CIFAR10.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "CIFAR10.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyPSpiVS7On+yQ5ckVukydUI",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/Mithun691/CIFAR10/blob/master/CIFAR10.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-Z1LlRgznREl",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "db14bfe7-935a-48ba-9205-aad6ab2942ab"
      },
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torchvision\n",
        "import torchvision.transforms as transforms\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "if torch.cuda.is_available():\n",
        "  device=torch.device(\"cuda\")\n",
        "else:\n",
        "  device=torch.device(\"cpu\")\n",
        "\n",
        "n_epochs=20\n",
        "batch_size=25\n",
        "learning_rate=0.1\n",
        "\n",
        "transform=transforms.Compose([transforms.ToTensor(),transforms.Normalize((0.5,0.5,0.5),(0.5,0.5,0.5))])\n",
        "train_dataset=torchvision.datasets.CIFAR10(root=\"./data\",train=True,transform=transform,download=True)\n",
        "test_dataset=torchvision.datasets.CIFAR10(root=\"./data\",train=False,transform=transform,download=True)\n",
        "train_loader=torch.utils.data.DataLoader(dataset=train_dataset,batch_size=batch_size,shuffle=True)\n",
        "test_loader=torch.utils.data.DataLoader(dataset=test_dataset,batch_size=batch_size,shuffle=False)\n",
        "\n",
        "map_labels={0:\"plane\",1:\"car\",2:\"bird\",3:\"cat\",4:\"deer\",5:\"dog\",6:\"frog\",7:\"horse\",8:\"ship\",9:\"truck\"}\n",
        "examples=iter(train_loader)\n",
        "samples,labels=examples.next()\n",
        "print(samples.shape,labels.shape)\n",
        "images=samples.numpy()\n",
        "\n",
        "for i in range(25):\n",
        "  plt.subplot(5,5,i+1)\n",
        "  img_view=np.moveaxis(images[i],0,2)\n",
        "  plt.imshow(img_view)\n",
        "\n",
        "for i in range(25):\n",
        "  print(map_labels[labels[i].item()])\n",
        "\n",
        "print(torch.cuda.device_count())\n",
        "print(torch.cuda.get_device_name(0))"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Files already downloaded and verified\n",
            "Files already downloaded and verified\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "torch.Size([25, 3, 32, 32]) torch.Size([25])\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "deer\n",
            "frog\n",
            "frog\n",
            "bird\n",
            "frog\n",
            "bird\n",
            "dog\n",
            "ship\n",
            "frog\n",
            "frog\n",
            "truck\n",
            "cat\n",
            "ship\n",
            "bird\n",
            "cat\n",
            "car\n",
            "horse\n",
            "truck\n",
            "car\n",
            "cat\n",
            "plane\n",
            "horse\n",
            "deer\n",
            "ship\n",
            "plane\n",
            "1\n",
            "Tesla K80\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 25 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OACqFuXNokmB",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "ea3d1230-1d46-4e73-c47e-0699d5f2e6cc"
      },
      "source": [
        "import torch.nn.functional as F\n",
        "\n",
        "class ResNet50(nn.Module):\n",
        "  def __init__(self):\n",
        "    super(ResNet50,self).__init__()\n",
        "    self.conv1a=nn.Conv2d(3,96,3)\n",
        "    self.bn1a=nn.BatchNorm2d(96)\n",
        "    self.conv1b=nn.Conv2d(96,96,3,padding=1)\n",
        "    self.bn1b=nn.BatchNorm2d(96)\n",
        "    self.conv1c=nn.Conv2d(96,96,3,padding=1)\n",
        "    self.bn1c=nn.BatchNorm2d(96)\n",
        "    self.conv1d=nn.Conv2d(96,96,3,padding=1)\n",
        "    self.bn1d=nn.BatchNorm2d(96)\n",
        "    self.conv1e=nn.Conv2d(96,96,3,padding=1)\n",
        "    self.bn1e=nn.BatchNorm2d(96)\n",
        "    self.conv1f=nn.Conv2d(96,96,3,padding=1)\n",
        "    self.bn1f=nn.BatchNorm2d(96)\n",
        "    self.conv1g=nn.Conv2d(96,96,3,padding=1)\n",
        "    self.bn1g=nn.BatchNorm2d(96)\n",
        "    self.mp1=nn.MaxPool2d(2,2)\n",
        "    self.conv2a=nn.Conv2d(96,192,3,padding=1)\n",
        "    self.bn2a=nn.BatchNorm2d(192)\n",
        "    self.conv2b=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2b=nn.BatchNorm2d(192)\n",
        "    self.conv2c=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2c=nn.BatchNorm2d(192)\n",
        "    self.conv2d=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2d=nn.BatchNorm2d(192)\n",
        "    self.conv2e=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2e=nn.BatchNorm2d(192)\n",
        "    self.conv2f=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2f=nn.BatchNorm2d(192)\n",
        "    self.conv2g=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2g=nn.BatchNorm2d(192)\n",
        "    self.conv2h=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2h=nn.BatchNorm2d(192)\n",
        "    self.conv2i=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2i=nn.BatchNorm2d(192)\n",
        "    self.conv2j=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2j=nn.BatchNorm2d(192)\n",
        "    self.conv2k=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2k=nn.BatchNorm2d(192)\n",
