colab-notebooks/15_FFNetworksWithPyTorch.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "15_FFNetworksWithPyTorch.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/satyajitghana/PadhAI-Course/blob/master/15_FFNetworksWithPyTorch.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UuAb0vJNGYt4",
        "colab_type": "text"
      },
      "source": [
        "## Outline\n",
        "\n",
        "* Recap of data\n",
        "* Feedforward network with Pytorch tensors and autograd\n",
        "* Using Pytorch's NN -> Functional, Linear, Sequential & Pytorch's Optim\n",
        "* Moving things to CUDA"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zi3i20NTF2LY",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "import numpy as np\n",
        "import math\n",
        "import matplotlib.pyplot as plt\n",
        "import matplotlib.colors\n",
        "import pandas as pd\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import accuracy_score, mean_squared_error, log_loss\n",
        "from tqdm import tqdm_notebook \n",
        "import seaborn as sns\n",
        "import time\n",
        "from IPython.display import HTML\n",
        "import warnings\n",
        "warnings.filterwarnings('ignore')\n",
        "\n",
        "sns.set()\n",
        "\n",
        "from sklearn.preprocessing import OneHotEncoder\n",
        "from sklearn.datasets import make_blobs\n",
        "\n",
        "import torch"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "S-2v19fdGBpV",
        "colab_type": "code",
        "outputId": "aa425242-cb00-4378-a40a-b33d3913a666",
        "cellView": "both",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "torch.manual_seed(0)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<torch._C.Generator at 0x7f82c2b726f0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 34
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KaNnf16vGFC8",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "my_cmap = 'nipy_spectral'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "t3h91SrKGem4",
        "colab_type": "text"
      },
      "source": [
        "## Generate Dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7CiU06B-GG-6",
        "colab_type": "code",
        "outputId": "e83ca0e4-d23e-40f0-8149-4bcc1b71af81",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "data, labels = make_blobs(n_samples=1000, centers=4, n_features=2, random_state=0)\n",
        "print(data.shape, labels.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(1000, 2) (1000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XOOGMkBAGIMy",
        "colab_type": "code",
        "outputId": "df5b8aa4-a744-4d8e-f82a-56a28aa1f40f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 268
        }
      },
      "source": [
        "plt.scatter(data[:,0], data[:,1], c=labels, cmap=my_cmap)\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXYAAAD7CAYAAAB+B7/XAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Eqblj3GsGMx6",
        "colab_type": "code",
        "outputId": "1b0a87e5-93f1-40b4-b01b-216f4d5522fa",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "X_train, X_val, Y_train, Y_val = train_test_split(data, labels, stratify=labels, random_state=0)\n",
        "print(X_train.shape, X_val.shape, labels.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(750, 2) (250, 2) (1000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LnPj_IrhGiYe",
        "colab_type": "text"
      },
      "source": [
        "## Using torch tensors and autograd"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "x3quRDoBOdvE",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "X_train, Y_train, X_val, Y_val = map(torch.tensor, (X_train, Y_train, X_val, Y_val))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "S8gNaribO9aB",
        "colab_type": "code",
        "outputId": "be25dbac-f2f6-4e72-9294-bd33747b791b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "print(X_train.shape, Y_train.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([750, 2]) torch.Size([750])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "h7jI7WUax89R",
        "colab_type": "code",
