{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.10.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"gpu","dataSources":[{"sourceId":16017,"databundleVersionId":708189,"sourceType":"competition"}],"dockerImageVersionId":30587,"isInternetEnabled":false,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"# import libraries\nimport torch\nfrom torch.utils.data import DataLoader, Dataset\nimport torchvision\nimport torchvision.transforms as transforms\nfrom torch.utils.data.sampler import SubsetRandomSampler\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport pandas as pd\nimport torch.nn as nn\nimport torch.nn.functional as F","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:35:35.195286Z","iopub.execute_input":"2023-12-09T02:35:35.195699Z","iopub.status.idle":"2023-12-09T02:35:39.060734Z","shell.execute_reply.started":"2023-12-09T02:35:35.195666Z","shell.execute_reply":"2023-12-09T02:35:39.059960Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Dataset classes definition \n","metadata":{}},{"cell_type":"markdown","source":"\n### Train Dataset class","metadata":{}},{"cell_type":"code","source":"class TrainMNIST(Dataset):\n    \n    def __init__(self, file_path, transform=None):\n        device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n        df = pd.read_csv(file_path)\n        self.data = torch.from_numpy(df.values).float().to(device)\n        self.transform = transform\n        \n    def __len__(self):\n        return len(self.data)\n    \n    def __getitem__(self, index):\n        # (C,H,W) because it's a tensor\n        image = self.data[index, 1:].view(1, 28, 28).float()  # Assuming the first column is the label\n        label = self.data[index, 0].byte()  # Assuming the first column is the label\n            \n        return image, label","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:35:39.062318Z","iopub.execute_input":"2023-12-09T02:35:39.062730Z","iopub.status.idle":"2023-12-09T02:35:39.069690Z","shell.execute_reply.started":"2023-12-09T02:35:39.062697Z","shell.execute_reply":"2023-12-09T02:35:39.068697Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Test Dataset class","metadata":{}},{"cell_type":"code","source":"class TestMNIST(Dataset):\n    \n    def __init__(self, file_path, transform=None):\n        device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n        df = pd.read_csv(file_path)\n        self.data = torch.from_numpy(df.values).float().to(device)\n        self.transform = transform\n        \n    def __len__(self):\n        return len(self.data)\n    \n    def __getitem__(self, index):\n        # (C,H,W) because it's a tensor\n        image = self.data[index, 1:].view(1, 28, 28).float()  # Assuming the first column is the label            \n        return image","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:35:39.070758Z","iopub.execute_input":"2023-12-09T02:35:39.071040Z","iopub.status.idle":"2023-12-09T02:35:39.081914Z","shell.execute_reply.started":"2023-12-09T02:35:39.071017Z","shell.execute_reply":"2023-12-09T02:35:39.081198Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Load Data","metadata":{}},{"cell_type":"code","source":"train_data = TrainMNIST('/kaggle/input/Kannada-MNIST/train.csv', transform=transforms.ToTensor())\ntest_data = TestMNIST('/kaggle/input/Kannada-MNIST/test.csv', transform=transforms.ToTensor())","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:35:39.082954Z","iopub.execute_input":"2023-12-09T02:35:39.083276Z","iopub.status.idle":"2023-12-09T02:35:46.920866Z","shell.execute_reply.started":"2023-12-09T02:35:39.083244Z","shell.execute_reply":"2023-12-09T02:35:46.919854Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## init data Loaders","metadata":{}},{"cell_type":"code","source":"num_workers = 0\nbatch_size = 20\n\ntrain_loader = DataLoader(train_data, batch_size = batch_size, num_workers = num_workers)\ntest_loader = DataLoader(test_data, batch_size = batch_size, num_workers = num_workers)","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:35:46.923941Z","iopub.execute_input":"2023-12-09T02:35:46.924244Z","iopub.status.idle":"2023-12-09T02:35:46.929330Z","shell.execute_reply.started":"2023-12-09T02:35:46.924218Z","shell.execute_reply":"2023-12-09T02:35:46.928411Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Defining the Nerual Network's Architecture","metadata":{}},{"cell_type":"markdown","source":"### The NN architecture","metadata":{}},{"cell_type":"code","source":"class Net(nn.Module):\n    def __init__(self):\n        super(Net,self).__init__()\n        Conv2D1 = 64\n        Conv2D2 = 128\n        Dense = 128\n        Output = 10\n        #  extract features using a 3x3 kernel\n        self.conv1 = nn.Conv2d(1, Conv2D1, 3)\n        # resize for better compute\n        self.maxPool = nn.MaxPool2d(2,2)\n        # extract features from features using 3x3 kernel with 2 stride \n        self.conv2 = nn.Conv2d(Conv2D1, Conv2D2, 3,2)\n\n        self.fc1 = nn.Linear(128 * 3 * 3, Dense)\n        self.fc2 = nn.Linear(Dense, Output)\n\n\n    def forward(self,x):\n        # using leaky relu to prevent \"dead\" outputs\n        \n        # out 26 x 26\n        x = F.leaky_relu(self.conv1(x))\n        # out 13 x 13\n        x = self.maxPool(x)\n        # out 6 x 6\n        x = F.leaky_relu(self.conv2(x))\n        # out 3 x 3\n        x = self.maxPool(x)\n        \n        # flatten to # of neurons input * (tensor out size)\n        x = x.view(-1, 128 * 3 * 3)\n        \n        # Fully connected layers with ReLU activation\n        x = F.leaky_relu(self.fc1(x))\n        x = self.fc2(x)\n\n        return x\n\nmodel = Net()\ndevice = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\nmodel.to(device)\nprint(model)\nprint(next(model.parameters()).is_cuda)","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:35:46.930520Z","iopub.execute_input":"2023-12-09T02:35:46.930812Z","iopub.status.idle":"2023-12-09T02:35:46.967566Z","shell.execute_reply.started":"2023-12-09T02:35:46.930788Z","shell.execute_reply":"2023-12-09T02:35:46.966727Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Loss Function and Optimizer","metadata":{}},{"cell_type":"code","source":"criterion = nn.CrossEntropyLoss()\n\noptimizer = torch.optim.SGD(model.parameters(), lr=0.0069)","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:35:46.968664Z","iopub.execute_input":"2023-12-09T02:35:46.971120Z","iopub.status.idle":"2023-12-09T02:35:46.975055Z","shell.execute_reply.started":"2023-12-09T02:35:46.971095Z","shell.execute_reply":"2023-12-09T02:35:46.974202Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Training","metadata":{}},{"cell_type":"code","source":"# number of epochs to train the model\nn_MaxEpochs = 100\ntarget_loss = 0.00000069\n\n# initialize tracker for minimum validation loss\ntrain_loss_min = np.Inf # set initial \"min\" to infinity\n\nfor epoch in range(n_MaxEpochs):\n    # monitor training loss\n    train_loss = 0.0\n    \n    ###################\n    # train the model #\n    ###################\n    model.train() # prep model for training\n    for data, target in train_loader:\n        # clear the gradients of all optimized variables\n        optimizer.zero_grad()\n        # forward pass: compute predicted outputs by passing inputs to the model\n        output = model(data)\n        # calculate the loss\n        loss = criterion(output, target)\n        # backward pass: compute gradient of the loss with respect to model parameters\n        loss.backward()\n        # perform a single optimization step (parameter update)\n        optimizer.step()\n        # update running training loss\n        train_loss += loss.item()*data.size(0)\n        \n   \n    train_loss = train_loss/len(train_loader.dataset)\n    \n    print('Epoch: {} \\tTraining Loss: {:.6f}'.format(\n        epoch+1, \n        train_loss,\n        ))\n    if train_loss <= target_loss:\n        print(\"Target loss achieved, ending training and saving model...