        "    self.conv2l=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2l=nn.BatchNorm2d(192)\n",
        "    self.conv2m=nn.Conv2d(192,192,3,padding=1)\n",
        "    self.bn2m=nn.BatchNorm2d(192)\n",
        "    self.mp2=nn.MaxPool2d(2,2)\n",
        "    self.conv3a=nn.Conv2d(192,256,3,padding=1)\n",
        "    self.bn3a=nn.BatchNorm2d(256)\n",
        "    self.conv3b=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3b=nn.BatchNorm2d(256)\n",
        "    self.conv3c=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3c=nn.BatchNorm2d(256)\n",
        "    self.conv3d=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3d=nn.BatchNorm2d(256)\n",
        "    self.conv3e=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3e=nn.BatchNorm2d(256)\n",
        "    self.conv3f=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3f=nn.BatchNorm2d(256)\n",
        "    self.conv3g=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3g=nn.BatchNorm2d(256)\n",
        "    self.conv3h=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3h=nn.BatchNorm2d(256)\n",
        "    self.conv3i=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3i=nn.BatchNorm2d(256)\n",
        "    self.conv3j=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3j=nn.BatchNorm2d(256)\n",
        "    self.conv3k=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3k=nn.BatchNorm2d(256)\n",
        "    self.conv3l=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3l=nn.BatchNorm2d(256)\n",
        "    self.conv3m=nn.Conv2d(256,256,3,padding=1)\n",
        "    self.bn3m=nn.BatchNorm2d(256)\n",
        "    self.mp3=nn.MaxPool2d(2,2)\n",
        "    self.conv4a=nn.Conv2d(256,256,3)\n",
        "    self.conv4b=nn.Conv2d(256,192,1)\n",
        "    self.conv4c=nn.Conv2d(192,10,1)\n",
        "    \n",
        "  def forward(self,x):\n",
        "    conv1a=self.bn1a(F.relu(self.conv1a(x)))\n",
        "    conv1b=self.bn1b(F.relu(self.conv1b(conv1a)))\n",
        "    conv1c=self.bn1c(F.relu(self.conv1c(conv1b)))\n",
        "    conv1d=self.bn1d(F.relu(self.conv1d(conv1c)))+conv1a\n",
        "    conv1e=self.bn1e(F.relu(self.conv1e(conv1d)))\n",
        "    conv1f=self.bn1f(F.relu(self.conv1f(conv1e)))+conv1c\n",
        "    conv1g=self.bn1g(F.relu(self.conv1g(conv1f)))\n",
        "    mp1=self.mp1(conv1g+conv1e)\n",
        "    conv2a=self.bn2a(F.relu(self.conv2a(mp1)))\n",
        "    conv2b=self.bn2b(F.relu(self.conv2b(conv2a)))\n",
        "    conv2c=self.bn2c(F.relu(self.conv2c(conv2b)))\n",
        "    conv2d=self.bn2d(F.relu(self.conv2d(conv2c)))+conv2a\n",
        "    conv2e=self.bn2e(F.relu(self.conv2e(conv2d)))\n",
        "    conv2f=self.bn2f(F.relu(self.conv2f(conv2e)))+conv2c\n",
        "    conv2g=self.bn2g(F.relu(self.conv2g(conv2f)))\n",
        "    conv2h=self.bn2h(F.relu(self.conv2h(conv2g)))+conv2e\n",
        "    conv2i=self.bn2i(F.relu(self.conv2i(conv2h)))\n",
        "    conv2j=self.bn2j(F.relu(self.conv2j(conv2i)))+conv2g\n",
        "    conv2k=self.bn2k(F.relu(self.conv2k(conv2j)))\n",
        "    conv2l=self.bn2l(F.relu(self.conv2l(conv2k)))+conv2i\n",
        "    conv2m=self.bn2m(F.relu(self.conv2m(conv2l)))\n",
        "    mp2=self.mp2(conv2m+conv2k)\n",
        "    conv3a=self.bn3a(F.relu(self.conv3a(mp2)))\n",
        "    conv3b=self.bn3b(F.relu(self.conv3b(conv3a)))\n",
        "    conv3c=self.bn3c(F.relu(self.conv3c(conv3b)))\n",
        "    conv3d=self.bn3d(F.relu(self.conv3d(conv3c)))+conv3a\n",
        "    conv3e=self.bn3e(F.relu(self.conv3e(conv3d)))\n",
        "    conv3f=self.bn3f(F.relu(self.conv3f(conv3e)))+conv3c\n",
        "    conv3g=self.bn3g(F.relu(self.conv3g(conv3f)))\n",
        "    conv3h=self.bn3h(F.relu(self.conv3h(conv3g)))+conv3e\n",
        "    conv3i=self.bn3i(F.relu(self.conv3i(conv3h)))\n",
        "    conv3j=self.bn3j(F.relu(self.conv3j(conv3i)))+conv3g\n",
        "    conv3k=self.bn3k(F.relu(self.conv3k(conv3j)))\n",
        "    conv3l=self.bn3l(F.relu(self.conv3l(conv3k)))+conv3i\n",
        "    conv3m=self.bn3m(F.relu(self.conv3m(conv3l)))\n",
        "    mp3=self.mp3(conv3m+conv3k)\n",
        "    conv4a=F.relu(self.conv4a(mp3))\n",
        "    conv4b=F.relu(self.conv4b(conv4a))\n",
        "    conv4c=F.relu(self.conv4c(conv4b))\n",
        "    out=torch.mean(conv4c.view(conv4c.size(0),conv4c.size(1),-1),dim=2)\n",
        "    return out\n",
        "\n",
        "model=ResNet50()\n",
        "\n",
        "loss=nn.CrossEntropyLoss()\n",
        "optimizer=torch.optim.SGD(model.parameters(),lr=learning_rate)\n",
        "\n",
        "\n",
        "for epoch in range(n_epochs):\n",
        "  for i,(features,labels) in enumerate(train_loader):\n",
        "    labels_pred=model(features)\n",
        "    l=loss(labels_pred,labels)\n",
        "    \n",
        "    l.backward()\n",
        "    optimizer.step()\n",
        "    optimizer.zero_grad()\n",
        "    \n",
        "    if(i%10==9):\n",
        "      print(epoch,i,l)\n",
        "  \n",
        "  Filename=\"model_\"+str(epoch)+\".pth\"\n",
        "  torch.save(model.state_dict(),Filename)\n",
        "\n",
        "  with torch.no_grad():\n",
        "    n_samples=0\n",
        "    n_correct=0\n",
        "    for features,labels in test_loader:\n",
