resnet_freeze.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "3aa403c8",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torchvision.models as models\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "from prepare_data import *\n",
    "from data import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "a5c28013",
   "metadata": {},
   "outputs": [],
   "source": [
    "def prepare_model(device, model):\n",
    "    # load model, criterion, optimizer, and learning rate scheduler\n",
    "    \n",
    "    # Create an empty model and move it to the designated computing device\n",
    "    model=model.to(device)\n",
    "\n",
    "\n",
    "    return model\n",
    "\n",
    "def train_single_model(model, criterion, optimizer, scheduler, device, dataloaders,epoch):\n",
    "    \n",
    "    # prepare the model\n",
    "    model=prepare_model(device,model)\n",
    "    \n",
    "    # store the results of our training\n",
    "    training_losses=[]\n",
    "    training_accs=[]\n",
    "    val_losses=[]\n",
    "    val_accs=[]\n",
    "\n",
    "    for i in range(epoch):\n",
    "        # train the model in each epoch\n",
    "        model.train()\n",
    "        for batch_idx, (data, target) in enumerate(dataloaders[0]):\n",
    "            data, target = data.to(device), target.to(device)\n",
    "            optimizer.zero_grad() # clear the old gradients \n",
    "            output = model(data) # compute outputs of the fc layer \n",
    "            loss = criterion(output, target) \n",
    "            loss.backward() # compute gradient for every variables with requires_grad=True\n",
    "            optimizer.step()\n",
    "        if scheduler is not None:\n",
    "            scheduler.step()\n",
    "        \n",
    "        print(\"finish a epoch of training\")\n",
    "            \n",
    "        # test the model with the training data:\n",
    "        model.eval() # sets model in evaluation (inference) mode. \n",
    "        training_loss = 0 \n",
    "        correct = 0\n",
    "        with torch.no_grad(): # stop storing gradients for the variables\n",
    "            for data, target in dataloaders[0]:\n",
    "                data, target = data.to(device), target.to(device)\n",
    "                output = model(data)\n",
    "                training_loss+=criterion(output, target)\n",
    "                pred = output.argmax(dim=1, keepdim=True) # get the index of maximum fc output. \n",
    "                correct += pred.eq(target.view_as(pred)).sum().item()\n",
    "        training_loss /= len(dataloaders[0].dataset)\n",
    "        training_losses.append(training_loss)\n",
    "        training_acc=correct / len(dataloaders[0].dataset)\n",
    "        training_accs.append(training_acc)\n",
    "        \n",
    "        \n",
    "        # test the model with validation data in each epoch\n",
    "        model.eval() # sets model in evaluation (inference) mode.  \n",
    "        val_loss = 0 \n",
    "        correct = 0\n",
    "        with torch.no_grad(): # stop storing gradients for the variables\n",
    "            for data, target in dataloaders[1]:\n",
    "                data, target = data.to(device), target.to(device)\n",
    "                output = model(data)\n",
    "                val_loss+=criterion(output, target)\n",
    "                pred = output.argmax(dim=1, keepdim=True) # get the index of maximum fc output. \n",
    "                correct += pred.eq(target.view_as(pred)).sum().item()\n",
    "        val_loss /= len(dataloaders[1].dataset)\n",
    "        val_losses.append(val_loss)\n",
    "        val_acc=correct / len(dataloaders[1].dataset)\n",
    "        val_accs.append(val_acc)\n",
    "        print(\"epoch_num: \"+str(i))\n",
    "        print(\"train_loss: \"+str(training_loss))\n",
    "        print(\"val_loss: \"+str(val_loss))\n",
    "\n",
    "    torch.save(model.state_dict(), \"food-101/state_dict_model3.pt\")\n",
    "    return model,training_losses,training_accs,val_losses,val_accs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "02a6e43f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# freeze all the weights for resnet\n",
    "class resnet(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(resnet, self).__init__()\n",
    "        temp=models.resnet18(pretrained=True)\n",
    "        for param in temp.parameters():\n",
    "            param.requires_grad = False\n",
    "        num_ftrs = temp.fc.in_features\n",
    "        temp.fc = nn.Linear(num_ftrs, 20)\n",
    "        self.model=temp\n",
    "        \n",
    "    def forward(self, x):\n",
    "        x=self.model(x)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "0dd0cb61",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "finish a epoch of training\n",
      "epoch_num: 0\n",
      "train_loss: tensor(0.0805, device='cuda:0')\n",
      "val_loss: tensor(0.0483, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 1\n",
      "train_loss: tensor(0.0672, device='cuda:0')\n",
      "val_loss: tensor(0.0406, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 2\n",
      "train_loss: tensor(0.0593, device='cuda:0')\n",
      "val_loss: tensor(0.0359, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 3\n",
      "train_loss: tensor(0.0541, device='cuda:0')\n",
      "val_loss: tensor(0.0330, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 4\n",
      "train_loss: tensor(0.0504, device='cuda:0')\n",
