ResNet9_Barlow_Twins.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "ResNet9_Barlow_Twins.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/hrishi508/Self-and-Semi-Supervised-Learning/blob/main/ResNet9_Barlow_Twins.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sofe8YH8BOEN"
      },
      "source": [
        "## Imports"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KyagcRRwMeo5"
      },
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "from torchvision import datasets\n",
        "from torch.autograd import Variable\n",
        "from torch.utils.data import Dataset\n",
        "from torch.utils.data import DataLoader\n",
        "from torchvision.transforms import ToTensor\n",
        "import torchvision.transforms as transforms\n",
        "# from torchviz import make_dot\n",
        "\n",
        "import time\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.manifold import TSNE\n",
        "from sklearn.decomposition import PCA\n",
        "from sklearn.preprocessing import StandardScaler"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DMyDn3HSLhq9",
        "outputId": "e2a04f5d-a549-4ed3-e4f7-d74f45ee7a05"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')\n",
        "\n",
        "%cd /content/drive/MyDrive"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n",
            "/content/drive/MyDrive\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
        "device"
      ],
      "metadata": {
        "id": "1_a4y6W_bmJ3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "876540d2-6466-42aa-8fad-ca2db23b1ee8"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "'cpu'"
            ]
          },
          "metadata": {},
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ty1XcIxvBShg"
      },
      "source": [
        "## Data Preparation\n",
        "\n",
        "Here, we are extracting 500 labelled images from the the training set for the self-supervised learning. Using labelled data sounds counter-intuitive but we are not using the labels in the training of the network, we are only using them for the TSNE visualization of the embedding vectors after every epoch."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zYLWrg3vOr87"
      },
      "source": [
        "transform = transforms.Compose([\n",
        "    transforms.ToTensor(),    \n",
        "    transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),   \n",
        "    \n",
        "])\n",
        "\n",
        "training_data = datasets.STL10(\n",
        "    root=\"data\",\n",
        "    split=\"train\",\n",
        "    transform=transform,\n",
        "    download=False\n",
        ")\n",
        "\n",
        "# test_data = datasets.STL10(\n",
        "#     root=\"data\",\n",
        "#     split='test',\n",
        "#     transform=transform,\n",
        "#     download=False\n",
        "# )\n",
        "\n",
        "# unlabelled = datasets.STL10(\n",
        "#     root=\"data\",\n",
        "#     split=\"unlabeled\",\n",
        "#     transform=transform,\n",
        "#     download=False\n",
        "# )"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G1Ma9i5_SvAa"
      },
      "source": [
        "labelled_500, _ = torch.utils.data.random_split(training_data, [500, 4500])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ae8z9kyaUNGu"
      },
      "source": [
        "## HYPERPARAMETERS\n",
        "BATCH_SIZE = 64  #64\n",
        "LR = 0.001      #0.001\n",
        "EPOCHS = 50\n",
        "\n",
        "train_dataloader = DataLoader(labelled_500, batch_size=BATCH_SIZE, shuffle=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mMsoyhwD-t5q"
      },
      "source": [
        "def imshow(imgt):\n",
        "    img = imgt.clone()\n",
        "    t1 = torch.tensor([0.485, 0.456, 0.406])\n",
        "    t2 = torch.tensor([0.229, 0.224, 0.225])\n",
        "    img[0]*=t2[0]\n",
        "    img[1]*=t2[1]\n",
        "    img[2]*=t2[2]\n",
        "\n",
        "    img[0]+=t1[0]\n",
        "    img[1]+=t1[1]\n",
        "    img[2]+=t1[2]\n",
        "\n",
        "    npimg = img.numpy()\n",
        "    plt.imshow(np.transpose(npimg, (1, 2, 0)))\n",
        "    plt.show()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 302
        },
        "id": "xs0BekaJRcdT",
        "outputId": "a72ce95a-3089-43dd-860f-a7a4a44e1b83"
      },
      "source": [
        "labels_map = {\n",
        "    0: \"airplane\",\n",
        "    1: \"bird\",\n",
        "    2: \"car\",\n",
        "    3: \"cat\",\n",
        "    4: \"deer\",\n",
        "    5: \"dog\",\n",
        "    6: \"horse\",\n",
        "    7: \"monkey\",\n",
        "    8: \"ship\",\n",
        "    9: \"truck\",\n",
        "}\n",
        "img, label = training_data[155]\n",
        "print(labels_map[label])\n",
        "print(img.shape)\n",
        "imshow(img)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "monkey\n",
            "torch.Size([3, 96, 96])\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TFrlZqG3UgIv"
      },
      "source": [
        "train_features, train_labels = next(iter(train_dataloader))\n",