        "      pred=model(features)\n",
        "      _,predictions=torch.max(pred,1)\n",
        "      n_correct+=(predictions==labels).sum().item()\n",
        "      n_samples+=labels.shape[0]\n",
        "    acc=100.0*n_correct/n_samples\n",
        "    print(acc)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0 9 tensor(2.2898, grad_fn=<NllLossBackward>)\n",
            "0 19 tensor(2.2169, grad_fn=<NllLossBackward>)\n",
            "0 29 tensor(2.2969, grad_fn=<NllLossBackward>)\n",
            "0 39 tensor(2.3563, grad_fn=<NllLossBackward>)\n",
            "0 49 tensor(2.2509, grad_fn=<NllLossBackward>)\n",
            "0 59 tensor(2.2807, grad_fn=<NllLossBackward>)\n",
            "0 69 tensor(2.3016, grad_fn=<NllLossBackward>)\n",
            "0 79 tensor(2.2453, grad_fn=<NllLossBackward>)\n",
            "0 89 tensor(2.2344, grad_fn=<NllLossBackward>)\n",
            "0 99 tensor(2.2987, grad_fn=<NllLossBackward>)\n",
            "0 109 tensor(2.1148, grad_fn=<NllLossBackward>)\n",
            "0 119 tensor(2.5845, grad_fn=<NllLossBackward>)\n",
            "0 129 tensor(2.1634, grad_fn=<NllLossBackward>)\n",
            "0 139 tensor(2.2964, grad_fn=<NllLossBackward>)\n",
            "0 149 tensor(2.0789, grad_fn=<NllLossBackward>)\n",
            "0 159 tensor(2.1932, grad_fn=<NllLossBackward>)\n",
            "0 169 tensor(2.3116, grad_fn=<NllLossBackward>)\n",
            "0 179 tensor(2.0928, grad_fn=<NllLossBackward>)\n",
            "0 189 tensor(2.1235, grad_fn=<NllLossBackward>)\n",
            "0 199 tensor(1.9550, grad_fn=<NllLossBackward>)\n",
            "0 209 tensor(2.2161, grad_fn=<NllLossBackward>)\n",
            "0 219 tensor(2.1950, grad_fn=<NllLossBackward>)\n",
            "0 229 tensor(2.0577, grad_fn=<NllLossBackward>)\n",
            "0 239 tensor(1.8848, grad_fn=<NllLossBackward>)\n",
            "0 249 tensor(2.2318, grad_fn=<NllLossBackward>)\n",
            "0 259 tensor(2.0453, grad_fn=<NllLossBackward>)\n",
            "0 269 tensor(2.1615, grad_fn=<NllLossBackward>)\n",
            "0 279 tensor(2.0344, grad_fn=<NllLossBackward>)\n",
            "0 289 tensor(2.1094, grad_fn=<NllLossBackward>)\n",
            "0 299 tensor(2.0664, grad_fn=<NllLossBackward>)\n",
            "0 309 tensor(1.9139, grad_fn=<NllLossBackward>)\n",
            "0 319 tensor(2.0214, grad_fn=<NllLossBackward>)\n",
            "0 329 tensor(1.9735, grad_fn=<NllLossBackward>)\n",
            "0 339 tensor(2.0110, grad_fn=<NllLossBackward>)\n",
            "0 349 tensor(2.2064, grad_fn=<NllLossBackward>)\n",
            "0 359 tensor(2.1797, grad_fn=<NllLossBackward>)\n",
            "0 369 tensor(2.1022, grad_fn=<NllLossBackward>)\n",
            "0 379 tensor(1.9233, grad_fn=<NllLossBackward>)\n",
            "0 389 tensor(1.7567, grad_fn=<NllLossBackward>)\n",
            "0 399 tensor(1.7399, grad_fn=<NllLossBackward>)\n",
            "0 409 tensor(1.9253, grad_fn=<NllLossBackward>)\n",
            "0 419 tensor(2.3118, grad_fn=<NllLossBackward>)\n",
            "0 429 tensor(2.0980, grad_fn=<NllLossBackward>)\n",
            "0 439 tensor(2.0921, grad_fn=<NllLossBackward>)\n",
            "0 449 tensor(1.9285, grad_fn=<NllLossBackward>)\n",
            "0 459 tensor(2.0305, grad_fn=<NllLossBackward>)\n",
            "0 469 tensor(1.7541, grad_fn=<NllLossBackward>)\n",
            "0 479 tensor(1.7363, grad_fn=<NllLossBackward>)\n",
            "0 489 tensor(1.8794, grad_fn=<NllLossBackward>)\n",
            "0 499 tensor(1.7860, grad_fn=<NllLossBackward>)\n",
            "0 509 tensor(2.5018, grad_fn=<NllLossBackward>)\n",
            "0 519 tensor(1.8544, grad_fn=<NllLossBackward>)\n",
            "0 529 tensor(1.9368, grad_fn=<NllLossBackward>)\n",
            "0 539 tensor(1.9239, grad_fn=<NllLossBackward>)\n",
            "0 549 tensor(1.9613, grad_fn=<NllLossBackward>)\n",
            "0 559 tensor(2.0493, grad_fn=<NllLossBackward>)\n",
            "0 569 tensor(1.9754, grad_fn=<NllLossBackward>)\n",
            "0 579 tensor(1.4707, grad_fn=<NllLossBackward>)\n",
            "0 589 tensor(2.0023, grad_fn=<NllLossBackward>)\n",
            "0 599 tensor(1.6399, grad_fn=<NllLossBackward>)\n",
            "0 609 tensor(1.9024, grad_fn=<NllLossBackward>)\n",
            "0 619 tensor(1.9787, grad_fn=<NllLossBackward>)\n",
            "0 629 tensor(1.8340, grad_fn=<NllLossBackward>)\n",
            "0 639 tensor(2.1347, grad_fn=<NllLossBackward>)\n",
            "0 649 tensor(1.8256, grad_fn=<NllLossBackward>)\n",
            "0 659 tensor(1.7569, grad_fn=<NllLossBackward>)\n",
            "0 669 tensor(1.6097, grad_fn=<NllLossBackward>)\n",
            "0 679 tensor(1.8054, grad_fn=<NllLossBackward>)\n",
            "0 689 tensor(1.8414, grad_fn=<NllLossBackward>)\n",
            "0 699 tensor(1.7612, grad_fn=<NllLossBackward>)\n",
            "0 709 tensor(1.7517, grad_fn=<NllLossBackward>)\n",
            "0 719 tensor(2.0057, grad_fn=<NllLossBackward>)\n",
            "0 729 tensor(1.7574, grad_fn=<NllLossBackward>)\n",
            "0 739 tensor(1.4037, grad_fn=<NllLossBackward>)\n",
            "0 749 tensor(2.0782, grad_fn=<NllLossBackward>)\n",
            "0 759 tensor(1.8652, grad_fn=<NllLossBackward>)\n",
            "0 769 tensor(2.0098, grad_fn=<NllLossBackward>)\n",
