LoRa-ResNET.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "33095728",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:26:00.268958Z",
     "iopub.status.busy": "2024-05-30T09:26:00.268579Z",
     "iopub.status.idle": "2024-05-30T09:26:06.705538Z",
     "shell.execute_reply": "2024-05-30T09:26:06.704768Z"
    },
    "papermill": {
     "duration": 6.449134,
     "end_time": "2024-05-30T09:26:06.707916",
     "exception": false,
     "start_time": "2024-05-30T09:26:00.258782",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "import torchvision\n",
    "import torchvision.transforms as transforms\n",
    "import torchvision.datasets as datasets\n",
    "\n",
    "import os\n",
    "import random\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "48f8577c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:26:06.727543Z",
     "iopub.status.busy": "2024-05-30T09:26:06.726544Z",
     "iopub.status.idle": "2024-05-30T09:29:57.408439Z",
     "shell.execute_reply": "2024-05-30T09:29:57.407267Z"
    },
    "papermill": {
     "duration": 230.693795,
     "end_time": "2024-05-30T09:29:57.410782",
     "exception": false,
     "start_time": "2024-05-30T09:26:06.716987",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "!conda install -y gdown\n",
    "# https://drive.google.com/file/d/13vbSVWSeZkOAdP_bIxYhOSWtWSpyWBD0/view?usp=share_link\n",
    "!gdown --id 13vbSVWSeZkOAdP_bIxYhOSWtWSpyWBD0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "02d8a6b5",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:29:57.700200Z",
     "iopub.status.busy": "2024-05-30T09:29:57.699815Z",
     "iopub.status.idle": "2024-05-30T09:30:27.564665Z",
     "shell.execute_reply": "2024-05-30T09:30:27.563473Z"
    },
    "papermill": {
     "duration": 29.990327,
     "end_time": "2024-05-30T09:30:27.567026",
     "exception": false,
     "start_time": "2024-05-30T09:29:57.576699",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "! unzip folder.zip"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "c29ba112",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:27.942076Z",
     "iopub.status.busy": "2024-05-30T09:30:27.941306Z",
     "iopub.status.idle": "2024-05-30T09:30:28.918234Z",
     "shell.execute_reply": "2024-05-30T09:30:28.917108Z"
    },
    "papermill": {
     "duration": 1.15588,
     "end_time": "2024-05-30T09:30:28.920589",
     "exception": false,
     "start_time": "2024-05-30T09:30:27.764709",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__notebook__.ipynb  folder.zip\tlora-nonlora\r\n"
     ]
    }
   ],
   "source": [
    "! ls"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "519d9287",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:29.272161Z",
     "iopub.status.busy": "2024-05-30T09:30:29.271428Z",
     "iopub.status.idle": "2024-05-30T09:30:29.279989Z",
     "shell.execute_reply": "2024-05-30T09:30:29.279243Z"
    },
    "papermill": {
     "duration": 0.186499,
     "end_time": "2024-05-30T09:30:29.282006",
     "exception": false,
     "start_time": "2024-05-30T09:30:29.095507",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "SEED = 1234\n",
    "\n",
    "random.seed(SEED)\n",
    "np.random.seed(SEED)\n",
    "torch.manual_seed(SEED)\n",
    "torch.cuda.manual_seed(SEED)\n",
    "torch.backends.cudnn.deterministic = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "bbf0e8b1",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:29.635286Z",
     "iopub.status.busy": "2024-05-30T09:30:29.634928Z",
     "iopub.status.idle": "2024-05-30T09:30:29.642241Z",
     "shell.execute_reply": "2024-05-30T09:30:29.641484Z"
    },
    "papermill": {
     "duration": 0.187173,
     "end_time": "2024-05-30T09:30:29.644123",
     "exception": false,
     "start_time": "2024-05-30T09:30:29.456950",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "train_transforms = transforms.Compose([\n",
    "                           transforms.RandomHorizontalFlip(),\n",
    "                           transforms.RandomRotation(10),\n",
    "                           transforms.RandomCrop((224, 224), pad_if_needed=True),\n",
    "                           transforms.ToTensor(),\n",
    "                           transforms.Normalize((0.485, 0.456, 0.406),(0.229, 0.224, 0.225))\n",
