DRL_04_Deep_Learning_with_PyTorch.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "DRL_04_Deep_Learning_with_PyTorch.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XfVHUVdarSLc"
      },
      "source": [
        "DEEP REINFORCEMENT LEARNING EXPLAINED - 04\n",
        "# **Deep Learning with PyTorch**\n",
        "## First contact with PyTorch for beginners\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tz6Mq31Y7Da6"
      },
      "source": [
        "1. Import required libraries"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MEXURnHMSamQ"
      },
      "source": [
        "# https://github.com/pytorch/vision/issues/1938#issuecomment-789986996\n",
        "from six.moves import urllib\n",
        "opener = urllib.request.build_opener()\n",
        "opener.addheaders = [('User-agent', 'Mozilla/5.0')]\n",
        "urllib.request.install_opener(opener)"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ymKullWooWG_"
      },
      "source": [
        "import torch \n",
        "import torchvision"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JqshwflVodU9"
      },
      "source": [
        "import numpy as np \n",
        "import matplotlib.pyplot as plt \n",
        "EPOCHS = 10 \n",
        "BATCH_SIZE= 64"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6BXWpS7Z7PtV"
      },
      "source": [
        "2. Load and Preprocessing Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SOTtMGEpofuO"
      },
      "source": [
        "xy_trainPT = torchvision.datasets.MNIST(root='./data', train=True, download=True,transform=torchvision.transforms.Compose([torchvision.transforms.ToTensor()])) \n",
        "xy_trainPT_loader = torch.utils.data.DataLoader(xy_trainPT, batch_size=BATCH_SIZE)"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FrpgRE-Ioh20",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 280
        },
        "outputId": "d3d8f233-9094-44ff-e289-a67c26655b66"
      },
      "source": [
        "fig = plt.figure(figsize=(25, 4)) \n",
        "for idx in np.arange(20):\n",
        "   image, label = xy_trainPT [idx]\n",
        "   ax = fig.add_subplot(2, 20/2, idx+1, xticks=[], yticks=[])\n",
        "   ax.imshow(torch.squeeze(image, dim = 0).numpy(), cmap=plt.cm.binary)\n",
        "   ax.set_title(str(label))"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1800x288 with 20 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Rdj9Akq8o2Kc",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4c99e0ca-88b0-4368-842f-cdf4954cf348"
      },
      "source": [
        "image, _ = xy_trainPT [0] \n",
        "print(image.size())\n",
        "image_flatten = image.view(image.shape[0], -1)\n",
        "print (image_flatten.size())"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([1, 28, 28])\n",
            "torch.Size([1, 784])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4ZDe0fSy9Sih"
      },
      "source": [
        "Tensors"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6Y-xgaNZo2bL",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "90c24471-d77d-4226-88d0-3c7bd3ab88ab"
      },
      "source": [
        "a = torch.randn(2, 3)\n",
        "print(a)"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tensor([[-0.3776, -1.1870,  0.1614],\n",
            "        [-1.4291,  0.0740, -1.5337]])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OO8AMsZlo2d7",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "bc2b1e4f-280d-4882-83a2-87b38cf3d7a4"
      },
      "source": [
        "print(a.size())\n",
        "print(a.dim())"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([2, 3])\n",
            "2\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_D0yCIlRo2hD",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "deb9d2e6-28de-4b60-cf73-05a27864a64d"
      },
      "source": [
        "matrix=torch.zeros([2, 4], dtype=torch.int32)\n",
        "print(matrix)"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tensor([[0, 0, 0, 0],\n",
            "        [0, 0, 0, 0]], dtype=torch.int32)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "k3sGanzjpGdd",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b2eee27d-4322-40ef-cb3f-95704300d644"
      },
      "source": [
        "print(matrix.dtype)"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.int32\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "It7j4KlGpGgX",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "18575675-54ea-4f81-ace6-e643b2639463"
      },
      "source": [
        "b = torch.zeros(2, 3)\n",
        "print(b)"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tensor([[0., 0., 0.],\n",
            "        [0., 0., 0.]])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Jhr3PFE7pGi6",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f4ca654c-ef81-4bb1-8f64-0122ca747b5d"
      },
      "source": [
        "c = torch.ones(2, 3)\n",
        "print(c)"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tensor([[1., 1., 1.],\n",
            "        [1., 1., 1.]])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qdAX3gpJpSSB",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a2ea951a-4df5-46d3-f262-3a8bf87e24e6"
      },
      "source": [
        "c[0,0]=222 \n",
        "print(c)"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tensor([[222.,   1.,   1.],\n",
            "        [  1.,   1.,   1.]])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hLv6NXRLpirf",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "93334e3f-af83-4dc5-dceb-e002d461d8b4"
      },
      "source": [
        "x = torch.Tensor([[1,2,3,4], [5,6,7,8], [9,10,11,12]]) \n",
        "print (x)\n",
        "print (\"x column 1: \", x[:, 1])\n",
        "print (\"x row 0: \", x[0, :])\n",
        "print (\"x rows 0,1 & cols 1,2: \\n\", x[0:2, 1:3])"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tensor([[ 1.,  2.,  3.,  4.],\n",