            "0 779 tensor(1.8566, grad_fn=<NllLossBackward>)\n",
            "0 789 tensor(1.4729, grad_fn=<NllLossBackward>)\n",
            "0 799 tensor(1.9444, grad_fn=<NllLossBackward>)\n",
            "0 809 tensor(1.8718, grad_fn=<NllLossBackward>)\n",
            "0 819 tensor(1.7962, grad_fn=<NllLossBackward>)\n",
            "0 829 tensor(1.8187, grad_fn=<NllLossBackward>)\n",
            "0 839 tensor(1.5688, grad_fn=<NllLossBackward>)\n",
            "0 849 tensor(1.7920, grad_fn=<NllLossBackward>)\n",
            "0 859 tensor(1.7387, grad_fn=<NllLossBackward>)\n",
            "0 869 tensor(2.0036, grad_fn=<NllLossBackward>)\n",
            "0 879 tensor(1.8357, grad_fn=<NllLossBackward>)\n",
            "0 889 tensor(1.8325, grad_fn=<NllLossBackward>)\n",
            "0 899 tensor(1.6095, grad_fn=<NllLossBackward>)\n",
            "0 909 tensor(1.9048, grad_fn=<NllLossBackward>)\n",
            "0 919 tensor(1.7921, grad_fn=<NllLossBackward>)\n",
            "0 929 tensor(1.6611, grad_fn=<NllLossBackward>)\n",
            "0 939 tensor(1.7563, grad_fn=<NllLossBackward>)\n",
            "0 949 tensor(1.8365, grad_fn=<NllLossBackward>)\n",
            "0 959 tensor(1.5085, grad_fn=<NllLossBackward>)\n",
            "0 969 tensor(1.6695, grad_fn=<NllLossBackward>)\n",
            "0 979 tensor(1.5376, grad_fn=<NllLossBackward>)\n",
            "0 989 tensor(2.0953, grad_fn=<NllLossBackward>)\n",
            "0 999 tensor(1.6016, grad_fn=<NllLossBackward>)\n",
            "0 1009 tensor(1.6635, grad_fn=<NllLossBackward>)\n",
            "0 1019 tensor(1.6093, grad_fn=<NllLossBackward>)\n",
            "0 1029 tensor(1.6251, grad_fn=<NllLossBackward>)\n",
            "0 1039 tensor(1.5406, grad_fn=<NllLossBackward>)\n",
            "0 1049 tensor(1.8412, grad_fn=<NllLossBackward>)\n",
            "0 1059 tensor(1.8238, grad_fn=<NllLossBackward>)\n",
            "0 1069 tensor(2.0604, grad_fn=<NllLossBackward>)\n",
            "0 1079 tensor(1.7518, grad_fn=<NllLossBackward>)\n",
            "0 1089 tensor(1.7053, grad_fn=<NllLossBackward>)\n",
            "0 1099 tensor(1.7748, grad_fn=<NllLossBackward>)\n",
            "0 1109 tensor(1.8061, grad_fn=<NllLossBackward>)\n",
            "0 1119 tensor(2.0023, grad_fn=<NllLossBackward>)\n",
            "0 1129 tensor(1.6747, grad_fn=<NllLossBackward>)\n",
            "0 1139 tensor(1.9992, grad_fn=<NllLossBackward>)\n",
            "0 1149 tensor(1.7422, grad_fn=<NllLossBackward>)\n",
            "0 1159 tensor(1.8162, grad_fn=<NllLossBackward>)\n",
            "0 1169 tensor(1.8021, grad_fn=<NllLossBackward>)\n",
            "0 1179 tensor(1.3702, grad_fn=<NllLossBackward>)\n",
            "0 1189 tensor(1.4543, grad_fn=<NllLossBackward>)\n",
            "0 1199 tensor(1.7087, grad_fn=<NllLossBackward>)\n",
            "0 1209 tensor(1.8675, grad_fn=<NllLossBackward>)\n",
            "0 1219 tensor(1.3627, grad_fn=<NllLossBackward>)\n",
            "0 1229 tensor(1.6743, grad_fn=<NllLossBackward>)\n",
            "0 1239 tensor(1.8050, grad_fn=<NllLossBackward>)\n",
            "0 1249 tensor(1.7329, grad_fn=<NllLossBackward>)\n",
            "0 1259 tensor(1.6085, grad_fn=<NllLossBackward>)\n",
            "0 1269 tensor(1.4036, grad_fn=<NllLossBackward>)\n",
            "0 1279 tensor(1.9220, grad_fn=<NllLossBackward>)\n",
            "0 1289 tensor(1.3624, grad_fn=<NllLossBackward>)\n",
            "0 1299 tensor(1.7157, grad_fn=<NllLossBackward>)\n",
            "0 1309 tensor(1.6292, grad_fn=<NllLossBackward>)\n",
            "0 1319 tensor(1.5495, grad_fn=<NllLossBackward>)\n",
            "0 1329 tensor(1.3412, grad_fn=<NllLossBackward>)\n",
            "0 1339 tensor(1.6841, grad_fn=<NllLossBackward>)\n",
            "0 1349 tensor(1.5163, grad_fn=<NllLossBackward>)\n",
            "0 1359 tensor(1.6264, grad_fn=<NllLossBackward>)\n",
            "0 1369 tensor(1.3927, grad_fn=<NllLossBackward>)\n",
            "0 1379 tensor(1.6061, grad_fn=<NllLossBackward>)\n",
            "0 1389 tensor(1.5990, grad_fn=<NllLossBackward>)\n",
            "0 1399 tensor(1.5313, grad_fn=<NllLossBackward>)\n",
            "0 1409 tensor(1.4382, grad_fn=<NllLossBackward>)\n",
            "0 1419 tensor(1.5507, grad_fn=<NllLossBackward>)\n",
            "0 1429 tensor(1.4352, grad_fn=<NllLossBackward>)\n",
            "0 1439 tensor(1.6536, grad_fn=<NllLossBackward>)\n",
            "0 1449 tensor(1.9097, grad_fn=<NllLossBackward>)\n",
            "0 1459 tensor(1.8425, grad_fn=<NllLossBackward>)\n",
            "0 1469 tensor(1.6071, grad_fn=<NllLossBackward>)\n",
            "0 1479 tensor(1.7081, grad_fn=<NllLossBackward>)\n",
            "0 1489 tensor(1.4120, grad_fn=<NllLossBackward>)\n",
            "0 1499 tensor(1.5923, grad_fn=<NllLossBackward>)\n",
            "0 1509 tensor(1.6725, grad_fn=<NllLossBackward>)\n",
            "0 1519 tensor(1.7337, grad_fn=<NllLossBackward>)\n",
            "0 1529 tensor(1.6699, grad_fn=<NllLossBackward>)\n",
            "0 1539 tensor(1.8015, grad_fn=<NllLossBackward>)\n",
            "0 1549 tensor(1.3600, grad_fn=<NllLossBackward>)\n",
            "0 1559 tensor(1.4370, grad_fn=<NllLossBackward>)\n",
            "0 1569 tensor(1.0905, grad_fn=<NllLossBackward>)\n",
            "0 1579 tensor(1.6642, grad_fn=<NllLossBackward>)\n",