        "# tf = train_features.unsqueeze(0)\n",
        "# print(tf.shape)\n",
        "# print(train_labels)\n",
        "img = train_features[0]\n",
        "imshow(img)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "X5jf24Lc3-d7"
      },
      "source": [
        "## Data Augmentation\n",
        "\n",
        "We have defined several augmentations, with the hyperparameters taken from the original work on Barlow Twins."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sSPZwO60Admy"
      },
      "source": [
        "def augment(x):\n",
        "    t = transforms.RandomApply([\n",
        "        transforms.RandomHorizontalFlip(0.5),\n",
        "        transforms.RandomRotation(15),\n",
        "        transforms.GaussianBlur(9, sigma=(0.1, 2.0)),\n",
        "        transforms.RandomSolarize(0.5, 0.2),\n",
        "        transforms.ColorJitter(0.4, 0.4, 0.2, 0.1) \n",
        "    ], p=0.7)\n",
        "\n",
        "    im1 = t(x)\n",
        "    im2 = t(x)\n",
        "\n",
        "    return im1, im2\n",
        "    # return x.clone(), x.clone()\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BpH-UbwsC8xg"
      },
      "source": [
        "a, b = augment(img)\n",
        "imshow(a)\n",
        "imshow(b)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2oQIBiBhMaGm"
      },
      "source": [
        "## Train function\n",
        "This function contains our implementaion of the training pseudocode as given in the paper on Barlow Twins."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "E2gLPsi9dhkJ"
      },
      "source": [
        "def loss_fn(C, lambd):\n",
        "    loss = 0\n",
        "    temp = (C - torch.eye(C.size(0), requires_grad=False).to(device)).pow(2)\n",
        "    temp2 = temp * ((torch.ones(C.shape, requires_grad=False)-torch.eye(C.size(0), requires_grad=False))*lambd).to(device)\n",
        "    loss = torch.sum(temp2)\n",
        "    return loss"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "O-ROd9ydAsg_"
      },
      "source": [
        "def train_model(model, dataloader, loss_func, optimizer, lambd, num_epochs=100):\n",
        "    since = time.time()\n",
        "    model = model.to(device)\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        print('Epoch {}/{}'.format(epoch, num_epochs - 1))\n",
        "        print('-' * 10)\n",
        "        \n",
        "        model.train()\n",
        "\n",
        "        running_loss = 0.0\n",
        "\n",
        "        for inputs, _ in dataloader:\n",
        "            inputs = inputs.to(device)\n",
        "\n",
        "            optimizer.zero_grad()\n",
        "\n",
        "            ix1, ix2 = augment(inputs)                     \n",
        "            x1 = torch.squeeze(model(ix1))\n",
        "            x2 = torch.squeeze(model(ix2))\n",
        "            # print(x1-x2)\n",
        "\n",
        "            \n",
        "            x1 = (x1 - torch.std_mean(x1, 0)[1])/(torch.std_mean(x1, 0)[0] + 1e-9)\n",
        "            x2 = (x2 - torch.std_mean(x2, 0)[1])/(torch.std_mean(x2, 0)[0] + 1e-9)\n",
        "            \n",
        "            C = torch.matmul(x1.T, x2)/len(inputs)\n",
        "            # print(C)\n",
        "            # print(C.shape)\n",
        "            \n",
        "            loss = loss_func(C, lambd)\n",
        "            # make_dot(loss, params=dict(list(model_ft.named_parameters()))).render(\"model_torchviz\", format=\"png\")\n",
        "            # break\n",
        "\n",
        "            loss.backward()\n",
        "            torch.nn.utils.clip_grad_norm_(model.parameters(), 1000)\n",
        "            optimizer.step()\n",
        "\n",
        "            running_loss += loss.item() * inputs.size(0)\n",
        "\n",
        "        epoch_loss = running_loss / len(dataloader.dataset)\n",
        "\n",
        "        print('Loss: {:.4f}'.format(epoch_loss))\n",
        "\n",
        "        print()\n",
        "        e = visualize_embeddings_500(model, dataloader, epoch, \"BOTH\", False)\n",
        "        print(\"Viz done!\")\n",
        "        print()\n",
        "\n",
        "    time_elapsed = time.time() - since\n",
        "    print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))\n",
        "\n",
        "    # return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Debugging code (ignore)"
      ],
      "metadata": {
        "id": "My0FClcb7wHB"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cLxBx3Y_jNtF"
      },
      "source": [
        "# a = torch.tensor([[3,3,3],[3,3,3]]) / torch.tensor([3,3,3])\n",
        "# a"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UskkAMWHMByP"
      },
      "source": [
        "# x1 = torch.ones((64,128))*30\n",
        "# print(x1)\n",
        "# print(torch.std_mean(x1, 0)[0])\n",
        "# x1 = (x1 - torch.std_mean(x1, 0)[1])/torch.std_mean(x1, 0)[0]\n",
        "# print(x1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zxAWDs_QlFF8"
      },
      "source": [
        "# def set_parameter_requires_grad(model, feature_extracting):\n",