\")\n        torch.save(model.state_dict(),'model.pt')\n        break\n    # save model if validation loss has decreased\n    if train_loss <= train_loss_min:\n        print('Training loss decreased ({:.6f} --> {:.8f}).  Saving model ...'.format(\n        train_loss_min,\n        train_loss))\n        torch.save(model.state_dict(), 'model.pt')\n        train_loss_min = train_loss","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:35:46.976603Z","iopub.execute_input":"2023-12-09T02:35:46.976967Z","iopub.status.idle":"2023-12-09T02:38:04.439928Z","shell.execute_reply.started":"2023-12-09T02:35:46.976937Z","shell.execute_reply":"2023-12-09T02:38:04.438932Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Testing the model","metadata":{}},{"cell_type":"markdown","source":"### Visualization","metadata":{}},{"cell_type":"code","source":"model.load_state_dict(torch.load('model.pt'))","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:38:04.441097Z","iopub.execute_input":"2023-12-09T02:38:04.441395Z","iopub.status.idle":"2023-12-09T02:38:04.451192Z","shell.execute_reply.started":"2023-12-09T02:38:04.441368Z","shell.execute_reply":"2023-12-09T02:38:04.450289Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"\n# obtain one batch of test images\ndataiter = iter(test_loader)\nimages = next(dataiter)\n\n# get sample outputs\noutput = model(images)\n# convert output probabilities to predicted class\n_, preds = torch.max(output, 1)\n# prep images for display\nimages = images.cpu().numpy()\n\n# plot the images in the batch, along with predicted and true labels\nfig = plt.figure(figsize=(25, 4))\nfor idx in np.arange(20):\n    ax = fig.add_subplot(2, int(20/2), idx+1, xticks=[], yticks=[])\n    ax.imshow(np.squeeze(images[idx]), cmap='gray')\n    ax.set_title(\"{}\".format(str(preds[idx].item())))","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:38:04.452163Z","iopub.execute_input":"2023-12-09T02:38:04.452438Z","iopub.status.idle":"2023-12-09T02:38:05.214472Z","shell.execute_reply.started":"2023-12-09T02:38:04.452408Z","shell.execute_reply":"2023-12-09T02:38:05.213123Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def predict(model, dataloader):\n    prediction_list = []\n    for i, batch in enumerate(dataloader):\n        outputs = model(batch)\n        _, predicted = torch.max(outputs.data, 1) \n        prediction_list.append(predicted.cpu())\n    return prediction_list","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:38:05.215655Z","iopub.status.idle":"2023-12-09T02:38:05.215986Z","shell.execute_reply.started":"2023-12-09T02:38:05.215827Z","shell.execute_reply":"2023-12-09T02:38:05.215843Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test_loader = DataLoader(test_data, batch_size=1, num_workers=num_workers)\n\npredictions = predict(model,test_loader)\np = []\nfor a in predictions:\n    p.append(a.item())\npredictions = np.array(p)","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:38:05.217004Z","iopub.status.idle":"2023-12-09T02:38:05.217442Z","shell.execute_reply.started":"2023-12-09T02:38:05.217217Z","shell.execute_reply":"2023-12-09T02:38:05.217239Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test_data = pd.read_csv(\"/kaggle/input/Kannada-MNIST/test.csv\")\n\n\nsubmission = pd.DataFrame(data={\n    \"Id\": test_data.index,\n    \"Label\": predictions\n})\nsubmission.to_csv(\"submission.csv\", index=None)","metadata":{"execution":{"iopub.status.busy":"2023-12-09T02:38:05.219082Z","iopub.status.idle":"2023-12-09T02:38:05.219524Z","shell.execute_reply.started":"2023-12-09T02:38:05.219297Z","shell.execute_reply":"2023-12-09T02:38:05.219318Z"},"trusted":true},"execution_count":null,"outputs":[]}]}