            "0 1589 tensor(1.4796, grad_fn=<NllLossBackward>)\n",
            "0 1599 tensor(1.5926, grad_fn=<NllLossBackward>)\n",
            "0 1609 tensor(1.3910, grad_fn=<NllLossBackward>)\n",
            "0 1619 tensor(1.8170, grad_fn=<NllLossBackward>)\n",
            "0 1629 tensor(1.6007, grad_fn=<NllLossBackward>)\n",
            "0 1639 tensor(1.4821, grad_fn=<NllLossBackward>)\n",
            "0 1649 tensor(1.2020, grad_fn=<NllLossBackward>)\n",
            "0 1659 tensor(1.5673, grad_fn=<NllLossBackward>)\n",
            "0 1669 tensor(1.3165, grad_fn=<NllLossBackward>)\n",
            "0 1679 tensor(1.4133, grad_fn=<NllLossBackward>)\n",
            "0 1689 tensor(1.4789, grad_fn=<NllLossBackward>)\n",
            "0 1699 tensor(1.3400, grad_fn=<NllLossBackward>)\n",
            "0 1709 tensor(1.3076, grad_fn=<NllLossBackward>)\n",
            "0 1719 tensor(2.2213, grad_fn=<NllLossBackward>)\n",
            "0 1729 tensor(1.5171, grad_fn=<NllLossBackward>)\n",
            "0 1739 tensor(1.6294, grad_fn=<NllLossBackward>)\n",
            "0 1749 tensor(1.4319, grad_fn=<NllLossBackward>)\n",
            "0 1759 tensor(1.5074, grad_fn=<NllLossBackward>)\n",
            "0 1769 tensor(1.6470, grad_fn=<NllLossBackward>)\n",
            "0 1779 tensor(1.0760, grad_fn=<NllLossBackward>)\n",
            "0 1789 tensor(1.5381, grad_fn=<NllLossBackward>)\n",
            "0 1799 tensor(1.5345, grad_fn=<NllLossBackward>)\n",
            "0 1809 tensor(2.1370, grad_fn=<NllLossBackward>)\n",
            "0 1819 tensor(1.6426, grad_fn=<NllLossBackward>)\n",
            "0 1829 tensor(1.3253, grad_fn=<NllLossBackward>)\n",
            "0 1839 tensor(1.3741, grad_fn=<NllLossBackward>)\n",
            "0 1849 tensor(1.4138, grad_fn=<NllLossBackward>)\n",
            "0 1859 tensor(1.4265, grad_fn=<NllLossBackward>)\n",
            "0 1869 tensor(1.8263, grad_fn=<NllLossBackward>)\n",
            "0 1879 tensor(1.5706, grad_fn=<NllLossBackward>)\n",
            "0 1889 tensor(1.5689, grad_fn=<NllLossBackward>)\n",
            "0 1899 tensor(1.5568, grad_fn=<NllLossBackward>)\n",
            "0 1909 tensor(2.0519, grad_fn=<NllLossBackward>)\n",
            "0 1919 tensor(1.7130, grad_fn=<NllLossBackward>)\n",
            "0 1929 tensor(1.4368, grad_fn=<NllLossBackward>)\n",
            "0 1939 tensor(1.4617, grad_fn=<NllLossBackward>)\n",
            "0 1949 tensor(1.2811, grad_fn=<NllLossBackward>)\n",
            "0 1959 tensor(0.9758, grad_fn=<NllLossBackward>)\n",
            "0 1969 tensor(1.8409, grad_fn=<NllLossBackward>)\n",
            "0 1979 tensor(0.9265, grad_fn=<NllLossBackward>)\n",
            "0 1989 tensor(1.5985, grad_fn=<NllLossBackward>)\n",
            "0 1999 tensor(1.3738, grad_fn=<NllLossBackward>)\n",
            "44.84\n",
            "1 9 tensor(0.9221, grad_fn=<NllLossBackward>)\n",
            "1 19 tensor(1.1033, grad_fn=<NllLossBackward>)\n",
            "1 29 tensor(1.5446, grad_fn=<NllLossBackward>)\n",
            "1 39 tensor(1.2264, grad_fn=<NllLossBackward>)\n",
            "1 49 tensor(1.3883, grad_fn=<NllLossBackward>)\n",
            "1 59 tensor(1.5523, grad_fn=<NllLossBackward>)\n",
            "1 69 tensor(1.5843, grad_fn=<NllLossBackward>)\n",
            "1 79 tensor(1.6665, grad_fn=<NllLossBackward>)\n",
            "1 89 tensor(1.5086, grad_fn=<NllLossBackward>)\n",
            "1 99 tensor(1.2917, grad_fn=<NllLossBackward>)\n",
            "1 109 tensor(1.1421, grad_fn=<NllLossBackward>)\n",
            "1 119 tensor(1.5326, grad_fn=<NllLossBackward>)\n",
            "1 129 tensor(1.7220, grad_fn=<NllLossBackward>)\n",
            "1 139 tensor(1.7164, grad_fn=<NllLossBackward>)\n",
            "1 149 tensor(1.5238, grad_fn=<NllLossBackward>)\n",
            "1 159 tensor(1.6432, grad_fn=<NllLossBackward>)\n",
            "1 169 tensor(1.4342, grad_fn=<NllLossBackward>)\n",
            "1 179 tensor(1.3347, grad_fn=<NllLossBackward>)\n",
            "1 189 tensor(1.3021, grad_fn=<NllLossBackward>)\n",
            "1 199 tensor(1.6617, grad_fn=<NllLossBackward>)\n",
            "1 209 tensor(1.4714, grad_fn=<NllLossBackward>)\n",
            "1 219 tensor(1.5914, grad_fn=<NllLossBackward>)\n",
            "1 229 tensor(1.4677, grad_fn=<NllLossBackward>)\n",
            "1 239 tensor(1.4642, grad_fn=<NllLossBackward>)\n",
            "1 249 tensor(1.3100, grad_fn=<NllLossBackward>)\n",
            "1 259 tensor(1.4425, grad_fn=<NllLossBackward>)\n",
            "1 269 tensor(1.1471, grad_fn=<NllLossBackward>)\n",
            "1 279 tensor(1.2851, grad_fn=<NllLossBackward>)\n",
            "1 289 tensor(1.3760, grad_fn=<NllLossBackward>)\n",
            "1 299 tensor(1.1429, grad_fn=<NllLossBackward>)\n",
            "1 309 tensor(1.3067, grad_fn=<NllLossBackward>)\n",
            "1 319 tensor(1.2910, grad_fn=<NllLossBackward>)\n",
            "1 329 tensor(1.3919, grad_fn=<NllLossBackward>)\n",
            "1 339 tensor(1.6971, grad_fn=<NllLossBackward>)\n",
            "1 349 tensor(1.5578, grad_fn=<NllLossBackward>)\n",
            "1 359 tensor(1.6589, grad_fn=<NllLossBackward>)\n",
            "1 369 tensor(1.3461, grad_fn=<NllLossBackward>)\n",
            "1 379 tensor(1.4991, grad_fn=<NllLossBackward>)\n",
            "1 389 tensor(1.0470, grad_fn=<NllLossBackward>)\n",