        "#     if feature_extracting:\n",
        "#         for param in model.parameters():\n",
        "            # param.requires_grad = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yVLVLmgMcMEV"
      },
      "source": [
        "## Facebook train function (ignore)\n",
        "This function contains Facebook's [implementaion](https://github.com/facebookresearch/barlowtwins) of the training pseudocode as given in the paper on Barlow Twins."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "979MCenQWxij"
      },
      "source": [
        "def off_diagonal(x):\n",
        "    # return a flattened view of the off-diagonal elements of a square matrix\n",
        "    n, m = x.shape\n",
        "    assert n == m\n",
        "    return x.flatten()[:-1].view(n - 1, n + 1)[:, 1:].flatten()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SZTL9Wmn0tgE"
      },
      "source": [
        "def train_model_2(model, dataloader, optimizer, lambd, num_epochs=100):\n",
        "    since = time.time()\n",
        "    model = model.to(device)\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        print('Epoch {}/{}'.format(epoch, num_epochs - 1))\n",
        "        print('-' * 10)\n",
        "        \n",
        "        model.train()  # Set model to training mode\n",
        "\n",
        "        running_loss = 0.0\n",
        "\n",
        "            # Iterate over data.\n",
        "        for inputs, _ in dataloader:\n",
        "            inputs = inputs.to(device)\n",
        "\n",
        "            # zero the parameter gradients\n",
        "            optimizer.zero_grad()\n",
        "\n",
        "            ix1, ix2 = augment(inputs)                     \n",
        "            x1 = model(ix1)\n",
        "            x2 = model(ix2)\n",
        "            # print(x1.shape)\n",
        "            # print(x1-x2)\n",
        "            \n",
        "            m = nn.BatchNorm1d(x1.size(1), affine = False)\n",
        "            C = (m(x1).T @ m(x2))/len(inputs)\n",
        "            on_diag = torch.diagonal(C).add_(-1).pow_(2).sum()\n",
        "            off_diag = off_diagonal(C).pow_(2).sum()\n",
        "            loss = on_diag + lambd * off_diag\n",
        "            loss.backward()\n",
        "            optimizer.step()\n",
        "\n",
        "            # statistics\n",
        "            running_loss += loss.item() * inputs.size(0)\n",
        "\n",
        "        epoch_loss = running_loss / len(dataloader.dataset)\n",
        "\n",
        "        print('Loss: {:.4f}'.format(epoch_loss))\n",
        "\n",
        "        print()\n",
        "\n",
        "    time_elapsed = time.time() - since\n",
        "    print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))\n",
        "\n",
        "    return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### LARS Optimizer implemented by Facebook (only useful for very large batch sizes)\n",
        "Even a batch size of 256 was causing an out of memory error on colab, so we did not use this optimizer."
      ],
      "metadata": {
        "id": "dJ-qNZCD-HVz"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3oASYbsUgKyu"
      },
      "source": [
        "from torch import optim\n",
        "class LARS(optim.Optimizer):\n",
        "    def __init__(self, params, lr, weight_decay=0, momentum=0.9, eta=0.001,\n",
        "                 weight_decay_filter=False, lars_adaptation_filter=False):\n",
        "        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum,\n",
        "                        eta=eta, weight_decay_filter=weight_decay_filter,\n",
        "                        lars_adaptation_filter=lars_adaptation_filter)\n",
        "        super().__init__(params, defaults)\n",
        "\n",
        "\n",
        "    def exclude_bias_and_norm(self, p):\n",
        "        return p.ndim == 1\n",
        "\n",
        "    @torch.no_grad()\n",
        "    def step(self):\n",
        "        for g in self.param_groups:\n",
        "            for p in g['params']:\n",
        "                dp = p.grad\n",
        "\n",
        "                if dp is None:\n",
        "                    continue\n",
        "\n",
        "                if not g['weight_decay_filter'] or not self.exclude_bias_and_norm(p):\n",
        "                    dp = dp.add(p, alpha=g['weight_decay'])\n",
        "\n",
        "                if not g['lars_adaptation_filter'] or not self.exclude_bias_and_norm(p):\n",
        "                    param_norm = torch.norm(p)\n",
        "                    update_norm = torch.norm(dp)\n",
        "                    one = torch.ones_like(param_norm)\n",
        "                    q = torch.where(param_norm > 0.,\n",
        "                                    torch.where(update_norm > 0,\n",
        "                                                (g['eta'] * param_norm / update_norm), one), one)\n",
        "                    dp = dp.mul(q)\n",
        "\n",
        "                param_state = self.state[p]\n",
