Vision-Transformer.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2C2jy1YgjtIp"
      },
      "source": [
        "### Libraries"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "1e2Celhijn7B"
      },
      "outputs": [],
      "source": [
        "import random\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sns\n",
        "\n",
        "import torch\n",
        "from torch import nn\n",
        "from torch.utils.data import DataLoader\n",
        "from torchvision import datasets, transforms\n",
        "\n",
        "from tqdm import tqdm\n",
        "\n",
        "import warnings\n",
        "warnings.filterwarnings('ignore')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bE7RVrJLL2fQ"
      },
      "source": [
        "## Vision-Transformer\n",
        "\n",
        "![ViT.png](data:image/png;base64,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)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FDOApeXMkKDt"
      },
      "source": [
        "### 2-D Positional Embedding"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Uy3URNzEkN7F"
      },
      "outputs": [],
      "source": [
        "class PatchEmbedding(nn.Module):\n",
        "  def __init__(self, imgSize, patchSize, embedDim, channels=3):\n",
        "    super(PatchEmbedding, self).__init__()\n",
        "\n",
        "    assert imgSize % patchSize == 0, \"Patch Dimensions should be divisible by Image Dimensions\"\n",
        "\n",
        "    self.imgSize = imgSize\n",
        "    self.patchSize = patchSize\n",
        "    self.nPatches = (self.imgSize * self.imgSize)// (self.patchSize * self.patchSize)\n",
        "\n",
        "    self.proj = nn.Conv2d(channels, embedDim, kernel_size=patchSize, stride=patchSize)\n",
        "\n",
        "  def forward(self, x):\n",
        "    x = self.proj(x)\n",
        "    x = x.flatten(2)\n",
        "    x = x.transpose(1, 2)\n",
        "\n",
        "    return x"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xDj8lDFF0aMV"
      },
      "source": [
        "### Multi-Head Attention"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "5LFKLy1303FK"
      },
      "outputs": [],
      "source": [
        "class MultiHeadAttention(nn.Module):\n",
        "  def __init__(self, dim, numHeads, qkv_bias=True, attnP=0, projP=0):\n",
        "    super(MultiHeadAttention, self).__init__()\n",
        "\n",
        "    self.numHeads = numHeads\n",
        "    self.dim = dim\n",
        "    self.headDim = dim // numHeads\n",
        "    self.scale = self.headDim ** -0.5\n",
        "\n",
        "    self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)\n",
        "    self.attnDrop = nn.Dropout(attnP)\n",
        "    self.proj = nn.Linear(dim, dim)\n",
        "    self.projDrop = nn.Dropout(projP)\n",
        "\n",
        "  def forward(self, x):\n",
        "    B, N, C = x.shape\n",
        "    qkv = self.qkv(x).reshape(B, N, 3, self.numHeads, self.headDim).permute(2, 0, 3, 1, 4)\n",
        "    Q, K, V = qkv[0], qkv[1], qkv[2]\n",
        "\n",
        "    attention = (Q @ K.transpose(-1, -2)) * self.scale\n",
        "    attention = attention.softmax(dim=-1)\n",
        "    attention = self.attnDrop(attention)\n",
        "\n",
        "    x = (attention @ V).transpose(1, 2).reshape(B, N, C)\n",
        "    x = self.proj(x)\n",
        "    x = self.projDrop(x)\n",
        "\n",
        "    return x"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pW10LyKlF_8m"
      },
      "source": [
        "### Multi-Layer Perceptron"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "3ZUKmfw-GG1m"
      },
      "outputs": [],
      "source": [
        "class MLP(nn.Module):\n",
        "  def __init__(self, inFeatures, hiddenFeatures, outFeatures, dropout=0.1):\n",
        "    super(MLP, self).__init__()\n",
        "\n",
        "    self.fc1 = nn.Linear(inFeatures, hiddenFeatures)\n",
        "    self.act = nn.GELU()\n",
        "    self.fc2 = nn.Linear(hiddenFeatures, outFeatures)\n",
        "    self.dropout = nn.Dropout(dropout)\n",
        "\n",
        "  def forward(self, x):\n",
        "    x = self.fc1(x)\n",
        "    x = self.act(x)\n",
        "    x = self.dropout(x)\n",
        "    x = self.fc2(x)\n",
        "    x = self.dropout(x)\n",
        "\n",
        "    return x"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kjZDgRueJVuZ"
      },
      "source": [
        "### Block\n",
        "\n",
        "* Combination of Multi-Head Attention and MLP\n",
        "* Layer Normalization is done prior to Multi-Head Attention and MLP, unlike normal transformer"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "QTgSfn8aHOTV"