            "1 399 tensor(2.0029, grad_fn=<NllLossBackward>)\n",
            "1 409 tensor(1.5939, grad_fn=<NllLossBackward>)\n",
            "1 419 tensor(1.2596, grad_fn=<NllLossBackward>)\n",
            "1 429 tensor(1.2822, grad_fn=<NllLossBackward>)\n",
            "1 439 tensor(1.2401, grad_fn=<NllLossBackward>)\n",
            "1 449 tensor(1.6220, grad_fn=<NllLossBackward>)\n",
            "1 459 tensor(1.4799, grad_fn=<NllLossBackward>)\n",
            "1 469 tensor(1.1721, grad_fn=<NllLossBackward>)\n",
            "1 479 tensor(1.3612, grad_fn=<NllLossBackward>)\n",
            "1 489 tensor(1.4567, grad_fn=<NllLossBackward>)\n",
            "1 499 tensor(1.2240, grad_fn=<NllLossBackward>)\n",
            "1 509 tensor(1.6157, grad_fn=<NllLossBackward>)\n",
            "1 519 tensor(1.6203, grad_fn=<NllLossBackward>)\n",
            "1 529 tensor(1.4352, grad_fn=<NllLossBackward>)\n",
            "1 539 tensor(1.5475, grad_fn=<NllLossBackward>)\n",
            "1 549 tensor(1.2161, grad_fn=<NllLossBackward>)\n",
            "1 559 tensor(1.2851, grad_fn=<NllLossBackward>)\n",
            "1 569 tensor(1.2902, grad_fn=<NllLossBackward>)\n",
            "1 579 tensor(1.1847, grad_fn=<NllLossBackward>)\n",
            "1 589 tensor(1.2205, grad_fn=<NllLossBackward>)\n",
            "1 599 tensor(1.6926, grad_fn=<NllLossBackward>)\n",
            "1 609 tensor(1.4737, grad_fn=<NllLossBackward>)\n",
            "1 619 tensor(1.0540, grad_fn=<NllLossBackward>)\n",
            "1 629 tensor(1.8031, grad_fn=<NllLossBackward>)\n",
            "1 639 tensor(1.3638, grad_fn=<NllLossBackward>)\n",
            "1 649 tensor(1.2504, grad_fn=<NllLossBackward>)\n",
            "1 659 tensor(1.1062, grad_fn=<NllLossBackward>)\n",
            "1 669 tensor(1.7431, grad_fn=<NllLossBackward>)\n",
            "1 679 tensor(1.2810, grad_fn=<NllLossBackward>)\n",
            "1 689 tensor(1.4076, grad_fn=<NllLossBackward>)\n",
            "1 699 tensor(1.3301, grad_fn=<NllLossBackward>)\n",
            "1 709 tensor(1.5319, grad_fn=<NllLossBackward>)\n",
            "1 719 tensor(1.5815, grad_fn=<NllLossBackward>)\n",
            "1 729 tensor(1.7322, grad_fn=<NllLossBackward>)\n",
            "1 739 tensor(1.0493, grad_fn=<NllLossBackward>)\n",
            "1 749 tensor(1.3169, grad_fn=<NllLossBackward>)\n",
            "1 759 tensor(1.4067, grad_fn=<NllLossBackward>)\n",
            "1 769 tensor(1.2645, grad_fn=<NllLossBackward>)\n",
            "1 779 tensor(0.9214, grad_fn=<NllLossBackward>)\n",
            "1 789 tensor(1.2013, grad_fn=<NllLossBackward>)\n",
            "1 799 tensor(1.1437, grad_fn=<NllLossBackward>)\n",
            "1 809 tensor(1.5025, grad_fn=<NllLossBackward>)\n",
            "1 819 tensor(1.9667, grad_fn=<NllLossBackward>)\n",
            "1 829 tensor(1.2209, grad_fn=<NllLossBackward>)\n",
            "1 839 tensor(1.2760, grad_fn=<NllLossBackward>)\n",
            "1 849 tensor(1.2587, grad_fn=<NllLossBackward>)\n",
            "1 859 tensor(1.7168, grad_fn=<NllLossBackward>)\n",
            "1 869 tensor(1.1009, grad_fn=<NllLossBackward>)\n",
            "1 879 tensor(1.3630, grad_fn=<NllLossBackward>)\n",
            "1 889 tensor(1.3235, grad_fn=<NllLossBackward>)\n",
            "1 899 tensor(1.1882, grad_fn=<NllLossBackward>)\n",
            "1 909 tensor(1.0726, grad_fn=<NllLossBackward>)\n",
            "1 919 tensor(1.1383, grad_fn=<NllLossBackward>)\n",
            "1 929 tensor(1.7161, grad_fn=<NllLossBackward>)\n",
            "1 939 tensor(1.3867, grad_fn=<NllLossBackward>)\n",
            "1 949 tensor(1.0196, grad_fn=<NllLossBackward>)\n",
            "1 959 tensor(1.1564, grad_fn=<NllLossBackward>)\n",
            "1 969 tensor(1.1899, grad_fn=<NllLossBackward>)\n",
            "1 979 tensor(1.3426, grad_fn=<NllLossBackward>)\n",
            "1 989 tensor(1.2982, grad_fn=<NllLossBackward>)\n",
            "1 999 tensor(1.0892, grad_fn=<NllLossBackward>)\n",
            "1 1009 tensor(1.5314, grad_fn=<NllLossBackward>)\n",
            "1 1019 tensor(1.2051, grad_fn=<NllLossBackward>)\n",
            "1 1029 tensor(1.3791, grad_fn=<NllLossBackward>)\n",
            "1 1039 tensor(1.2682, grad_fn=<NllLossBackward>)\n",
            "1 1049 tensor(1.2644, grad_fn=<NllLossBackward>)\n",
            "1 1059 tensor(0.8938, grad_fn=<NllLossBackward>)\n",
            "1 1069 tensor(1.4352, grad_fn=<NllLossBackward>)\n",
            "1 1079 tensor(1.4226, grad_fn=<NllLossBackward>)\n",
            "1 1089 tensor(1.2571, grad_fn=<NllLossBackward>)\n",
            "1 1099 tensor(1.2618, grad_fn=<NllLossBackward>)\n",
            "1 1109 tensor(1.3224, grad_fn=<NllLossBackward>)\n",
            "1 1119 tensor(1.5727, grad_fn=<NllLossBackward>)\n",
            "1 1129 tensor(1.3105, grad_fn=<NllLossBackward>)\n",
            "1 1139 tensor(1.1326, grad_fn=<NllLossBackward>)\n",
            "1 1149 tensor(1.1866, grad_fn=<NllLossBackward>)\n",
            "1 1159 tensor(1.3630, grad_fn=<NllLossBackward>)\n",
            "1 1169 tensor(1.1453, grad_fn=<NllLossBackward>)\n",
            "1 1179 tensor(1.4184, grad_fn=<NllLossBackward>)\n",
            "1 1189 tensor(1.0520, grad_fn=<NllLossBackward>)\n",
            "1 1199 tensor(1.0162, grad_fn=<NllLossBackward>)\n",
            "1 1209 tensor(1.0244, grad_fn=<NllLossBackward>)\n",