            "        [ 5.,  6.,  7.,  8.],\n",
            "        [ 9., 10., 11., 12.]])\n",
            "x column 1:  tensor([ 2.,  6., 10.])\n",
            "x row 0:  tensor([1., 2., 3., 4.])\n",
            "x rows 0,1 & cols 1,2: \n",
            " tensor([[2., 3.],\n",
            "        [6., 7.]])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y1L2LCappldj",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7aa4d4f2-5aba-49f0-e289-f1d28145e11f"
      },
      "source": [
        "x = np.array([[1,2], [3,4], [5,6]])\n",
        "print (x)"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[1 2]\n",
            " [3 4]\n",
            " [5 6]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gDWjCTAnpom_",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1a80a020-d846-413b-8010-fa06576df687"
      },
      "source": [
        "y=torch.from_numpy(x)\n",
        "print(y)"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tensor([[1, 2],\n",
            "        [3, 4],\n",
            "        [5, 6]])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YZzpIZLWpoph",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f5fda203-950f-4a55-8399-2b6ea359c705"
      },
      "source": [
        "z = y.numpy()\n",
        "print (z)"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[1 2]\n",
            " [3 4]\n",
            " [5 6]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ou_uj-prposE",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "47c3a7f4-6376-4076-9e11-65fd6bc388a7"
      },
      "source": [
        "one_d = torch.arange(0,16)\n",
        "print (one_d)\n",
        "two_d= one_d.reshape(4,4)\n",
        "print (two_d)\n",
        "print(two_d.size())"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tensor([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15])\n",
            "tensor([[ 0,  1,  2,  3],\n",
            "        [ 4,  5,  6,  7],\n",
            "        [ 8,  9, 10, 11],\n",
            "        [12, 13, 14, 15]])\n",
            "torch.Size([4, 4])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "d-ZJNT9k-7VK"
      },
      "source": [
        "3. Define Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yBU6rHwXpoue"
      },
      "source": [
        "modelPT= torch.nn.Sequential(            \n",
        "         torch.nn.Linear(784,10),\n",
        "         torch.nn.Sigmoid(), \n",
        "         torch.nn.Linear(10,10), \n",
        "         torch.nn.LogSoftmax(dim=1) \n",
        "         )"
      ],
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QWCOGdcyAkZa"
      },
      "source": [
        "4. Define the Optimizer and the Loss function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AMBYGpDcpowy"
      },
      "source": [
        "criterion = torch.nn.NLLLoss()"
      ],
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XTcaKe3BpozZ"
      },
      "source": [
        " optimizer = torch.optim.SGD(modelPT.parameters(), lr=0.01)"
      ],
      "execution_count": 21,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "J6bw76PFJ0_b"
      },
      "source": [
        "5. Train the model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RrT9OLC6po2R",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e4f9b6d3-329b-4c1b-86cc-6ab15da0b005"
      },
      "source": [
        "for e in range(EPOCHS):\n",
        "    running_loss = 0\n",
        "    for images, labels in xy_trainPT_loader:\n",
        "        images = images.view(images.shape[0], -1)\n",
        "        output = modelPT(images)\n",
        "        loss = criterion(output, labels)\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "        optimizer.zero_grad()\n",
        "        running_loss += loss.item()\n",
        "    print(\"Epoch {} - Training loss: {}\".format(e, running_loss/len(xy_trainPT_loader)))"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 0 - Training loss: 2.2140613255470294\n",
            "Epoch 1 - Training loss: 1.9499023151296033\n",
            "Epoch 2 - Training loss: 1.651940277136211\n",
            "Epoch 3 - Training loss: 1.4045752183969087\n",
            "Epoch 4 - Training loss: 1.2070220914730894\n",
            "Epoch 5 - Training loss: 1.0513942113309018\n",
            "Epoch 6 - Training loss: 0.9299559522666403\n",
            "Epoch 7 - Training loss: 0.8349341942684483\n",
            "Epoch 8 - Training loss: 0.7600495802250498\n",
            "Epoch 9 - Training loss: 0.7004144790012445\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rl_6fnL23PSn"
      },
      "source": [
        "6. Evaluate and use the Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wRV71lahq6Yv",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9fe17804-8dfa-4db9-a675-7754b71661fa"
      },
      "source": [
        "xy_testPT = torchvision.datasets.MNIST(root='./data', train=False, download=True, \n",
        "            transform=torchvision.transforms.Compose([torchvision.transforms.ToTensor()]))\n",
        "xy_test_loaderPT = torch.utils.data.DataLoader(xy_testPT)\n",
        "correct_count, all_count = 0, 0\n",
        "for images,labels in xy_test_loaderPT:\n",
        "  for i in range(len(labels)):\n",
        "    img = images[i].view(1, 784)\n",
        "    logps = modelPT(img)\n",
        "    ps = torch.exp(logps)\n",
        "    probab = list(ps.detach().numpy()[0])\n",
        "    pred_label = probab.index(max(probab))\n",
        "    true_label = labels.numpy()[i]\n",
        "    if(true_label == pred_label):\n",
        "      correct_count += 1\n",
        "    all_count += 1\n",
        "print(\"\\nAccuracy of the model =\",(correct_count/all_count))         "
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "Accuracy of the model = 0.8594\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}