    "                       ])\n",
    "\n",
    "test_transforms = transforms.Compose([\n",
    "                           transforms.CenterCrop((224, 224)),\n",
    "                           transforms.ToTensor(),\n",
    "                           transforms.Normalize((0.485, 0.456, 0.406),(0.229, 0.224, 0.225))\n",
    "                       ])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "7c578e04",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:29.994423Z",
     "iopub.status.busy": "2024-05-30T09:30:29.994042Z",
     "iopub.status.idle": "2024-05-30T09:30:30.023354Z",
     "shell.execute_reply": "2024-05-30T09:30:30.022576Z"
    },
    "papermill": {
     "duration": 0.206271,
     "end_time": "2024-05-30T09:30:30.025376",
     "exception": false,
     "start_time": "2024-05-30T09:30:29.819105",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "train_data = datasets.ImageFolder('/kaggle/working/lora-nonlora/train', train_transforms)\n",
    "valid_data = datasets.ImageFolder('/kaggle/working/lora-nonlora/valid', test_transforms)\n",
    "test_data = datasets.ImageFolder('/kaggle/working/lora-nonlora/test', test_transforms)\n",
    "\n",
    "#import os\n",
    "\n",
    "#print(len(os.listdir('data/dogs-vs-cats/train')))\n",
    "\n",
    "#n_train_examples = int(len(train_data)*0.9)\n",
    "#n_valid_examples = n_test_examples = len(train_data) - n_train_examples\n",
    "\n",
    "#train_data, valid_data = torch.utils.data.random_split(train_data, [n_train_examples, n_valid_examples])\n",
    "#train_data, test_data = torch.utils.data.random_split(train_data, [n_train_examples-n_valid_examples, n_test_examples])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f18fdc2b",
   "metadata": {
    "papermill": {
     "duration": 0.176654,
     "end_time": "2024-05-30T09:30:30.386371",
     "exception": false,
     "start_time": "2024-05-30T09:30:30.209717",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "https://github.com/facebook/fb.resnet.torch/issues/180\n",
    "https://github.com/bamos/densenet.pytorch/blob/master/compute-cifar10-mean.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "f1e9310a",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:30.739024Z",
     "iopub.status.busy": "2024-05-30T09:30:30.738265Z",
     "iopub.status.idle": "2024-05-30T09:30:30.743405Z",
     "shell.execute_reply": "2024-05-30T09:30:30.742457Z"
    },
    "papermill": {
     "duration": 0.186141,
     "end_time": "2024-05-30T09:30:30.745979",
     "exception": false,
     "start_time": "2024-05-30T09:30:30.559838",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of training examples: 3156\n",
      "Number of validation examples: 80\n",
      "Number of testing examples: 1000\n"
     ]
    }
   ],
   "source": [
    "print(f'Number of training examples: {len(train_data)}')\n",
    "print(f'Number of validation examples: {len(valid_data)}')\n",
    "print(f'Number of testing examples: {len(test_data)}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "39680a15",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:31.097645Z",
     "iopub.status.busy": "2024-05-30T09:30:31.096921Z",
     "iopub.status.idle": "2024-05-30T09:30:31.102985Z",
     "shell.execute_reply": "2024-05-30T09:30:31.102112Z"
    },
    "papermill": {
     "duration": 0.183493,
     "end_time": "2024-05-30T09:30:31.104910",
     "exception": false,
     "start_time": "2024-05-30T09:30:30.921417",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "BATCH_SIZE = 64\n",
    "\n",
    "train_iterator = torch.utils.data.DataLoader(train_data, shuffle=True, batch_size=BATCH_SIZE)\n",
    "# print(train_data)\n",
    "# for data, label in train_iterator:\n",
    "#     print(data)\n",
    "#     print(label)\n",
    "valid_iterator = torch.utils.data.DataLoader(valid_data, batch_size=BATCH_SIZE)\n",
    "test_iterator = torch.utils.data.DataLoader(test_data, batch_size=BATCH_SIZE)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d86e2a75",
   "metadata": {
    "papermill": {
     "duration": 0.175598,
     "end_time": "2024-05-30T09:30:31.454299",
     "exception": false,
     "start_time": "2024-05-30T09:30:31.278701",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "https://discuss.pytorch.org/t/why-does-the-resnet-model-given-by-pytorch-omit-biases-from-the-convolutional-layer/10990/4\n",
