n"
          },
          "output_type": "display_data",
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "### Trying WRN-40-2"
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "9.5M parameters is probably overkill for MNIST. Let's build a more reasonable network instead:"
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "# WRN-40-2\nmodel = nn.Sequential(\n    nn.Conv2d(1, 8, 3, padding=1, bias=False),           # conv1\n    *wresgroup(8, 8, k=2, N=6, dropout=0.3),             # conv2\n    *wresgroup(16, 16, k=2, N=6, stride=2, dropout=0.3), # conv3\n    *wresgroup(32, 32, k=2, N=6, stride=2, dropout=0.3), # conv4\n    nn.BatchNorm2d(64),\n    nn.ReLU(inplace=True),\n    nn.AvgPool2d(7),                                     # the output of the last wresgroup is 7x7\n    Flatten(),\n    nn.Linear(64, 10)\n)",
      "execution_count": 27,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "learn = Learner(data, model, loss_func = nn.CrossEntropyLoss(), metrics=accuracy)",
      "execution_count": 28,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "print(learn.summary())",
      "execution_count": 29,
      "outputs": [
        {
          "name": "stdout",
          "text": "======================================================================\nLayer (type)         Output Shape         Param #    Trainable \n======================================================================\nConv2d               [1, 8, 28, 28]       72         True      \n______________________________________________________________________\nBatchNorm2d          [1, 8, 28, 28]       16         True      \n______________________________________________________________________\nReLU                 [1, 8, 28, 28]       0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      1,152      True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nDropout              [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      2,304      True      \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      128        True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      2,304      True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nDropout              [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      2,304      True      \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      256        True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      2,304      True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nDropout              [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      2,304      True      \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      256        True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      2,304      True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nDropout              [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      2,304      True      \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      256        True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      2,304      True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nDropout              [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      2,304      True      \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      256        True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      2,304      True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nDropout              [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      2,304      True      \n______________________________________________________________________\nConv2d               [1, 16, 28, 28]      256        True      \n______________________________________________________________________\nBatchNorm2d          [1, 16, 28, 28]      32         True      \n______________________________________________________________________\nReLU                 [1, 16, 28, 28]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      4,608      True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nDropout              [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      9,216      True      \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      512        True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      9,216      True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nDropout              [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      9,216      True      \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      1,024      True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      9,216      True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nDropout              [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      9,216      True      \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      1,024      True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      9,216      True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nDropout              [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      9,216      True      \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      1,024      True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      9,216      True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nDropout              [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      9,216      True      \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      1,024      True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      9,216      True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nDropout              [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      9,216      True      \n______________________________________________________________________\nConv2d               [1, 32, 14, 14]      1,024      True      \n______________________________________________________________________\nBatchNorm2d          [1, 32, 14, 14]      64         True      \n______________________________________________________________________\nReLU                 [1, 32, 14, 14]      0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        18,432     True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nDropout              [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        36,864     True      \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        