        "                if 'mu' not in param_state:\n",
        "                    param_state['mu'] = torch.zeros_like(p)\n",
        "                mu = param_state['mu']\n",
        "                mu.mul_(g['momentum']).add_(dp)\n",
        "\n",
        "                p.add_(mu, alpha=-g['lr'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5wjutBz2gr5R"
      },
      "source": [
        "optimizer = LARS(model_ft.parameters(), lr=0, weight_decay=1e-6,\n",
        "                     weight_decay_filter=True,\n",
        "                     lars_adaptation_filter=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## ResNet 9 Model"
      ],
      "metadata": {
        "id": "0KZPAtqxKxDl"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3C1oz5ASdxD0"
      },
      "source": [
        "def conv_block(in_channels, out_channels, pool=False):\n",
        "    layers = [\n",
        "                nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1), \n",
        "                nn.BatchNorm2d(out_channels), \n",
        "                nn.ReLU(inplace=True)\n",
        "              ]\n",
        "\n",
        "    if pool: \n",
        "      layers.append(nn.MaxPool2d(2))\n",
        "    \n",
        "    return nn.Sequential(*layers)\n",
        "\n",
        "class ResNet9(nn.Module):\n",
        "\n",
        "    def __init__(self, in_channels, num_classes):\n",
        "        super(ResNet9, self).__init__()\n",
        "        \n",
        "        self.conv1 = conv_block(in_channels, 64,pool=True)\n",
        "        self.conv2 = conv_block(64, 128, pool=True) \n",
        "        self.res1 = nn.Sequential(conv_block(128, 128), \n",
        "                                  conv_block(128, 128))\n",
        "        \n",
        "        self.conv3 = conv_block(128, 256, pool=True)\n",
        "        self.conv4 = conv_block(256, 512, pool=True)\n",
        "        self.res2 = nn.Sequential(conv_block(512, 512), \n",
        "                                  conv_block(512, 512))\n",
        "        \n",
        "        self.classifier = nn.Sequential(nn.MaxPool2d(6), \n",
        "                                        nn.Flatten(), \n",
        "                                        nn.Dropout(0.2),\n",
        "                                        nn.Linear(512, 128))\n",
        "        \n",
        "    def forward(self, xb):\n",
        "        out = self.conv1(xb)\n",
        "        out = self.conv2(out)\n",
        "        out = self.res1(out) + out\n",
        "        out = self.conv3(out)\n",
        "        out = self.conv4(out)\n",
        "        out = self.res2(out) + out\n",
        "        out = self.classifier(out)\n",
        "        return out"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sDWWLkYuKhia"
      },
      "source": [
        "model = ResNet9(3, 10)\n",
        "optimizer = torch.optim.Adam(params=model.parameters(), lr=LR)\n",
        "print(model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RVSQHAh_dM2N"
      },
      "source": [
        "## PCA function\n",
        "**visualize_embeddings_100**: This function generates 100 embedding vectors and gives their PCA or TSNE visualization as specified by the user.\n",
        "\n",
        "**visualize_embeddings_500**: This function generates 500 embedding vectors and gives their PCA or TSNE visualization as specified by the user."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PSZH21RCV5ll"
      },
      "source": [
        "def visualize_embeddings_100(model_ft, train_dataloader, epoch_number, method = \"TSNE\", visualize = True, N_components = 2, vector_size = 128, classes = 10):\n",
        "    cnt = 0\n",
        "    tmp = [0 for i in range(vector_size)]\n",
        "    x = [tmp for i in range(1000)]\n",
        "    vec_cnt = [0 for i in range(classes)]\n",
        "    i = 0\n",
        "    principalComponents = None\n",
        "\n",
        "    model_ft.eval()\n",
        "\n",
        "    while cnt < 100:\n",
        "\n",
        "        inputs, labels = next(iter(train_dataloader))\n",
        "        \n",
        "        for i in range(BATCH_SIZE):\n",
        "            if vec_cnt[labels[i]] == 10:\n",
        "                continue\n",
        "\n",
        "            if cnt == 100:\n",
        "                break\n",
        "            \n",
        "            input = inputs[i].unsqueeze(0)\n",
        "            input = input.to(device)\n",
        "            vec = torch.squeeze(model_ft(input))\n",
        "            \n",
        "            x[(labels[i]*10 + vec_cnt[labels[i]])] = vec.detach().numpy()\n",
        "            vec_cnt[labels[i]] += 1\n",
        "            cnt += 1\n",
        "\n",
        "\n",
        "    z = StandardScaler().fit_transform(x)\n",
        "\n",
        "    if method == \"TSNE\":\n",
        "        principalComponents = TSNE(n_components=N_components, learning_rate='auto', init='random').fit_transform(z)\n",
        "    \n",
        "    elif method == \"PCA\":\n",
        "        pca = PCA(n_components=N_components)\n",