            "1 1219 tensor(1.1375, grad_fn=<NllLossBackward>)\n",
            "1 1229 tensor(1.1665, grad_fn=<NllLossBackward>)\n",
            "1 1239 tensor(1.2832, grad_fn=<NllLossBackward>)\n",
            "1 1249 tensor(1.1552, grad_fn=<NllLossBackward>)\n",
            "1 1259 tensor(1.5185, grad_fn=<NllLossBackward>)\n",
            "1 1269 tensor(1.2141, grad_fn=<NllLossBackward>)\n",
            "1 1279 tensor(1.6757, grad_fn=<NllLossBackward>)\n",
            "1 1289 tensor(1.3755, grad_fn=<NllLossBackward>)\n",
            "1 1299 tensor(1.2917, grad_fn=<NllLossBackward>)\n",
            "1 1309 tensor(1.5009, grad_fn=<NllLossBackward>)\n",
            "1 1319 tensor(1.1052, grad_fn=<NllLossBackward>)\n",
            "1 1329 tensor(0.9686, grad_fn=<NllLossBackward>)\n",
            "1 1339 tensor(1.3920, grad_fn=<NllLossBackward>)\n",
            "1 1349 tensor(1.1468, grad_fn=<NllLossBackward>)\n",
            "1 1359 tensor(1.2406, grad_fn=<NllLossBackward>)\n",
            "1 1369 tensor(1.3166, grad_fn=<NllLossBackward>)\n",
            "1 1379 tensor(1.3986, grad_fn=<NllLossBackward>)\n",
            "1 1389 tensor(1.4358, grad_fn=<NllLossBackward>)\n",
            "1 1399 tensor(0.9227, grad_fn=<NllLossBackward>)\n",
            "1 1409 tensor(1.5163, grad_fn=<NllLossBackward>)\n",
            "1 1419 tensor(1.4156, grad_fn=<NllLossBackward>)\n",
            "1 1429 tensor(1.3073, grad_fn=<NllLossBackward>)\n",
            "1 1439 tensor(1.2569, grad_fn=<NllLossBackward>)\n",
            "1 1449 tensor(1.4181, grad_fn=<NllLossBackward>)\n",
            "1 1459 tensor(1.3745, grad_fn=<NllLossBackward>)\n",
            "1 1469 tensor(0.9568, grad_fn=<NllLossBackward>)\n",
            "1 1479 tensor(1.2453, grad_fn=<NllLossBackward>)\n",
            "1 1489 tensor(1.1090, grad_fn=<NllLossBackward>)\n",
            "1 1499 tensor(1.3346, grad_fn=<NllLossBackward>)\n",
            "1 1509 tensor(0.9957, grad_fn=<NllLossBackward>)\n",
            "1 1519 tensor(1.2274, grad_fn=<NllLossBackward>)\n",
            "1 1529 tensor(1.1463, grad_fn=<NllLossBackward>)\n",
            "1 1539 tensor(1.3058, grad_fn=<NllLossBackward>)\n",
            "1 1549 tensor(1.1170, grad_fn=<NllLossBackward>)\n",
            "1 1559 tensor(1.1582, grad_fn=<NllLossBackward>)\n",
            "1 1569 tensor(1.5375, grad_fn=<NllLossBackward>)\n",
            "1 1579 tensor(0.8231, grad_fn=<NllLossBackward>)\n",
            "1 1589 tensor(1.1305, grad_fn=<NllLossBackward>)\n",
            "1 1599 tensor(1.3161, grad_fn=<NllLossBackward>)\n",
            "1 1609 tensor(1.1581, grad_fn=<NllLossBackward>)\n",
            "1 1619 tensor(1.5603, grad_fn=<NllLossBackward>)\n",
            "1 1629 tensor(0.5680, grad_fn=<NllLossBackward>)\n",
            "1 1639 tensor(1.1569, grad_fn=<NllLossBackward>)\n",
            "1 1649 tensor(2.0279, grad_fn=<NllLossBackward>)\n",
            "1 1659 tensor(1.0196, grad_fn=<NllLossBackward>)\n",
            "1 1669 tensor(1.1703, grad_fn=<NllLossBackward>)\n",
            "1 1679 tensor(1.3146, grad_fn=<NllLossBackward>)\n",
            "1 1689 tensor(1.1939, grad_fn=<NllLossBackward>)\n",
            "1 1699 tensor(0.9940, grad_fn=<NllLossBackward>)\n",
            "1 1709 tensor(1.3432, grad_fn=<NllLossBackward>)\n",
            "1 1719 tensor(1.2555, grad_fn=<NllLossBackward>)\n",
            "1 1729 tensor(1.3086, grad_fn=<NllLossBackward>)\n",
            "1 1739 tensor(1.6185, grad_fn=<NllLossBackward>)\n",
            "1 1749 tensor(0.9096, grad_fn=<NllLossBackward>)\n",
            "1 1759 tensor(0.8947, grad_fn=<NllLossBackward>)\n",
            "1 1769 tensor(1.1928, grad_fn=<NllLossBackward>)\n",
            "1 1779 tensor(1.0774, grad_fn=<NllLossBackward>)\n",
            "1 1789 tensor(1.2737, grad_fn=<NllLossBackward>)\n",
            "1 1799 tensor(0.9567, grad_fn=<NllLossBackward>)\n",
            "1 1809 tensor(1.1511, grad_fn=<NllLossBackward>)\n",
            "1 1819 tensor(1.0317, grad_fn=<NllLossBackward>)\n",
            "1 1829 tensor(1.2360, grad_fn=<NllLossBackward>)\n",
            "1 1839 tensor(1.3575, grad_fn=<NllLossBackward>)\n",
            "1 1849 tensor(1.2297, grad_fn=<NllLossBackward>)\n",
            "1 1859 tensor(1.7322, grad_fn=<NllLossBackward>)\n",
            "1 1869 tensor(1.2350, grad_fn=<NllLossBackward>)\n",
            "1 1879 tensor(1.0892, grad_fn=<NllLossBackward>)\n",
            "1 1889 tensor(1.3614, grad_fn=<NllLossBackward>)\n",
            "1 1899 tensor(1.0770, grad_fn=<NllLossBackward>)\n",
            "1 1909 tensor(1.7168, grad_fn=<NllLossBackward>)\n",
            "1 1919 tensor(1.0040, grad_fn=<NllLossBackward>)\n",
            "1 1929 tensor(1.0096, grad_fn=<NllLossBackward>)\n",
            "1 1939 tensor(0.8801, grad_fn=<NllLossBackward>)\n",
            "1 1949 tensor(1.4071, grad_fn=<NllLossBackward>)\n",
            "1 1959 tensor(1.1325, grad_fn=<NllLossBackward>)\n",
            "1 1969 tensor(1.1720, grad_fn=<NllLossBackward>)\n",
            "1 1979 tensor(0.9139, grad_fn=<NllLossBackward>)\n",
            "1 1989 tensor(0.9571, grad_fn=<NllLossBackward>)\n",
            "1 1999 tensor(1.3241, grad_fn=<NllLossBackward>)\n",
            "57.36\n",
            "2 9 tensor(0.7387, grad_fn=<NllLossBackward>)\n",