        "outputId": "a4444db3-8b54-4a91-bc78-a3c8a2879959",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 175
        }
      },
      "source": [
        "a = torch.rand([2, 4])\n",
        "print(a)\n",
        "print(a.exp())\n",
        "print(a.exp().sum())\n",
        "print(a.exp().sum(0))\n",
        "print(a.exp().sum(-1))\n",
        "print(a.exp().sum(-1).unsqueeze(-1))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tensor([[0.4698, 0.8409, 0.6222, 0.5540],\n",
            "        [0.9910, 0.2100, 0.1434, 0.6268]])\n",
            "tensor([[1.5997, 2.3185, 1.8629, 1.7402],\n",
            "        [2.6938, 1.2337, 1.1542, 1.8716]])\n",
            "tensor(14.4747)\n",
            "tensor([4.2935, 3.5522, 3.0172, 3.6118])\n",
            "tensor([7.5214, 6.9533])\n",
            "tensor([[7.5214],\n",
            "        [6.9533]])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Z57ivURJOqB-",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def model(x):\n",
        "  a1 = torch.matmul(x, weights1) + bias1 # (N, 2) x (2, 2) -> (N, 2)\n",
        "  h1 = a1.sigmoid() # (N, 2)\n",
        "  a2 = torch.matmul(h1, weights2) + bias2 # (N, 2) x (2, 4) -> (N, 4)\n",
        "  h2 = a2.exp()/a2.exp().sum(-1).unsqueeze(-1) # (N, 4)\n",
        "  return h2"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XWuBSu0vROfQ",
        "colab_type": "code",
        "outputId": "e78d7c19-34c9-48a6-bbc5-02c57f38d0b9",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "y_hat = torch.tensor([[0.1, 0.2, 0.3, 0.4], [0.8, 0.1, 0.05, 0.05]])\n",
        "y = torch.tensor([2, 0])\n",
        "\n",
        "(-y_hat[range(y_hat.shape[0]), y].log()).mean().item()\n",
        "\n",
        "(torch.argmax(y_hat, dim=1) == y).float().mean().item()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.5"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 42
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K0DJId7vR7aJ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def loss_fn(y_hat, y):\n",
        "  return -(y_hat[range(y.shape[0]), y].log()).mean()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vZa-xupaa2i3",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def accuracy(y_hat, y):\n",
        "  pred = torch.argmax(y_hat, dim=1)\n",
        "  return (pred == y).float().mean()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eGNj27K0YcO_",
        "colab_type": "code",
        "outputId": "36ade679-393e-4c52-872d-c721e9b3c6a2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 306
        }
      },
      "source": [
        "torch.manual_seed(0)\n",
        "weights1 = torch.randn(2, 2) / math.sqrt(2)\n",
        "weights1.requires_grad_()\n",
        "bias1 = torch.zeros(2, requires_grad=True)\n",
        "\n",
        "weights2 = torch.randn(2, 4) / math.sqrt(2)\n",
        "weights2.requires_grad_()\n",
        "bias2 = torch.zeros(4, requires_grad=True)\n",
        "\n",
        "learning_rate = 0.2\n",
        "epochs = 10000\n",
        "\n",
        "X_train = X_train.float()\n",
        "Y_train = Y_train.long()\n",
        "\n",
        "loss_arr = []\n",
        "acc_arr = []\n",
        "\n",
        "for epoch in range(epochs):\n",
        "  y_hat = model(X_train)\n",
        "  loss = loss_fn(y_hat, Y_train)\n",
        "  loss.backward()\n",
        "  loss_arr.append(loss.item())\n",
        "  acc_arr.append(accuracy(y_hat, Y_train))\n",
        "\n",
        "  with torch.no_grad():\n",
        "    weights1 -= weights1.grad * learning_rate\n",
        "    bias1 -= bias1.grad * learning_rate\n",
        "    weights2 -= weights2.grad * learning_rate\n",
        "    bias2 -= bias2.grad * learning_rate\n",
        "    weights1.grad.zero_()\n",
        "    bias1.grad.zero_()\n",
        "    weights2.grad.zero_()\n",
        "    bias2.grad.zero_()\n",
        "\n",
        "plt.plot(loss_arr, 'r-')\n",
        "plt.plot(acc_arr, 'b-')\n",
        "plt.show()\n",
        "print('Loss before training', loss_arr[0])\n",
        "print('Loss after training', loss_arr[-1])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAD9CAYAAACyYrxEAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Loss before training 1.5456441640853882\n",