        "        principalComponents = pca.fit_transform(z)\n",
        "\n",
        "    elif method == \"BOTH\":\n",
        "        principalComponents1 = TSNE(n_components=N_components, learning_rate='auto', init='random').fit_transform(z)\n",
        "\n",
        "        pca = PCA(n_components=N_components)\n",
        "        principalComponents2 = pca.fit_transform(z)\n",
        "\n",
        "        principalDf = pd.DataFrame(data = principalComponents1, columns = ['principal component 1', 'principal component 2'])\n",
        "        fig = plt.figure(figsize = (8,8))\n",
        "        ax = fig.add_subplot(1,1,1) \n",
        "\n",
        "        #plt.ylim([-30,30])\n",
        "\n",
        "        ax.set_xlabel('Principal Component 1', fontsize = 15)\n",
        "        ax.set_ylabel('Principal Component 2', fontsize = 15)\n",
        "        ax.set_title(method + ' Output', fontsize = 20)\n",
        "        colors = ['r', 'g', 'b','c','m','y','k','gold','grey','purple']\n",
        "        k=[0,2,6,8,9]\n",
        "        n = 10\n",
        "        for i in range(10):\n",
        "            ax.scatter(principalDf.loc[range((i*n),((i*n)+n)), 'principal component 1']\n",
        "                    , principalDf.loc[range((i*n),((i*n)+n)), 'principal component 2']\n",
        "                    , c = colors[i]\n",
        "                    , s = 50)\n",
        "\n",
        "        #ax.legend(targets)\n",
        "        plt.savefig('/content/drive/MyDrive/data/'+'TSNE_Output_Epoch_' + str(epoch_number) + '.png')\n",
        "        plt.close()\n",
        "\n",
        "        principalDf = pd.DataFrame(data = principalComponents2, columns = ['principal component 1', 'principal component 2'])\n",
        "        fig = plt.figure(figsize = (8,8))\n",
        "        ax = fig.add_subplot(1,1,1) \n",
        "\n",
        "        #plt.ylim([-30,30])\n",
        "\n",
        "        ax.set_xlabel('Principal Component 1', fontsize = 15)\n",
        "        ax.set_ylabel('Principal Component 2', fontsize = 15)\n",
        "        ax.set_title(method + ' Output', fontsize = 20)\n",
        "        colors = ['r', 'g', 'b','c','m','y','k','gold','grey','purple']\n",
        "        k=[0,2,6,8,9]\n",
        "        n = 10\n",
        "        for i in range(10):\n",
        "            ax.scatter(principalDf.loc[range((i*n),((i*n)+n)), 'principal component 1']\n",
        "                    , principalDf.loc[range((i*n),((i*n)+n)), 'principal component 2']\n",
        "                    , c = colors[i]\n",
        "                    , s = 50)\n",
        "\n",
        "        #ax.legend(targets)\n",
        "        plt.savefig('/content/drive/MyDrive/data/'+'PCA_Output_Epoch_' + str(epoch_number) + '.png')\n",
        "        plt.close()\n",
        "\n",
        "        return x\n",
        "\n",
        "        \n",
        "    principalDf = pd.DataFrame(data = principalComponents, columns = ['principal component 1', 'principal component 2'])\n",
        "    fig = plt.figure(figsize = (8,8))\n",
        "    ax = fig.add_subplot(1,1,1) \n",
        "\n",
        "    #plt.ylim([-30,30])\n",
        "\n",
        "    ax.set_xlabel('Principal Component 1', fontsize = 15)\n",
        "    ax.set_ylabel('Principal Component 2', fontsize = 15)\n",
        "    ax.set_title(method + ' Output', fontsize = 20)\n",
        "    colors = ['r', 'g', 'b','c','m','y','k','gold','grey','purple']\n",
        "    k=[0,2,6,8,9]\n",
        "    n = 10\n",
        "    for i in range(10):\n",
        "        ax.scatter(principalDf.loc[range((i*n),((i*n)+n)), 'principal component 1']\n",
        "                , principalDf.loc[range((i*n),((i*n)+n)), 'principal component 2']\n",
        "                , c = colors[i]\n",
        "                , s = 50)\n",
        "\n",
        "    #ax.legend(targets)\n",
        "    ax.grid()\n",
        "    plt.savefig('/content/drive/MyDrive/data/'+'TSNE_Output_Epoch_' + str(epoch_number) + '.png')\n",
        "    if visualize == False:\n",
        "        plt.close()\n",
        "\n",
        "    return x"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_0L-qnnYeKC-"
      },
      "source": [
        "def visualize_embeddings_500(model_ft, train_dataloader, epoch_number, method = \"TSNE\", visualize = True, N_components = 2, vector_size = 128, classes = 10):\n",
        "    cnt = 0\n",
        "    tmp = [0 for i in range(vector_size)]\n",
        "    x = [tmp for i in range(1000)]\n",
        "    vec_cnt = [0 for i in range(classes)]\n",
        "    i = 0\n",
        "    principalComponents = None\n",
        "\n",
        "    model_ft.eval()\n",
        "\n",
        "    while cnt < 500:\n",
        "\n",
        "        inputs, labels = next(iter(train_dataloader))\n",
        "        # print(labels)\n",
        "        \n",
        "        for i in range(BATCH_SIZE):\n",
        "\n",
        "            if cnt == 500:\n",
        "                break\n",
        "            \n",
        "            input = inputs[i].unsqueeze(0)\n",