            "2 19 tensor(1.2028, grad_fn=<NllLossBackward>)\n",
            "2 29 tensor(1.2387, grad_fn=<NllLossBackward>)\n",
            "2 39 tensor(1.0030, grad_fn=<NllLossBackward>)\n",
            "2 49 tensor(1.2724, grad_fn=<NllLossBackward>)\n",
            "2 59 tensor(1.4611, grad_fn=<NllLossBackward>)\n",
            "2 69 tensor(1.1690, grad_fn=<NllLossBackward>)\n",
            "2 79 tensor(0.8454, grad_fn=<NllLossBackward>)\n",
            "2 89 tensor(1.0466, grad_fn=<NllLossBackward>)\n",
            "2 99 tensor(0.9165, grad_fn=<NllLossBackward>)\n",
            "2 109 tensor(1.5478, grad_fn=<NllLossBackward>)\n",
            "2 119 tensor(1.2545, grad_fn=<NllLossBackward>)\n",
            "2 129 tensor(1.1127, grad_fn=<NllLossBackward>)\n",
            "2 139 tensor(1.0492, grad_fn=<NllLossBackward>)\n",
            "2 149 tensor(1.2220, grad_fn=<NllLossBackward>)\n",
            "2 159 tensor(0.8247, grad_fn=<NllLossBackward>)\n",
            "2 169 tensor(1.3875, grad_fn=<NllLossBackward>)\n",
            "2 179 tensor(0.6428, grad_fn=<NllLossBackward>)\n",
            "2 189 tensor(1.1965, grad_fn=<NllLossBackward>)\n",
            "2 199 tensor(1.0212, grad_fn=<NllLossBackward>)\n",
            "2 209 tensor(1.2835, grad_fn=<NllLossBackward>)\n",
            "2 219 tensor(1.1075, grad_fn=<NllLossBackward>)\n",
            "2 229 tensor(0.8080, grad_fn=<NllLossBackward>)\n",
            "2 239 tensor(0.9230, grad_fn=<NllLossBackward>)\n",
            "2 249 tensor(0.9534, grad_fn=<NllLossBackward>)\n",
            "2 259 tensor(0.8893, grad_fn=<NllLossBackward>)\n",
            "2 269 tensor(0.9007, grad_fn=<NllLossBackward>)\n",
            "2 279 tensor(0.8961, grad_fn=<NllLossBackward>)\n",
            "2 289 tensor(1.6660, grad_fn=<NllLossBackward>)\n",
            "2 299 tensor(0.9774, grad_fn=<NllLossBackward>)\n",
            "2 309 tensor(1.4052, grad_fn=<NllLossBackward>)\n",
            "2 319 tensor(1.2985, grad_fn=<NllLossBackward>)\n",
            "2 329 tensor(1.0280, grad_fn=<NllLossBackward>)\n",
            "2 339 tensor(1.2879, grad_fn=<NllLossBackward>)\n",
            "2 349 tensor(1.1859, grad_fn=<NllLossBackward>)\n",
            "2 359 tensor(1.2630, grad_fn=<NllLossBackward>)\n",
            "2 369 tensor(1.0518, grad_fn=<NllLossBackward>)\n",
            "2 379 tensor(0.8048, grad_fn=<NllLossBackward>)\n",
            "2 389 tensor(1.0497, grad_fn=<NllLossBackward>)\n",
            "2 399 tensor(0.8202, grad_fn=<NllLossBackward>)\n",
            "2 409 tensor(1.4307, grad_fn=<NllLossBackward>)\n",
            "2 419 tensor(1.1417, grad_fn=<NllLossBackward>)\n",
            "2 429 tensor(1.2552, grad_fn=<NllLossBackward>)\n",
            "2 439 tensor(0.9427, grad_fn=<NllLossBackward>)\n",
            "2 449 tensor(0.8791, grad_fn=<NllLossBackward>)\n",
            "2 459 tensor(0.9981, grad_fn=<NllLossBackward>)\n",
            "2 469 tensor(0.7703, grad_fn=<NllLossBackward>)\n",
            "2 479 tensor(1.2799, grad_fn=<NllLossBackward>)\n",
            "2 489 tensor(1.0785, grad_fn=<NllLossBackward>)\n",
            "2 499 tensor(1.2554, grad_fn=<NllLossBackward>)\n",
            "2 509 tensor(0.9445, grad_fn=<NllLossBackward>)\n",
            "2 519 tensor(1.0662, grad_fn=<NllLossBackward>)\n",
            "2 529 tensor(1.0401, grad_fn=<NllLossBackward>)\n",
            "2 539 tensor(0.8647, grad_fn=<NllLossBackward>)\n",
            "2 549 tensor(1.1142, grad_fn=<NllLossBackward>)\n",
            "2 559 tensor(1.0636, grad_fn=<NllLossBackward>)\n",
            "2 569 tensor(1.0796, grad_fn=<NllLossBackward>)\n",
            "2 579 tensor(1.1203, grad_fn=<NllLossBackward>)\n",
            "2 589 tensor(0.8721, grad_fn=<NllLossBackward>)\n",
            "2 599 tensor(0.9589, grad_fn=<NllLossBackward>)\n",
            "2 609 tensor(1.0145, grad_fn=<NllLossBackward>)\n",
            "2 619 tensor(0.9770, grad_fn=<NllLossBackward>)\n",
            "2 629 tensor(1.1209, grad_fn=<NllLossBackward>)\n",
            "2 639 tensor(0.9761, grad_fn=<NllLossBackward>)\n",
            "2 649 tensor(1.3182, grad_fn=<NllLossBackward>)\n",
            "2 659 tensor(1.2938, grad_fn=<NllLossBackward>)\n",
            "2 669 tensor(1.0983, grad_fn=<NllLossBackward>)\n",
            "2 679 tensor(0.8648, grad_fn=<NllLossBackward>)\n",
            "2 689 tensor(1.7550, grad_fn=<NllLossBackward>)\n",
            "2 699 tensor(1.2097, grad_fn=<NllLossBackward>)\n",
            "2 709 tensor(1.1653, grad_fn=<NllLossBackward>)\n",
            "2 719 tensor(0.9073, grad_fn=<NllLossBackward>)\n",
            "2 729 tensor(1.5877, grad_fn=<NllLossBackward>)\n",
            "2 739 tensor(1.2044, grad_fn=<NllLossBackward>)\n",
            "2 749 tensor(0.8132, grad_fn=<NllLossBackward>)\n",
            "2 759 tensor(1.6660, grad_fn=<NllLossBackward>)\n",
            "2 769 tensor(0.7857, grad_fn=<NllLossBackward>)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ze8lS6Sh2nXa",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "filename=\"model_0.pth\"\n",
        "loaded_model=ResNet()\n",
        "loaded_model.load_state_dict(torch.load(filename))\n",
        "loaded_model.eval()\n",
        "for i in model.parameters():\n",
        "  print(i)"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}