    "https://github.com/kuangliu/pytorch-cifar/blob/master/models/resnet.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "a2e75214",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:31.808018Z",
     "iopub.status.busy": "2024-05-30T09:30:31.807625Z",
     "iopub.status.idle": "2024-05-30T09:30:31.898256Z",
     "shell.execute_reply": "2024-05-30T09:30:31.897256Z"
    },
    "papermill": {
     "duration": 0.265911,
     "end_time": "2024-05-30T09:30:31.900323",
     "exception": false,
     "start_time": "2024-05-30T09:30:31.634412",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "print(torch.cuda.is_available())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "bbdeb87b",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:32.252239Z",
     "iopub.status.busy": "2024-05-30T09:30:32.251908Z",
     "iopub.status.idle": "2024-05-30T09:30:32.256427Z",
     "shell.execute_reply": "2024-05-30T09:30:32.255606Z"
    },
    "papermill": {
     "duration": 0.1822,
     "end_time": "2024-05-30T09:30:32.258297",
     "exception": false,
     "start_time": "2024-05-30T09:30:32.076097",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "cca23c4f",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:32.652046Z",
     "iopub.status.busy": "2024-05-30T09:30:32.651090Z",
     "iopub.status.idle": "2024-05-30T09:30:33.154146Z",
     "shell.execute_reply": "2024-05-30T09:30:33.153355Z"
    },
    "papermill": {
     "duration": 0.723138,
     "end_time": "2024-05-30T09:30:33.156610",
     "exception": false,
     "start_time": "2024-05-30T09:30:32.433472",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import torchvision.models as models\n",
    "\n",
    "model = models.resnet18().to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "29a467e1",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:33.509239Z",
     "iopub.status.busy": "2024-05-30T09:30:33.508885Z",
     "iopub.status.idle": "2024-05-30T09:30:33.514396Z",
     "shell.execute_reply": "2024-05-30T09:30:33.513580Z"
    },
    "papermill": {
     "duration": 0.18387,
     "end_time": "2024-05-30T09:30:33.517287",
     "exception": false,
     "start_time": "2024-05-30T09:30:33.333417",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ResNet(\n",
      "  (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
      "  (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "  (relu): ReLU(inplace=True)\n",
      "  (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "  (layer1): Sequential(\n",
      "    (0): BasicBlock(\n",
      "      (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    )\n",
      "    (1): BasicBlock(\n",
      "      (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    )\n",
      "  )\n",
      "  (layer2): Sequential(\n",
      "    (0): BasicBlock(\n",
      "      (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (downsample): Sequential(\n",
      "        (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "        (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (1): BasicBlock(\n",
      "      (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    )\n",
      "  )\n",
      "  (layer3): Sequential(\n",
      "    (0): BasicBlock(\n",
      "      (conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (downsample): Sequential(\n",
      "        (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (1): BasicBlock(\n",
      "      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    )\n",
      "  )\n",
      "  (layer4): Sequential(\n",
      "    (0): BasicBlock(\n",
      "      (conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (downsample): Sequential(\n",
      "        (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "        (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (1): BasicBlock(\n",
      "      (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    )\n",
      "  )\n",
      "  (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))\n",
      "  (fc): Linear(in_features=512, out_features=1000, bias=True)\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "print(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "a3f96ad2",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:33.865360Z",