        "            input = input.to(device)\n",
        "            vec = torch.squeeze(model_ft(input))\n",
        "            \n",
        "            # print((labels[i]*100 + vec_cnt[labels[i]]))\n",
        "            x[(labels[i]*100 + vec_cnt[labels[i]])] = vec.cpu().detach().numpy()\n",
        "            vec_cnt[labels[i]] += 1\n",
        "            cnt += 1\n",
        "            i += 1\n",
        "\n",
        "\n",
        "    z = StandardScaler().fit_transform(x)\n",
        "\n",
        "    if method == \"TSNE\":\n",
        "        principalComponents = TSNE(n_components=N_components, learning_rate='auto', init='random').fit_transform(z)\n",
        "    \n",
        "    elif method == \"PCA\":\n",
        "        pca = PCA(n_components=N_components)\n",
        "        principalComponents = pca.fit_transform(z)\n",
        "\n",
        "    elif method == \"BOTH\":\n",
        "        principalComponents1 = TSNE(n_components=N_components, learning_rate='auto', init='random').fit_transform(z)\n",
        "\n",
        "        pca = PCA(n_components=N_components)\n",
        "        principalComponents2 = pca.fit_transform(z)\n",
        "\n",
        "        principalDf = pd.DataFrame(data = principalComponents1, columns = ['principal component 1', 'principal component 2'])\n",
        "        fig = plt.figure(figsize = (8,8))\n",
        "        ax = fig.add_subplot(1,1,1) \n",
        "\n",
        "        #plt.ylim([-30,30])\n",
        "\n",
        "        ax.set_xlabel('Principal Component 1', fontsize = 15)\n",
        "        ax.set_ylabel('Principal Component 2', fontsize = 15)\n",
        "        ax.set_title('TSNE Output', fontsize = 20)\n",
        "        colors = ['r', 'g', 'b','c','m','y','k','gold','grey','purple']\n",
        "        k=[0,2,6,8,9]\n",
        "        # n = 10\n",
        "        for i in range(10):\n",
        "            ax.scatter(principalDf.loc[range((i*100),((i*100)+vec_cnt[i])), 'principal component 1']\n",
        "                    , principalDf.loc[range((i*100),((i*100)+vec_cnt[i])), 'principal component 2']\n",
        "                    , c = colors[i]\n",
        "                    , s = 25)\n",
        "\n",
        "        #ax.legend(targets)\n",
        "        ax.grid()\n",
        "        plt.savefig('/content/drive/MyDrive/data/'+'TSNE_Output_Epoch_' + str(epoch_number) + '.png')\n",
        "        plt.close()\n",
        "\n",
        "        principalDf = pd.DataFrame(data = principalComponents2, columns = ['principal component 1', 'principal component 2'])\n",
        "        fig = plt.figure(figsize = (8,8))\n",
        "        ax = fig.add_subplot(1,1,1) \n",
        "\n",
        "        #plt.ylim([-30,30])\n",
        "\n",
        "        ax.set_xlabel('Principal Component 1', fontsize = 15)\n",
        "        ax.set_ylabel('Principal Component 2', fontsize = 15)\n",
        "        ax.set_title('PCA Output', fontsize = 20)\n",
        "        colors = ['r', 'g', 'b','c','m','y','k','gold','grey','purple']\n",
        "        k=[0,2,6,8,9]\n",
        "        # n = 10\n",
        "        for i in range(10):\n",
        "            ax.scatter(principalDf.loc[range((i*100),((i*100)+vec_cnt[i])), 'principal component 1']\n",
        "                    , principalDf.loc[range((i*100),((i*100)+vec_cnt[i])), 'principal component 2']\n",
        "                    , c = colors[i]\n",
        "                    , s = 25)\n",
        "\n",
        "        #ax.legend(targets)\n",
        "        ax.grid()\n",
        "        plt.savefig('/content/drive/MyDrive/data/'+'PCA_Output_Epoch_' + str(epoch_number) + '.png')\n",
        "        plt.close()\n",
        "\n",
        "        return x\n",
        "    \n",
        "    principalDf = pd.DataFrame(data = principalComponents\n",
        "                , columns = ['principal component 1', 'principal component 2'])\n",
        "    fig = plt.figure(figsize = (8,8))\n",
        "    ax = fig.add_subplot(1,1,1) \n",
        "\n",
        "    #plt.ylim([-30,30])\n",
        "\n",
        "    ax.set_xlabel('Principal Component 1', fontsize = 15)\n",
        "    ax.set_ylabel('Principal Component 2', fontsize = 15)\n",
        "    ax.set_title('TSNE Output', fontsize = 20)\n",
        "    colors = ['r', 'g', 'b','c','m','y','k','gold','grey','purple']\n",
        "    k=[0,2,6,8,9]\n",
        "    # x = 50\n",
        "    for i in range(10):\n",
        "        ax.scatter(principalDf.loc[range((i*100),((i*100)+vec_cnt[i])), 'principal component 1']\n",
        "                , principalDf.loc[range((i*100),((i*100)+vec_cnt[i])), 'principal component 2']\n",
        "                , c = colors[i]\n",
        "                , s = 50)\n",
        "\n",
        "    #ax.legend(targets)\n",
        "    ax.grid()\n",
        "    plt.savefig('/content/drive/MyDrive/data/'+'TSNE_Output_Epoch_' + str(epoch_number) + '.png')\n",
        "\n",
        "    if visualize == False:\n",
        "        plt.close()\n",
        "\n",