      "val_loss: tensor(0.0308, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 5\n",
      "train_loss: tensor(0.0478, device='cuda:0')\n",
      "val_loss: tensor(0.0294, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 6\n",
      "train_loss: tensor(0.0463, device='cuda:0')\n",
      "val_loss: tensor(0.0286, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 7\n",
      "train_loss: tensor(0.0444, device='cuda:0')\n",
      "val_loss: tensor(0.0276, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 8\n",
      "train_loss: tensor(0.0431, device='cuda:0')\n",
      "val_loss: tensor(0.0268, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 9\n",
      "train_loss: tensor(0.0417, device='cuda:0')\n",
      "val_loss: tensor(0.0261, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 10\n",
      "train_loss: tensor(0.0412, device='cuda:0')\n",
      "val_loss: tensor(0.0258, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 11\n",
      "train_loss: tensor(0.0402, device='cuda:0')\n",
      "val_loss: tensor(0.0253, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 12\n",
      "train_loss: tensor(0.0394, device='cuda:0')\n",
      "val_loss: tensor(0.0249, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 13\n",
      "train_loss: tensor(0.0387, device='cuda:0')\n",
      "val_loss: tensor(0.0245, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 14\n",
      "train_loss: tensor(0.0383, device='cuda:0')\n",
      "val_loss: tensor(0.0243, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 15\n",
      "train_loss: tensor(0.0380, device='cuda:0')\n",
      "val_loss: tensor(0.0243, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 16\n",
      "train_loss: tensor(0.0373, device='cuda:0')\n",
      "val_loss: tensor(0.0241, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 17\n",
      "train_loss: tensor(0.0371, device='cuda:0')\n",
      "val_loss: tensor(0.0240, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 18\n",
      "train_loss: tensor(0.0366, device='cuda:0')\n",
      "val_loss: tensor(0.0238, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 19\n",
      "train_loss: tensor(0.0362, device='cuda:0')\n",
      "val_loss: tensor(0.0237, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 20\n",
      "train_loss: tensor(0.0359, device='cuda:0')\n",
      "val_loss: tensor(0.0234, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 21\n",
      "train_loss: tensor(0.0357, device='cuda:0')\n",
      "val_loss: tensor(0.0234, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 22\n",
      "train_loss: tensor(0.0351, device='cuda:0')\n",
      "val_loss: tensor(0.0231, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 23\n",
      "train_loss: tensor(0.0351, device='cuda:0')\n",
      "val_loss: tensor(0.0232, device='cuda:0')\n",
      "finish a epoch of training\n",
      "epoch_num: 24\n",
      "train_loss: tensor(0.0345, device='cuda:0')\n",
      "val_loss: tensor(0.0228, device='cuda:0')\n"
     ]
    }
   ],
   "source": [
    "training_address=\"food-101/train.csv\"\n",
    "test_address=\"food-101/test.csv\"\n",
    "train,val,test=get_dataloaders(training_address, test_address, args=None)\n",
    "model=resnet()\n",
    "criterion=nn.CrossEntropyLoss()\n",
    "optimizer=optim.Adam(model.parameters(), lr=0.0001)\n",
    "a=[train,val,test]\n",
    "m1,tl1,tc1,vl1,vc1=train_single_model(model, criterion, optimizer, scheduler=None, device=\"cuda:0\", dataloaders=a,epoch=25)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "431e335d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.plot(tc1, label='Training')\n",
    "plt.plot(vc1, label='Validation')\n",
    "plt.title('Training and Validation Accuracies')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "967df3b2",
   "metadata": {},
   "outputs": [],
   "source": [
    "tl=[]\n",
    "for i in tl1:\n",
    "    tl.append(i.cpu().item())\n",
    "vl=[]\n",
    "for i in vl1:\n",
    "    vl.append(i.cpu().item())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "ff214fc6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(tl, label='Training')\n",
    "plt.plot(vl, label='Validation')\n",
    "plt.title('Training and Validation Losses')\n",
    "plt.xlabel('Epoches')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "d71162fe",
   "metadata": {},
   "outputs": [],
   "source": [
    "model.eval() \n",
    "correct = 0\n",
    "device=\"cuda:0\"\n",
    "with torch.no_grad(): # stop storing gradients for the variables\n",
    "    for data, target in a[2]:\n",
    "        data, target = data.to(device), target.to(device)\n",
    "        output = model(data)\n",
    "        pred = output.argmax(dim=1, keepdim=True) # get the index of maximum fc output. \n",
    "        correct += pred.eq(target.view_as(pred)).sum().item()\n",
    "    val_acc=correct / len(a[2].dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "d1b5cfe7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6678"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "val_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "dbb0a045",
   "metadata": {},
   "outputs": [],
   "source": [
    "torch.save(model.state_dict(), \"food-101/resnet_freeze_full_layer.pt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "da7f811d",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (clean)",
   "language": "python",
   "name": "python3_clean"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}