            "Loss after training 0.19288592040538788\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1fYBscrQJvyr",
        "colab_type": "text"
      },
      "source": [
        "## Using NN.Functional"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PH5t0Tvg2MVq",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import torch.nn.functional as F"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "R7CMk4w52S5x",
        "colab_type": "code",
        "outputId": "8e9fe720-8d35-4b1f-c469-df133945ad2c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 303
        }
      },
      "source": [
        "torch.manual_seed(0)\n",
        "weights1 = torch.randn(2, 2) / math.sqrt(2)\n",
        "weights1.requires_grad_()\n",
        "bias1 = torch.zeros(2, requires_grad=True)\n",
        "\n",
        "weights2 = torch.randn(2, 4) / math.sqrt(2)\n",
        "weights2.requires_grad_()\n",
        "bias2 = torch.zeros(4, requires_grad=True)\n",
        "\n",
        "learning_rate = 0.2\n",
        "epochs = 10000\n",
        "\n",
        "loss_arr = []\n",
        "acc_arr = []\n",
        "\n",
        "for epoch in range(epochs):\n",
        "  y_hat = model(X_train)\n",
        "  loss = F.cross_entropy(y_hat, Y_train)\n",
        "  loss.backward()\n",
        "  loss_arr.append(loss.item())\n",
        "  acc_arr.append(accuracy(y_hat, Y_train))\n",
        "\n",
        "  with torch.no_grad():\n",
        "    weights1 -= weights1.grad * learning_rate\n",
        "    bias1 -= bias1.grad * learning_rate\n",
        "    weights2 -= weights2.grad * learning_rate\n",
        "    bias2 -= bias2.grad * learning_rate\n",
        "    weights1.grad.zero_()\n",
        "    bias1.grad.zero_()\n",
        "    weights2.grad.zero_()\n",
        "    bias2.grad.zero_()\n",
        "\n",
        "plt.plot(loss_arr, 'r-')\n",
        "plt.plot(acc_arr, 'b-')\n",
        "plt.show()\n",
        "print('Loss before training', loss_arr[0])\n",
        "print('Loss after training', loss_arr[-1])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAD7CAYAAABkO19ZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngJMCpUkifAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Loss before training 1.4111980199813843\n",
            "Loss after training 0.9561843872070312\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TeLqKpJMKNf_",
        "colab_type": "text"
      },
      "source": [
        "## Using NN.Parameter"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YTi--oeE9IMn",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import torch.nn as nn"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5-lao9OA3Buw",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class FirstNetwork(nn.Module):\n",
        "  \n",
        "  def __init__(self):\n",
        "    super().__init__()\n",
        "    torch.manual_seed(0)\n",
        "    self.weights1 = nn.Parameter(torch.randn(2, 2) / math.sqrt(2))\n",
        "    self.bias1 = nn.Parameter(torch.zeros(2))\n",
        "    self.weights2 = nn.Parameter(torch.randn(2, 4) / math.sqrt(2))\n",
        "    self.bias2 = nn.Parameter(torch.zeros(4))\n",
        "    \n",
        "  def forward(self, X):\n",
        "    a1 = torch.matmul(X, self.weights1) + self.bias1\n",
        "    h1 = a1.sigmoid()\n",
        "    a2 = torch.matmul(h1, self.weights2) + self.bias2\n",
        "    h2 = a2.exp()/a2.exp().sum(-1).unsqueeze(-1)\n",
        "    return h2"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K4L4tDXA6Fju",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def fit(epochs = 1000, learning_rate = 1):\n",
        "  loss_arr = []\n",
        "  acc_arr = []\n",
        "  for epoch in range(epochs):\n",
        "    y_hat = fn(X_train) # executes the forward function of the nn.Module\n",
        "    loss = F.cross_entropy(y_hat, Y_train)\n",
        "    loss_arr.append(loss.item())\n",
        "    acc_arr.append(accuracy(y_hat, Y_train))\n",
        "\n",
        "    loss.backward()\n",
        "    with torch.no_grad():\n",
        "      for param in fn.parameters():\n",
        "        param -= learning_rate * param.grad\n",
        "      fn.zero_grad()\n",
        "        \n",