      },
      "outputs": [],
      "source": [
        "class Block(nn.Module):\n",
        "  def __init__(self, dim, nHeads, mlp_ratio=4.0, qkv_bias=True, projP=0., attnP=0.):\n",
        "    super(Block, self).__init__()\n",
        "\n",
        "    self.layerNorm1 = nn.LayerNorm(dim, eps=1e-6)\n",
        "    self.attention = MultiHeadAttention(dim, nHeads, qkv_bias=qkv_bias, attnP = attnP, projP=projP)\n",
        "    self.layerNorm2 = nn.LayerNorm(dim, eps=1e-6)\n",
        "\n",
        "    hiddenFeatures = int(dim * mlp_ratio)\n",
        "\n",
        "    self.mlp = MLP(\n",
        "        inFeatures=dim,\n",
        "        hiddenFeatures=hiddenFeatures,\n",
        "        outFeatures=dim,\n",
        "    )\n",
        "\n",
        "  def forward(self, x):\n",
        "    x = x + self.attention(self.layerNorm1(x))\n",
        "    x = x + self.mlp(self.layerNorm2(x))\n",
        "\n",
        "    return x"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SGeSI11WMq89"
      },
      "source": [
        "### Vision-Transformer Assembly"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "dwMerTCHMqkA"
      },
      "outputs": [],
      "source": [
        "class VisionTransformer(nn.Module):\n",
        "  def __init__(self, imgSize=256, patchSize=32, channels=3, n_classes=1000, embed_dim=768, depth=12,\n",
        "                 nHeads=12, mlp_ratio=4., qkv_bias=True, p=0., attn_p=0.):\n",
        "    super(VisionTransformer, self).__init__()\n",
        "\n",
        "    self.patchEmbedding = PatchEmbedding(imgSize, patchSize, embed_dim, channels)\n",
        "    self.clsToken = nn.Parameter(torch.zeros(1, 1, embed_dim))\n",
        "    self.posToken = nn.Parameter(torch.zeros(1, 1 + self.patchEmbedding.nPatches, embed_dim))\n",
        "    self.posDrop = nn.Dropout(p=p)\n",
        "\n",
        "    self.blocks = nn.ModuleList([Block(embed_dim, nHeads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, projP=p, attnP=attn_p) for _ in range(depth)])\n",
        "\n",
        "    self.layerNorm = nn.LayerNorm(embed_dim, eps=1e-6)\n",
        "    self.head = nn.Linear(embed_dim, n_classes)\n",
        "\n",
        "  def forward(self, x):\n",
        "    B = x.shape[0]\n",
        "    x = self.patchEmbedding(x)\n",
        "\n",
        "    clsToken = self.clsToken.expand(B, -1, -1)\n",
        "\n",
        "    x = torch.cat((clsToken, x), dim=1)\n",
        "    x = x + self.posToken\n",
        "    x = self.posDrop(x)\n",
        "\n",
        "    for block in self.blocks:\n",
        "      x = block(x)\n",
        "\n",
        "    x = self.layerNorm(x)\n",
        "    x = x[:, 0]\n",
        "    x = self.head(x)\n",
        "\n",
        "    return x"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oCBewVX5XjdX"
      },
      "source": [
        "## Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HgGmGKrqXiO-"
      },
      "source": [
        "### Model-Configuration"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "imgSize = 32\n",
        "patchSize=32\n",
        "channels=3\n",
        "n_classes=10\n",
        "embed_dim=128\n",
        "depth=2\n",
        "nHeads=2\n",
        "mlp_ratio=2\n",
        "\n",
        "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
        "\n",
        "model = VisionTransformer(imgSize=imgSize, patchSize=patchSize, channels=channels, n_classes=n_classes, embed_dim=embed_dim, depth=depth, nHeads=nHeads, mlp_ratio=mlp_ratio).to(device)"
      ],
      "metadata": {
        "id": "GDsRrMlaqUMI"
      },
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pRLeyNe4sDQJ",
        "outputId": "ee66b48a-25e4-4a84-d2fe-ca8649d5f6ce"
      },
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "VisionTransformer(\n",
            "  (patchEmbedding): PatchEmbedding(\n",
            "    (proj): Conv2d(3, 128, kernel_size=(32, 32), stride=(32, 32))\n",
            "  )\n",
            "  (posDrop): Dropout(p=0.0, inplace=False)\n",
            "  (blocks): ModuleList(\n",
            "    (0-1): 2 x Block(\n",
            "      (layerNorm1): LayerNorm((128,), eps=1e-06, elementwise_affine=True)\n",
            "      (attention): MultiHeadAttention(\n",
            "        (qkv): Linear(in_features=128, out_features=384, bias=True)\n",
            "        (attnDrop): Dropout(p=0.0, inplace=False)\n",
            "        (proj): Linear(in_features=128, out_features=128, bias=True)\n",
            "        (projDrop): Dropout(p=0.0, inplace=False)\n",
            "      )\n",
            "      (layerNorm2): LayerNorm((128,), eps=1e-06, elementwise_affine=True)\n",
            "      (mlp): MLP(\n",