        "    return x"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kjueVZwPXiMS"
      },
      "source": [
        "# model = model.to(device)\n",
        "# p = visualize_embeddings_100(model, train_dataloader, 1, \"BOTH\", False)\n",
        "# p = visualize_embeddings_500(model, train_dataloader, 1, \"BOTH\", False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PMjlNI0ecCEf"
      },
      "source": [
        "### PCA debugging code (ignore)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Gj_D9ig_UXWx",
        "outputId": "889f605c-557b-436e-b66c-b367d4541468"
      },
      "source": [
        "##### Experiment PCA with 500 samples\n",
        "# cnt = 0\n",
        "# i=0\n",
        "# tmp = [0 for i in range(128)]\n",
        "# x = [tmp for i in range(1000)]\n",
        "# vec_cnt = [0 for i in range(10)]\n",
        "\n",
        "# model_ft.eval()\n",
        "\n",
        "# while cnt < 500:\n",
        "\n",
        "#     inputs, labels = next(iter(train_dataloader))\n",
        "#     # print(labels)\n",
        "    \n",
        "#     for i in range(BATCH_SIZE):\n",
        "\n",
        "#         if cnt == 500:\n",
        "#             break\n",
        "        \n",
        "#         input = inputs[i].unsqueeze(0)\n",
        "#         input = input.to(device)\n",
        "#         vec = torch.squeeze(model_ft(input))\n",
        "        \n",
        "#         # print((labels[i]*100 + vec_cnt[labels[i]]))\n",
        "#         x[(labels[i]*100 + vec_cnt[labels[i]])] = vec.detach().numpy()\n",
        "#         vec_cnt[labels[i]] += 1\n",
        "#         cnt += 1\n",
        "#         i += 1\n",
        "\n",
        "# x = StandardScaler().fit_transform(x)\n",
        "\n",
        "# principalComponents = TSNE(n_components=2, learning_rate='auto', init='random').fit_transform(x)\n",
        "# principalDf = pd.DataFrame(data = principalComponents\n",
        "#              , columns = ['principal component 1', 'principal component 2'])\n",
        "# fig = plt.figure(figsize = (8,8))\n",
        "# ax = fig.add_subplot(1,1,1) \n",
        "\n",
        "# #plt.ylim([-30,30])\n",
        "\n",
        "# ax.set_xlabel('Principal Component 1', fontsize = 15)\n",
        "# ax.set_ylabel('Principal Component 2', fontsize = 15)\n",
        "# ax.set_title('TSNE Output', fontsize = 20)\n",
        "# colors = ['r', 'g', 'b','c','m','y','k','gold','grey','purple']\n",
        "# k=[0,2,6,8,9]\n",
        "# x = 50\n",
        "# for i in range(10):\n",
        "#     ax.scatter(principalDf.loc[range((i*100),((i*100)+vec_cnt[i])), 'principal component 1']\n",
        "#                , principalDf.loc[range((i*100),((i*100)+vec_cnt[i])), 'principal component 2']\n",
        "#                , c = colors[i]\n",
        "#                , s = 50)\n",
        "\n",
        "# #ax.legend(targets)\n",
        "# ax.grid()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "SKJNr_ppdxaR",
        "outputId": "a02edd53-c699-434a-a901-2c80d23b779c"
      },
      "source": [
        "# print(vec_cnt) #Trial 3"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[44, 37, 58, 68, 55, 41, 44, 36, 51, 66]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5YfNWUUEdOwR",
        "outputId": "e3de950e-137b-4e46-9715-3e504b379ef4"
      },
      "source": [
        "# print(vec_cnt) #Trial 2"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[29, 58, 62, 42, 67, 39, 59, 35, 69, 40]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ecLWKKs8b0vn",
        "outputId": "2f65c277-24c3-4a53-80b0-97972be4d3ec"
      },
      "source": [
        "# print(vec_cnt) #Trial 1\n",
        "# print(sum(vec_cnt))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[44, 46, 50, 47, 41, 56, 37, 66, 50, 63]\n",
            "500\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ob8gy4STniIG"
      },
      "source": [
        "# ##### PCA from trainloader\n",
        "# cnt = 0\n",
        "# i=0\n",
        "# tmp = [0 for i in range(128)]\n",
        "# x = [tmp for i in range(100)]\n",
        "# vec_cnt = [0 for i in range(10)]\n",
        "\n",
        "# model_ft.eval()\n",
        "\n",
        "# while cnt < 100:\n",
        "\n",
        "#     inputs, labels = next(iter(train_dataloader))\n",
        "    \n",
        "#     for i in range(BATCH_SIZE):\n",
        "#         if vec_cnt[labels[i]] == 10:\n",
        "#             i += 1\n",
        "#             continue\n",
        "\n",
        "#         if cnt == 100:\n",
        "#             break\n",
        "        \n",
        "#         input = inputs[i].unsqueeze(0)\n",
        "#         input = input.to(device)\n",
        "#         vec = torch.squeeze(model_ft(input))\n",
        "        \n",
        "#         x[(labels[i]*10 + vec_cnt[labels[i]])] = vec.detach().numpy()\n",
        "#         vec_cnt[labels[i]] += 1\n",
        "#         cnt += 1\n",
        "#         i += 1"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-3YFaSCUBb9u"
      },
      "source": [
        "# ##### PCA from dataset directly\n",
        "# cnt = 0\n",
        "# i=0\n",