     "iopub.status.busy": "2024-05-30T09:30:33.865011Z",
     "iopub.status.idle": "2024-05-30T09:30:33.869882Z",
     "shell.execute_reply": "2024-05-30T09:30:33.868989Z"
    },
    "papermill": {
     "duration": 0.178098,
     "end_time": "2024-05-30T09:30:33.871777",
     "exception": false,
     "start_time": "2024-05-30T09:30:33.693679",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "for param in model.parameters():\n",
    "    param.requires_grad = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "6f935cf0",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:34.213241Z",
     "iopub.status.busy": "2024-05-30T09:30:34.212405Z",
     "iopub.status.idle": "2024-05-30T09:30:34.217349Z",
     "shell.execute_reply": "2024-05-30T09:30:34.216519Z"
    },
    "papermill": {
     "duration": 0.177919,
     "end_time": "2024-05-30T09:30:34.219477",
     "exception": false,
     "start_time": "2024-05-30T09:30:34.041558",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Linear(in_features=512, out_features=1000, bias=True)\n"
     ]
    }
   ],
   "source": [
    "print(model.fc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "cb3b09c2",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:34.561501Z",
     "iopub.status.busy": "2024-05-30T09:30:34.560741Z",
     "iopub.status.idle": "2024-05-30T09:30:34.565883Z",
     "shell.execute_reply": "2024-05-30T09:30:34.565022Z"
    },
    "papermill": {
     "duration": 0.17795,
     "end_time": "2024-05-30T09:30:34.567820",
     "exception": false,
     "start_time": "2024-05-30T09:30:34.389870",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "model.fc = nn.Linear(in_features=512, out_features=2).to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "97c7adcd",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:34.932957Z",
     "iopub.status.busy": "2024-05-30T09:30:34.932288Z",
     "iopub.status.idle": "2024-05-30T09:30:34.937465Z",
     "shell.execute_reply": "2024-05-30T09:30:34.936623Z"
    },
    "papermill": {
     "duration": 0.198264,
     "end_time": "2024-05-30T09:30:34.939372",
     "exception": false,
     "start_time": "2024-05-30T09:30:34.741108",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "optimizer = optim.Adam(model.parameters())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "37b9c397",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:35.279640Z",
     "iopub.status.busy": "2024-05-30T09:30:35.279197Z",
     "iopub.status.idle": "2024-05-30T09:30:35.283824Z",
     "shell.execute_reply": "2024-05-30T09:30:35.282951Z"
    },
    "papermill": {
     "duration": 0.177517,
     "end_time": "2024-05-30T09:30:35.285698",
     "exception": false,
     "start_time": "2024-05-30T09:30:35.108181",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "criterion = nn.CrossEntropyLoss()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "5726e65f",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:35.626105Z",
     "iopub.status.busy": "2024-05-30T09:30:35.625773Z",
     "iopub.status.idle": "2024-05-30T09:30:35.631188Z",
     "shell.execute_reply": "2024-05-30T09:30:35.630187Z"
    },
    "papermill": {
     "duration": 0.178535,
     "end_time": "2024-05-30T09:30:35.633142",
     "exception": false,
     "start_time": "2024-05-30T09:30:35.454607",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def calculate_accuracy(fx, y):\n",
    "    preds = fx.max(1, keepdim=True)[1]\n",
    "    correct = preds.eq(y.view_as(preds)).sum()\n",
    "    acc = correct.float()/preds.shape[0]\n",
    "    return acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "2e7adddb",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:35.981288Z",
     "iopub.status.busy": "2024-05-30T09:30:35.980631Z",
     "iopub.status.idle": "2024-05-30T09:30:35.987930Z",
     "shell.execute_reply": "2024-05-30T09:30:35.986921Z"
    },
    "papermill": {
     "duration": 0.183927,
     "end_time": "2024-05-30T09:30:35.990164",
     "exception": false,
     "start_time": "2024-05-30T09:30:35.806237",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def train(model, device, iterator, optimizer, criterion):\n",
    "    \n",
    "    epoch_loss = 0\n",
    "    epoch_acc = 0\n",