            "        (fc1): Linear(in_features=128, out_features=256, bias=True)\n",
            "        (act): GELU(approximate='none')\n",
            "        (fc2): Linear(in_features=256, out_features=128, bias=True)\n",
            "        (dropout): Dropout(p=0.1, inplace=False)\n",
            "      )\n",
            "    )\n",
            "  )\n",
            "  (layerNorm): LayerNorm((128,), eps=1e-06, elementwise_affine=True)\n",
            "  (head): Linear(in_features=128, out_features=10, bias=True)\n",
            ")\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Dataset and DataLoader"
      ],
      "metadata": {
        "id": "pF2PFaHqqaAt"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "epochs = 10\n",
        "batchSize = 128\n",
        "lr = 1e-3\n",
        "\n",
        "transform = transforms.Compose([\n",
        "    transforms.ToTensor(),\n",
        "    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),\n",
        "])\n",
        "\n",
        "trainDataset = datasets.CIFAR10(root='./data', train=True, download=True, transform=transform)\n",
        "trainLoader = DataLoader(trainDataset, batchSize, shuffle=True, num_workers=2)\n",
        "\n",
        "totalBatches = len(trainLoader)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OXjrU_CKqYfG",
        "outputId": "e20dd607-3e35-463f-aeda-b3365fdd0f76"
      },
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Files already downloaded and verified\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(\"DataLoader Dimensions : \", trainLoader.dataset.data.shape)\n",
        "print(\"Total Batches : \", totalBatches)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FV7kEd2_qoEy",
        "outputId": "52afd21b-860e-452d-b5e1-a4d76b29fb43"
      },
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "DataLoader Dimensions :  (50000, 32, 32, 3)\n",
            "Total Batches :  391\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 32,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uUJ2zQvSP0Zp",
        "collapsed": true,
        "outputId": "fd1a4e54-c008-4b36-b2a2-8d11af40e9ae"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch : [1/10]: 100%|██████████| 391/391 [00:39<00:00,  9.94batch/s, Loss=1.48]\n",
            "Epoch : [2/10]: 100%|██████████| 391/391 [00:40<00:00,  9.66batch/s, Loss=1.35]\n",
            "Epoch : [3/10]: 100%|██████████| 391/391 [00:40<00:00,  9.67batch/s, Loss=1.33]\n",
            "Epoch : [4/10]: 100%|██████████| 391/391 [00:39<00:00, 10.02batch/s, Loss=1.23]\n",
            "Epoch : [5/10]: 100%|██████████| 391/391 [00:40<00:00,  9.58batch/s, Loss=1.33]\n",
            "Epoch : [6/10]: 100%|██████████| 391/391 [00:38<00:00, 10.03batch/s, Loss=1.41]\n",
            "Epoch : [7/10]: 100%|██████████| 391/391 [00:40<00:00,  9.73batch/s, Loss=1.34]\n",
            "Epoch : [8/10]: 100%|██████████| 391/391 [00:41<00:00,  9.46batch/s, Loss=1.3]\n",
            "Epoch : [9/10]: 100%|██████████| 391/391 [00:42<00:00,  9.26batch/s, Loss=1.28]\n",
            "Epoch : [10/10]: 100%|██████████| 391/391 [00:40<00:00,  9.54batch/s, Loss=1.17]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Training Complete!\n",
            "\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\n"
          ]
        }
      ],
      "source": [
        "criterion = nn.CrossEntropyLoss()\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=lr)\n",
        "\n",
        "for epoch in range(epochs):\n",
        "  model.train()\n",
        "\n",
        "  totalLoss = 0\n",
        "  correctPredictions = 0\n",
        "  totalPredictions = 0\n",
        "\n",
        "  batchBar = tqdm(trainLoader, desc=f\"Epoch : [{epoch + 1}/{epochs}]\", unit=\"batch\")\n",
        "  for i, data in enumerate(batchBar):\n",
        "    images, labels = data\n",
        "\n",
        "    images = images.to(device)\n",
        "    labels = labels.to(device)\n",
        "\n",
        "    optimizer.zero_grad()\n",
        "\n",
        "    outputs = model(images)\n",
        "    loss = criterion(outputs, labels)\n",
        "    loss.backward()\n",
        "    optimizer.step()\n",
        "\n",
        "    totalLoss += loss.item()\n",
        "    _, predicted = torch.max(outputs.data, 1)\n",
        "    totalPredictions += labels.size(0)\n",
        "    correctPredictions += (predicted == labels).sum().item()\n",
        "\n",
        "    batchBar.set_postfix(Loss=loss.item())\n",
        "\n",
        "print(\"Training Complete!\", end=\"\\n\\n\")\n",
        "\n",
        "torch.save(model.state_dict(), \"ViT_model.pth\")"
      ]
    }
  ],
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}