2,048      True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        36,864     True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nDropout              [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        36,864     True      \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        4,096      True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        36,864     True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nDropout              [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        36,864     True      \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        4,096      True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        36,864     True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nDropout              [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        36,864     True      \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        4,096      True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        36,864     True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nDropout              [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        36,864     True      \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        4,096      True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        36,864     True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nDropout              [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        36,864     True      \n______________________________________________________________________\nConv2d               [1, 64, 7, 7]        4,096      True      \n______________________________________________________________________\nBatchNorm2d          [1, 64, 7, 7]        128        True      \n______________________________________________________________________\nReLU                 [1, 64, 7, 7]        0          False     \n______________________________________________________________________\nAvgPool2d            [1, 64, 1, 1]        0          False     \n______________________________________________________________________\nFlatten              [1, 64]              0          False     \n______________________________________________________________________\nLinear               [1, 10]              650        True      \n______________________________________________________________________\n\nTotal params: 589,410\nTotal trainable params: 589,410\nTotal non-trainable params: 0\n\n",
          "output_type": "stream"
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "learn.lr_find()\nlearn.recorder.plot()",
      "execution_count": 30,
      "outputs": [
        {
          "data": {
            "text/html": "",
            "text/plain": "<IPython.core.display.HTML object>"
          },
          "output_type": "display_data",
          "metadata": {}
        },
        {
          "name": "stdout",
          "text": "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n",
          "output_type": "stream"
        },
        {
          "data": {
            "text/plain": "<Figure size 432x288 with 1 Axes>",
            "image/png": "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\n"
          },
          "output_type": "display_data",
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "learn.fit_one_cycle(12, 1e-2)",
      "execution_count": 31,
      "outputs": [
        {
          "data": {
            "text/html": "Total time: 04:48 <p><table style='width:375px; margin-bottom:10px'>\n  <tr>\n    <th>epoch</th>\n    <th>train_loss</th>\n    <th>valid_loss</th>\n    <th>accuracy</th>\n    <th>time</th>\n  </tr>\n  <tr>\n    <th>0</th>\n    <th>0.233642</th>\n    <th>0.573115</th>\n    <th>0.823400</th>\n    <th>00:23</th>\n  </tr>\n  <tr>\n    <th>1</th>\n    <th>0.135311</th>\n    <th>0.286319</th>\n    <th>0.918000</th>\n    <th>00:24</th>\n  </tr>\n  <tr>\n    <th>2</th>\n    <th>0.115002</th>\n    <th>0.124288</th>\n    <th>0.963800</th>\n    <th>00:24</th>\n  </tr>\n  <tr>\n    <th>3</th>\n    <th>0.096996</th>\n    <th>0.064454</th>\n    <th>0.980400</th>\n    <th>00:23</th>\n  </tr>\n  <tr>\n    <th>4</th>\n    <th>0.078536</th>\n    <th>0.173576</th>\n    <th>0.953800</th>\n    <th>00:23</th>\n  </tr>\n  <tr>\n    <th>5</th>\n    <th>0.064685</th>\n    <th>0.165209</th>\n    <th>0.950100</th>\n    <th>00:24</th>\n  </tr>\n  <tr>\n    <th>6</th>\n    <th>0.050055</th>\n    <th>0.037562</th>\n    <th>0.987500</th>\n    <th>00:23</th>\n  </tr>\n  <tr>\n    <th>7</th>\n    <th>0.041066</th>\n    <th>0.027783</th>\n    <th>0.991500</th>\n    <th>00:23</th>\n  </tr>\n  <tr>\n    <th>8</th>\n    <th>0.029880</th>\n    <th>0.017404</th>\n    <th>0.994600</th>\n    <th>00:23</th>\n  </tr>\n  <tr>\n    <th>9</th>\n    <th>0.020542</th>\n    <th>0.014314</th>\n    <th>0.994400</th>\n    <th>00:24</th>\n  </tr>\n  <tr>\n    <th>10</th>\n    <th>0.015938</th>\n    <th>0.012311</th>\n    <th>0.996000</th>\n    <th>00:24</th>\n  </tr>\n  <tr>\n    <th>11</th>\n    <th>0.016676</th>\n    <th>0.011243</th>\n    <th>0.996300</th>\n    <th>00:23</th>\n  </tr>\n</table>\n",
            "text/plain": "<IPython.core.display.HTML object>"
          },
          "output_type": "display_data",
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "terp = learn.interpret()\nterp.plot_top_losses(9, figsize=(7,7))",
      "execution_count": 32,
      "outputs": [
        {
          "data": {
            "text/plain": "<Figure size 504x504 with 9 Axes>",
            "image/png": "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\n"
          },
          "output_type": "display_data",
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ],
  "metadata": {
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3",
      "language": "python"
    },
    "language_info": {
      "pygments_lexer": "ipython3",
      "version": "3.7.1",
      "nbconvert_exporter": "python",
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "name": "python",
      "mimetype": "text/x-python",
      "file_extension": ".py"
    },
    "gist": {
      "id": "",
      "data": {
        "description": "Implementing WideResNet from scratch using fast.ai - MNIST",
        "public": true
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 2
}