    "    \n",
    "    model.train()\n",
    "    \n",
    "    for (x, y) in iterator:\n",
    "        \n",
    "        x = x.to(device)\n",
    "        y = y.to(device)\n",
    "        \n",
    "        optimizer.zero_grad()\n",
    "                \n",
    "        fx = model(x)\n",
    "        loss = criterion(fx, y)\n",
    "        \n",
    "        acc = calculate_accuracy(fx, y)\n",
    "        \n",
    "        loss.backward()\n",
    "        \n",
    "        optimizer.step()\n",
    "        \n",
    "        epoch_loss += loss.item()\n",
    "        epoch_acc += acc.item()\n",
    "        \n",
    "    return epoch_loss / len(iterator), epoch_acc / len(iterator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "3c280ebb",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:36.336239Z",
     "iopub.status.busy": "2024-05-30T09:30:36.335351Z",
     "iopub.status.idle": "2024-05-30T09:30:36.342391Z",
     "shell.execute_reply": "2024-05-30T09:30:36.341429Z"
    },
    "papermill": {
     "duration": 0.181965,
     "end_time": "2024-05-30T09:30:36.344314",
     "exception": false,
     "start_time": "2024-05-30T09:30:36.162349",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def evaluate(model, device, iterator, criterion):\n",
    "    \n",
    "    epoch_loss = 0\n",
    "    epoch_acc = 0\n",
    "    \n",
    "    model.eval()\n",
    "    \n",
    "    with torch.no_grad():\n",
    "        for (x, y) in iterator:\n",
    "\n",
    "            x = x.to(device)\n",
    "            y = y.to(device)\n",
    "\n",
    "            fx = model(x)\n",
    "            # print(x)\n",
    "            # print(fx)\n",
    "\n",
    "            loss = criterion(fx, y)\n",
    "\n",
    "            acc = calculate_accuracy(fx, y)\n",
    "\n",
    "            epoch_loss += loss.item()\n",
    "            epoch_acc += acc.item()\n",
    "        \n",
    "    return epoch_loss / len(iterator), epoch_acc / len(iterator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "7eb2da2d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:30:36.691352Z",
     "iopub.status.busy": "2024-05-30T09:30:36.690961Z",
     "iopub.status.idle": "2024-05-30T09:42:11.666022Z",
     "shell.execute_reply": "2024-05-30T09:42:11.665036Z"
    },
    "papermill": {
     "duration": 695.322249,
     "end_time": "2024-05-30T09:42:11.839136",
     "exception": false,
     "start_time": "2024-05-30T09:30:36.516887",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "| Epoch: 01 | Train Loss: 0.319 | Train Acc: 87.99% | Val. Loss: 0.326 | Val. Acc: 96.88% |\n",
      "| Epoch: 02 | Train Loss: 0.257 | Train Acc: 90.99% | Val. Loss: 0.382 | Val. Acc: 84.38% |\n",
      "| Epoch: 03 | Train Loss: 0.235 | Train Acc: 91.99% | Val. Loss: 0.220 | Val. Acc: 96.88% |\n",
      "| Epoch: 04 | Train Loss: 0.224 | Train Acc: 91.99% | Val. Loss: 0.182 | Val. Acc: 96.88% |\n",
      "| Epoch: 05 | Train Loss: 0.222 | Train Acc: 91.98% | Val. Loss: 0.191 | Val. Acc: 96.88% |\n",
      "| Epoch: 06 | Train Loss: 0.225 | Train Acc: 91.98% | Val. Loss: 0.177 | Val. Acc: 96.88% |\n",
      "| Epoch: 07 | Train Loss: 0.214 | Train Acc: 92.48% | Val. Loss: 0.164 | Val. Acc: 96.88% |\n",
      "| Epoch: 08 | Train Loss: 0.208 | Train Acc: 92.46% | Val. Loss: 0.153 | Val. Acc: 96.88% |\n",
      "| Epoch: 09 | Train Loss: 0.213 | Train Acc: 92.29% | Val. Loss: 0.154 | Val. Acc: 96.88% |\n",
      "| Epoch: 10 | Train Loss: 0.215 | Train Acc: 91.49% | Val. Loss: 0.144 | Val. Acc: 96.88% |\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "EPOCHS = 10\n",
    "SAVE_DIR = 'models'\n",
    "MODEL_SAVE_PATH = os.path.join(SAVE_DIR, 'resnet-lora.pt')\n",
    "\n",
    "best_valid_loss = float('inf')\n",
    "best_valid_acc = 0.0\n",
    "\n",
    "if not os.path.isdir(f'{SAVE_DIR}'):\n",
    "    os.makedirs(f'{SAVE_DIR}')\n",
    "\n",
    "for epoch in range(EPOCHS):\n",
    "    train_loss, train_acc = train(model, device, train_iterator, optimizer, criterion)\n",
    "    valid_loss, valid_acc = evaluate(model, device, valid_iterator, criterion)\n",
    "    \n",
    "    if valid_loss < best_valid_loss:\n",
    "        best_valid_loss = valid_loss\n",
    "        best_valid_acc = valid_acc\n",
    "\n",