        "# x = [[0] for i in range(100)]\n",
        "# vec_cnt = [0 for i in range(10)]\n",
        "\n",
        "# model_ft.eval()\n",
        "\n",
        "# while cnt < 100:\n",
        "\n",
        "#     input, label = training_data[i]\n",
        "\n",
        "#     if vec_cnt[label] == 10:\n",
        "#         i += 1\n",
        "#         continue\n",
        "    \n",
        "#     input = input.unsqueeze(0)\n",
        "#     input = input.to(device)\n",
        "#     vec = torch.squeeze(model_ft(input))\n",
        "    \n",
        "#     x[(label*10 + vec_cnt[label])] = vec.detach().numpy()\n",
        "#     vec_cnt[label] += 1\n",
        "#     cnt += 1\n",
        "#     i += 1"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "6sjJCKFrNqlN",
        "outputId": "fc928ac8-8836-43b5-8e8a-4962d0c278c6"
      },
      "source": [
        "# print(vec_cnt)\n",
        "# # print(x)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[10, 10, 10, 10, 10, 10, 10, 10, 10, 10]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "51ShYBOCMIXc"
      },
      "source": [
        "# temp = []\n",
        "# for i in range(100):\n",
        "#     for j in range(128):\n",
        "#         print(i, j)\n",
        "#         temp.append(x[i][j])\n",
        "\n",
        "#     x[i] = temp\n",
        "#     temp = []"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "HWISYIWCMeP5",
        "outputId": "5ee2f487-4207-43c0-c9e4-8dc25200ee0d"
      },
      "source": [
        "# print(len(x))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "100\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bRd8QN1WSAoy"
      },
      "source": [
        "# x = StandardScaler().fit_transform(x)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Y53tT8fJI9eq",
        "outputId": "48ee7e2d-e5d2-4f4e-9981-a331a716e335"
      },
      "source": [
        "# #### PCA\n",
        "# # pca = PCA(n_components=2)\n",
        "# # principalComponents = pca.fit_transform(x)\n",
        "# # principalDf = pd.DataFrame(data = principalComponents\n",
        "# #              , columns = ['principal component 1', 'principal component 2'])\n",
        "# # fig = plt.figure(figsize = (8,8))\n",
        "# # ax = fig.add_subplot(1,1,1) \n",
        "# # #plt.ylim([-30,30])\n",
        "# # ax.set_xlabel('Principal Component 1', fontsize = 15)\n",
        "# # ax.set_ylabel('Principal Component 2', fontsize = 15)\n",
        "# # ax.set_title('PCA Output', fontsize = 20)\n",
        "# # colors = ['r', 'g', 'b','c','m','y','k','gold','grey','purple']\n",
        "# # k=[0,2,6,8,9]\n",
        "# # x = 10\n",
        "# # for i in range(10):\n",
        "# #     ax.scatter(principalDf.loc[range((i*x),((i*x)+x)), 'principal component 1']\n",
        "# #                , principalDf.loc[range((i*x),((i*x)+x)), 'principal component 2']\n",
        "# #                , c = colors[i]\n",
        "# #                , s = 50)\n",
        "# # #ax.legend(targets)\n",
        "# # ax.grid()\n",
        "\n",
        "\n",
        "# #### TSNE\n",
        "# principalComponents = TSNE(n_components=2, learning_rate='auto', init='random').fit_transform(x)\n",
        "# principalDf = pd.DataFrame(data = principalComponents\n",
        "#              , columns = ['principal component 1', 'principal component 2'])\n",
        "# fig = plt.figure(figsize = (8,8))\n",
        "# ax = fig.add_subplot(1,1,1) \n",
        "\n",
        "# #plt.ylim([-30,30])\n",
        "\n",
        "# ax.set_xlabel('Principal Component 1', fontsize = 15)\n",
        "# ax.set_ylabel('Principal Component 2', fontsize = 15)\n",
        "# ax.set_title('TSNE Output', fontsize = 20)\n",
        "# colors = ['r', 'g', 'b','c','m','y','k','gold','grey','purple']\n",
        "# k=[0,2,6,8,9]\n",
        "# x = 10\n",
        "# for i in range(10):\n",
        "#     ax.scatter(principalDf.loc[range((i*x),((i*x)+x)), 'principal component 1']\n",
        "#                , principalDf.loc[range((i*x),((i*x)+x)), 'principal component 2']\n",
        "#                , c = colors[i]\n",
        "#                , s = 50)\n",
        "\n",
        "# #ax.legend(targets)\n",
        "# ax.grid()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "I9-0nQIMdZM1"
      },
      "source": [
        "## Training\n",
        "**torch.autograd.set_detect_anomaly(True)** was used to trace what part of the code was causing there to be Nan values.\n",
        "\n",
        "**Error was traced to be:** RuntimeError: Function 'PowBackward0' returned nan values in its 0th output."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Uxlu82mXtl0v"
      },
      "source": [
        "torch.autograd.set_detect_anomaly(True)\n",
        "train_model(model, train_dataloader, loss_fn, optimizer, lambd=5*1e-3, num_epochs=50) #labmd = 5*1e-3"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uWe0l277Uv68"
      },
      "source": [
        "# torch.save(model_ft.state_dict(), '/content/drive/MyDrive/data/weights_stl10_resnet_self_supervised.pth')"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}