        "  plt.plot(loss_arr, 'r-')\n",
        "  plt.plot(acc_arr, 'b-')\n",
        "  plt.show()      \n",
        "  print('Loss before training', loss_arr[0])\n",
        "  print('Loss after training', loss_arr[-1])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KkLjg3FS6V9c",
        "colab_type": "code",
        "outputId": "518a0ba1-fae1-4074-b786-b5cc8a6f6fd1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 303
        }
      },
      "source": [
        "fn = FirstNetwork()\n",
        "fit()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAD7CAYAAABgzo9kAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Loss before training 1.4111980199813843\n",
            "Loss after training 0.9939236044883728\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wKP_tS8cKSv9",
        "colab_type": "text"
      },
      "source": [
        "## Using NN.Linear and Optim"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zaS9b1OH9tBz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class FirstNetwork_v1(nn.Module):\n",
        "  \n",
        "  def __init__(self):\n",
        "    super().__init__()\n",
        "    torch.manual_seed(0)\n",
        "    self.lin1 = nn.Linear(2, 2)\n",
        "    self.lin2 = nn.Linear(2, 4)\n",
        "    \n",
        "  def forward(self, X):\n",
        "    a1 = self.lin1(X)\n",
        "    h1 = a1.sigmoid()\n",
        "    a2 = self.lin2(h1)\n",
        "    h2 = a2.exp()/a2.exp().sum(-1).unsqueeze(-1)\n",
        "    return h2"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CpIaKmU1CAmd",
        "colab_type": "code",
        "outputId": "dbd0fe97-2646-44f4-f8f2-21cb59f48485",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 303
        }
      },
      "source": [
        "fn = FirstNetwork_v1()\n",
        "fit()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAD7CAYAAABgzo9kAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Loss before training 1.395160436630249\n",
            "Loss after training 0.8738410472869873\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Bnj532qNC0By",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from torch import optim"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rffjj0tZCEC0",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def fit_v1(epochs = 1000, learning_rate = 1):\n",
        "  loss_arr = []\n",
        "  acc_arr = []\n",
        "  opt = optim.SGD(fn.parameters(), lr=learning_rate)\n",
        "  \n",
        "  for epoch in range(epochs):\n",
        "    y_hat = fn(X_train)\n",
        "    loss = F.cross_entropy(y_hat, Y_train)\n",
        "    loss_arr.append(loss.item())\n",
        "    acc_arr.append(accuracy(y_hat, Y_train))\n",
        "\n",
        "    loss.backward()\n",
        "    opt.step()\n",
        "    opt.zero_grad()\n",
        "        \n",
        "  plt.plot(loss_arr, 'r-')\n",
        "  plt.plot(acc_arr, 'b-')\n",
        "  plt.show()      \n",
        "  print('Loss before training', loss_arr[0])\n",
        "  print('Loss after training', loss_arr[-1])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FTiOuArHC9iD",
        "colab_type": "code",
        "outputId": "54fc88ec-19a4-4359-ab0e-91e6df6c038e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 303
        }
      },
      "source": [
        "fn = FirstNetwork_v1()\n",
        "fit_v1()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAD7CAYAAABgzo9kAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Loss before training 1.395160436630249\n",
            "Loss after training 0.8738410472869873\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VonDcC3FKbSR",
        "colab_type": "text"
      },
      "source": [
        "## Using NN.Sequential"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L-qr8a9cDdkm",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class FirstNetwork_v2(nn.Module):\n",
        "  \n",
        "  def __init__(self):\n",
        "    super().__init__()\n",
        "    torch.manual_seed(0)\n",
        "    self.net = nn.Sequential(\n",
        "        nn.Linear(2, 2), \n",
        "        nn.Sigmoid(), \n",
        "        nn.Linear(2, 4), \n",