    "        torch.save(model.state_dict(), MODEL_SAVE_PATH)\n",
    "    \n",
    "    print(f'| Epoch: {epoch+1:02} | Train Loss: {train_loss:.3f} | Train Acc: {train_acc*100:05.2f}% | Val. Loss: {valid_loss:.3f} | Val. Acc: {valid_acc*100:05.2f}% |')\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "7a4f9ff4",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:42:12.187051Z",
     "iopub.status.busy": "2024-05-30T09:42:12.186400Z",
     "iopub.status.idle": "2024-05-30T09:42:13.261291Z",
     "shell.execute_reply": "2024-05-30T09:42:13.260062Z"
    },
    "papermill": {
     "duration": 1.252109,
     "end_time": "2024-05-30T09:42:13.263223",
     "exception": false,
     "start_time": "2024-05-30T09:42:12.011114",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "| Test Loss: 0.144 | Test Acc: 96.88% |\n"
     ]
    }
   ],
   "source": [
    "model.load_state_dict(torch.load(MODEL_SAVE_PATH))\n",
    "\n",
    "test_loss, test_acc = evaluate(model, device, valid_iterator, criterion)\n",
    "\n",
    "print(f'| Test Loss: {test_loss:.3f} | Test Acc: {test_acc*100:05.2f}% |')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "e0a25cc3",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:42:13.611650Z",
     "iopub.status.busy": "2024-05-30T09:42:13.610859Z",
     "iopub.status.idle": "2024-05-30T09:45:06.383818Z",
     "shell.execute_reply": "2024-05-30T09:45:06.382729Z"
    },
    "papermill": {
     "duration": 172.948376,
     "end_time": "2024-05-30T09:45:06.385795",
     "exception": false,
     "start_time": "2024-05-30T09:42:13.437419",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Loss: 0.196 | Training Accuracy: 92.96%\n",
      "Test Loss: 0.375 | Test Accuracy: 89.47%\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x600 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "train_true_labels = []\n",
    "train_predicted_labels = []\n",
    "test_true_labels = []\n",
    "test_predicted_labels = []\n",
    "\n",
    "model.eval()\n",
    "\n",
    "train_loss, train_acc = evaluate(model, device, train_iterator, criterion)\n",
    "print(f'Training Loss: {train_loss:.3f} | Training Accuracy: {train_acc * 100:.2f}%')\n",
    "\n",
    "\n",
    "for (x, y) in train_iterator:\n",
    "    x = x.to(device)\n",
    "    y = y.to(device)\n",
    "\n",
    "    with torch.no_grad():\n",
    "        fx = model(x)\n",
    "\n",
    "\n",
    "    predicted_class = fx.argmax(1)\n",
    "    \n",
    "   \n",
    "    train_true_labels.extend(y.cpu().numpy())\n",
    "    train_predicted_labels.extend(predicted_class.cpu().numpy())\n",
    "\n",
    "test_loss, test_acc = evaluate(model, device, test_iterator, criterion)\n",
    "print(f'Test Loss: {test_loss:.3f} | Test Accuracy: {test_acc * 100:.2f}%')\n",
    "\n",
    "for (x, y) in test_iterator:\n",
    "    x = x.to(device)\n",
    "    y = y.to(device)\n",
    "\n",
    "    with torch.no_grad():\n",
    "        fx = model(x)\n",
    "\n",
    " \n",
    "    predicted_class = fx.argmax(1)\n",
    "    \n",
    "    \n",
    "    test_true_labels.extend(y.cpu().numpy())\n",
    "    test_predicted_labels.extend(predicted_class.cpu().numpy())\n",
    "\n",
    "\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "test_confusion = confusion_matrix(test_true_labels, test_predicted_labels)\n",
    "\n",
    "\n",
    "classes = ['Lora', 'nonLora']  \n",
    "plt.figure(figsize=(8, 6))\n",
    "sns.heatmap(test_confusion, annot=True, fmt='d', cmap='Blues', xticklabels=classes, yticklabels=classes)\n",
    "plt.xlabel('Predicted')\n",
    "plt.ylabel('True')\n",
    "plt.title('Test Set Confusion Matrix')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "e3b0a67c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:45:06.737978Z",
     "iopub.status.busy": "2024-05-30T09:45:06.737435Z",
     "iopub.status.idle": "2024-05-30T09:45:06.745413Z",
     "shell.execute_reply": "2024-05-30T09:45:06.744527Z"
    },
    "papermill": {
     "duration": 0.18281,
     "end_time": "2024-05-30T09:45:06.747442",
     "exception": false,
     "start_time": "2024-05-30T09:45:06.564632",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.106 0.106\n"
     ]
    }
   ],
   "source": [
    "# calculate false positives\n",
