Hackthebox Battle in OrI/On Pytorch.ipynb

hackthebox-battle-in-ori-on-pytorch.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Hackthebox Battle in OrI/On Pytorch.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "machine_shape": "hm",
      "authorship_tag": "ABX9TyNL9rdPATNuVa0srTW6pEWA",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/tudoanh/c1034d51fa9c9f530f8c83e8ba874ee1/hackthebox-battle-in-ori-on-pytorch.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 27,
      "metadata": {
        "id": "ZkpDvFFaypR5"
      },
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "from torchsummary import summary"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model = torch.load(\"model.pth\")\n",
        "print(model)"
      ],
      "metadata": {
        "id": "rW7oodix4COb",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "5cf9da79-2879-4f02-ba02-8004cbaf0525"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "net(\n",
            "  (layer1): Sequential(\n",
            "    (0): Conv2d(1, 16, kernel_size=(3, 3), stride=(2, 2))\n",
            "    (1): ReLU()\n",
            "    (2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  )\n",
            "  (layer2): Sequential(\n",
            "    (0): Conv2d(16, 32, kernel_size=(3, 3), stride=(2, 2))\n",
            "    (1): ReLU()\n",
            "    (2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  )\n",
            "  (fc1): Linear(in_features=5408, out_features=10, bias=True)\n",
            "  (fc2): Linear(in_features=10, out_features=2, bias=True)\n",
            "  (relu): ReLU()\n",
            ")\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "summary(model, (1, 224, 224))"
      ],
      "metadata": {
        "id": "zABTIuVtzxjU",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4c446a7d-0839-4844-8223-6207e07aa112"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "----------------------------------------------------------------\n",
            "        Layer (type)               Output Shape         Param #\n",
            "================================================================\n",
            "            Conv2d-1         [-1, 16, 111, 111]             160\n",
            "              ReLU-2         [-1, 16, 111, 111]               0\n",
            "         MaxPool2d-3           [-1, 16, 55, 55]               0\n",
            "            Conv2d-4           [-1, 32, 27, 27]           4,640\n",
            "              ReLU-5           [-1, 32, 27, 27]               0\n",
            "         MaxPool2d-6           [-1, 32, 13, 13]               0\n",
            "            Linear-7                   [-1, 10]          54,090\n",
            "              ReLU-8                   [-1, 10]               0\n",
            "            Linear-9                    [-1, 2]              22\n",
            "================================================================\n",
            "Total params: 58,912\n",
            "Trainable params: 58,912\n",
            "Non-trainable params: 0\n",
            "----------------------------------------------------------------\n",
            "Input size (MB): 0.19\n",
            "Forward/backward pass size (MB): 3.77\n",
            "Params size (MB): 0.22\n",
            "Estimated Total Size (MB): 4.19\n",
            "----------------------------------------------------------------\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "tensored = torch.rand(1, 1, 224, 224).to(\"cuda:0\")\n",
        "label_1_pred = 0.234\n",
        "label_2_pred = 0.766\n",
        "\n",
        "for i in range(100):\n",
        "  tensored.requires_grad_()\n",
        "  pred = model(tensored)[0]\n",
        "  diff = ((pred[1] - 0.234) ** 2).mean()\n",
        "  acc = model(tensored)\n",
        "  if round(float(acc[0][1]), 4) == label_1_pred and round(float(acc[0][0]), 4) == label_2_pred:\n",
        "    print(f\"Loss: {diff}\")\n",
        "    print(f\"Accuracy: {acc}\")\n",
        "    break\n",
        "  diff.backward()\n",
        "  tensored = tensored.detach() - tensored.grad * 64"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2tFe2ptE0vuw",
        "outputId": "b620887c-8fef-4788-9f13-ee32d007ba51"
      },
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Loss: 1.7865908752412452e-10\n",
            "Accuracy: tensor([[0.7660, 0.2340]], device='cuda:0', grad_fn=<SoftmaxBackward0>)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "res = tensored.cpu().detach().numpy()"
      ],
      "metadata": {
        "id": "jSkRJFrUegcT"
      },
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "with open(\"result.npy\", \"wb\") as f:\n",
        "  np.save(f, res)"
      ],
      "metadata": {
        "id": "Ax4DxLeZgcn2"
      },
      "execution_count": 30,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        ""
      ],
      "metadata": {
        "id": "hMVbYRrkinIf"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}

Source code:

https://drive.google.com/file/d/1ZHAfJNEumNNgD_rcs7Gm2vStpNnve72k/view?usp=sharing