        "        nn.Softmax()\n",
        "    )\n",
        "\n",
        "  def forward(self, X):\n",
        "    return self.net(X)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xohfM0eSDdei",
        "colab_type": "code",
        "outputId": "4d3693c7-b6a9-4c98-aed3-b57a0288a2ff",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 303
        }
      },
      "source": [
        "fn = FirstNetwork_v2()\n",
        "fit_v1()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAD7CAYAAABgzo9kAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Loss before training 1.395160436630249\n",
            "Loss after training 0.8738410472869873\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-Qts_kaoD4un",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def fit_v2(x, y, model, opt, loss_fn, epochs = 1000):\n",
        "  \n",
        "  for epoch in range(epochs):\n",
        "    loss = loss_fn(model(x), y)\n",
        "\n",
        "    loss.backward()\n",
        "    opt.step()\n",
        "    opt.zero_grad()\n",
        "    \n",
        "  return loss.item()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J6KfpPZKiNaC",
        "colab_type": "code",
        "outputId": "782d9ba1-4a97-4783-eb21-1a09755b1d90",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "fn = FirstNetwork_v2()\n",
        "loss_fn = F.cross_entropy\n",
        "opt = optim.SGD(fn.parameters(), lr=1)\n",
        "fit_v2(X_train, Y_train, fn, opt, loss_fn)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8738410472869873"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 60
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cUTJSDfPKd27",
        "colab_type": "text"
      },
      "source": [
        "## Running it on GPUs"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1Vd-_xX9EId0",
        "colab_type": "code",
        "outputId": "4f6d6c15-581d-431b-8bf8-c290943a462c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 87
        }
      },
      "source": [
        "%%time\n",
        "device = torch.device(\"cuda\")\n",
        "\n",
        "X_train=X_train.to(device)\n",
        "Y_train=Y_train.to(device)\n",
        "fn = FirstNetwork_v2()\n",
        "fn.to(device)\n",
        "tic = time.time()\n",
        "print('Final loss', fit_v2(X_train, Y_train, fn, opt, loss_fn))\n",
        "toc = time.time()\n",
        "print('Time taken', toc - tic)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Final loss 1.3951594829559326\n",
            "Time taken 0.6423451900482178\n",
            "CPU times: user 606 ms, sys: 36.6 ms, total: 643 ms\n",
            "Wall time: 645 ms\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1qVL_qwyoAHQ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class FirstNetwork_v3(nn.Module):\n",
        "  \n",
        "  def __init__(self):\n",
        "    super().__init__()\n",
        "    torch.manual_seed(0)\n",
        "    self.net = nn.Sequential(\n",
        "        nn.Linear(2, 1024*4), \n",
        "        nn.Sigmoid(), \n",
        "        nn.Linear(1024*4, 4), \n",
        "        nn.Softmax()\n",
        "    )\n",
        "\n",
        "  def forward(self, X):\n",
        "    return self.net(X)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "I92Jt55xodtb",
        "colab_type": "code",
        "outputId": "256e0bd5-32aa-411c-8d6a-ce8ee9a6f46c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 87
        }
      },
      "source": [
        "%%time\n",
        "device = torch.device(\"cpu\")\n",
        "\n",
        "X_train=X_train.to(device)\n",
        "Y_train=Y_train.to(device)\n",
        "fn = FirstNetwork_v3()\n",
        "fn.to(device)\n",
        "tic = time.time()\n",
        "print('Final loss', fit_v2(X_train, Y_train, fn, opt, loss_fn))\n",
        "toc = time.time()\n",
        "print('Time taken', toc - tic)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Final loss 1.3890225887298584\n",
            "Time taken 24.685086250305176\n",
            "CPU times: user 24.3 s, sys: 321 ms, total: 24.6 s\n",
            "Wall time: 24.7 s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CFsUIYtPsz6-",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}