    "false_positives = test_confusion.sum(axis=0) - np.diag(test_confusion)\n",
    "\n",
    "# calculate false negatives\n",
    "false_negatives = test_confusion.sum(axis=1) - np.diag(test_confusion)\n",
    "# calculate false positive rate\n",
    "false_positive_rate = false_positives.sum() / (false_positives.sum() + np.diag(test_confusion).sum())\n",
    "\n",
    "# calculate false negative rate\n",
    "false_negative_rate = false_negatives.sum() / (false_negatives.sum() + np.diag(test_confusion).sum())\n",
    "print(false_positive_rate,false_negative_rate)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9cb094fb",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:45:07.096376Z",
     "iopub.status.busy": "2024-05-30T09:45:07.095594Z",
     "iopub.status.idle": "2024-05-30T09:45:07.823940Z",
     "shell.execute_reply": "2024-05-30T09:45:07.822610Z"
    },
    "papermill": {
     "duration": 0.904204,
     "end_time": "2024-05-30T09:45:07.825517",
     "exception": true,
     "start_time": "2024-05-30T09:45:06.921313",
     "status": "failed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from PIL import Image\n",
    "import torch\n",
    "\n",
    "# Load and preprocess the image\n",
    "image_path = '/kaggle/working/lora-nonlora/test/nonLora/0_2487N.png'\n",
    "image = Image.open(image_path)\n",
    "\n",
    "# Apply the same preprocessing transformations used for the dataset\n",
    "image_transforms = transforms.Compose([\n",
    "    transforms.Resize((224, 224)),\n",
    "    transforms.ToTensor(),\n",
    "    transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))\n",
    "])\n",
    "preprocessed_image = image_transforms(image).unsqueeze(0)  # Add batch dimension\n",
    "\n",
    "# Ensure the model is in evaluation mode\n",
    "model.eval()\n",
    "\n",
    "# Move the preprocessed image to the same device used for training (CPU or GPU)\n",
    "preprocessed_image = preprocessed_image.to(device)\n",
    "\n",
    "# Pass the image through the model\n",
    "with torch.no_grad():\n",
    "    output = model(preprocessed_image)\n",
    "\n",
    "# Get the predicted class (assuming you have 2 classes)\n",
    "predicted_class = output.argmax().item()\n",
    "\n",
    "# Print the predicted class or use it for further processing\n",
    "print(f'Predicted class: {predicted_class}')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "28e092d7",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:22:45.658503Z",
     "iopub.status.busy": "2024-05-30T09:22:45.657652Z",
     "iopub.status.idle": "2024-05-30T09:22:58.189233Z",
     "shell.execute_reply": "2024-05-30T09:22:58.188233Z",
     "shell.execute_reply.started": "2024-05-30T09:22:45.658473Z"
    },
    "papermill": {
     "duration": null,
     "end_time": null,
     "exception": null,
     "start_time": null,
     "status": "pending"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from PIL import Image\n",
    "import os\n",
    "\n",
    "# Directory path containing the images you want to classify\n",
    "image_directory = '/kaggle/working/lora-nonlora/test/nonLora'\n",
    "\n",
    "# List of image file names in the directory\n",
    "image_filenames = os.listdir(image_directory)\n",
    "\n",
    "# Apply the same preprocessing transformations used for the dataset\n",
    "image_transforms = transforms.Compose([\n",
    "    transforms.Resize((224, 224)),\n",
    "    transforms.ToTensor(),\n",
    "    transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))\n",
    "])\n",
    "\n",
    "# Ensure the model is in evaluation mode\n",
    "model.eval()\n",
    "\n",
    "# Initialize a list to store predictions\n",
    "predictions = []\n",
    "\n",
    "# Loop through each image and get predictions\n",
    "for filename in image_filenames:\n",
    "    image_path = os.path.join(image_directory, filename)\n",
    "    image = Image.open(image_path)\n",
    "    \n",
    "    # Preprocess the image\n",
    "    preprocessed_image = image_transforms(image).unsqueeze(0)  # Add batch dimension\n",
    "\n",
    "    # Move the preprocessed image to the same device used for training (CPU or GPU)\n",
    "    preprocessed_image = preprocessed_image.to(device)\n",
    "\n",
    "    # Pass the image through the model\n",
    "    with torch.no_grad():\n",
    "        output = model(preprocessed_image)\n",
    "\n",
    "    # Get the predicted class (assuming you have 2 classes)\n",
    "    predicted_class = output.argmax().item()\n",
    "\n",
    "    # Store the filename and predicted class in the predictions list\n",
    "    predictions.append((filename, predicted_class))\n",
    "\n",
    "# Print or process the predictions\n",
    "for filename, predicted_class in predictions:\n",
    "    print(f'Image: {filename}, Predicted class: {predicted_class}')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "df946f02",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-05-30T09:21:31.846831Z",
     "iopub.status.busy": "2024-05-30T09:21:31.846398Z",
     "iopub.status.idle": "2024-05-30T09:21:41.952955Z",
     "shell.execute_reply": "2024-05-30T09:21:41.952022Z",
     "shell.execute_reply.started": "2024-05-30T09:21:31.846798Z"
    },
    "papermill": {
     "duration": null,
     "end_time": null,
     "exception": null,
     "start_time": null,
     "status": "pending"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "import torchvision.models as models\n",
    "\n",
    "\n",
    "# Load the model\n",
    "\n",
    "# model_checkpoint = torch.load('C:/Users/Abhijith/Desktop/loramodel/models/resnet-lora.pt')\n",
    "# model = model_checkpoint['model']\n",
    "model.eval()\n",
    "from PIL import Image\n",
    "import os\n",
    "\n",
    "# Directory path containing the images you want to classify\n",
    "image_directory = '/kaggle/working/lora-nonlora/test/Lora'\n",
    "\n",
    "# List of image file names in the directory\n",
    "image_filenames = os.listdir(image_directory)\n",
    "\n",
    "# Apply the same preprocessing transformations used for the dataset\n",
    "image_transforms = transforms.Compose([\n",
    "    transforms.Resize((224, 224)),\n",
    "    transforms.ToTensor(),\n",
    "    transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))\n",
    "])\n",
    "# You can now use the model for inference, e.g., for making predictions\n",
    "# Make sure to preprocess your input data appropriately before passing it to the model\n",
    "# Initialize a list to store predictions\n",
    "predictions = []\n",
    "\n",
    "# Loop through each image and get predictions\n",
    "for filename in image_filenames:\n",
    "    image_path = os.path.join(image_directory, filename)\n",
    "    image = Image.open(image_path)\n",
    "    \n",
    "    # Preprocess the image\n",
    "    preprocessed_image = image_transforms(image).unsqueeze(0)  # Add batch dimension\n",
    "\n",
    "    # Move the preprocessed image to the same device used for training (CPU or GPU)\n",
    "    preprocessed_image = preprocessed_image.to(device)\n",
    "\n",
    "    # Pass the image through the model\n",
    "    with torch.no_grad():\n",
    "        output = model(preprocessed_image)\n",
    "\n",
    "    # Get the predicted class (assuming you have 2 classes)\n",
    "    predicted_class = output.argmax().item()\n",
    "\n",
    "    # Store the filename and predicted class in the predictions list\n",
    "    predictions.append((filename, predicted_class))\n",
    "\n",
    "# Print or process the predictions\n",
    "for filename, predicted_class in predictions:\n",
    "    print(f'Image: {filename}, Predicted class: {predicted_class}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "486b2ec1",
   "metadata": {
    "papermill": {
     "duration": null,
     "end_time": null,
     "exception": null,
     "start_time": null,
     "status": "pending"
    },
    "tags": []
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kaggle": {
   "accelerator": "gpu",
   "dataSources": [],
   "dockerImageVersionId": 30699,
   "isGpuEnabled": true,
   "isInternetEnabled": true,
   "language": "python",
   "sourceType": "notebook"
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.13"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 1152.853871,
   "end_time": "2024-05-30T09:45:10.259019",
   "environment_variables": {},
   "exception": true,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2024-05-30T09:25:57.405148",
   "version": "2.5.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}