Recognition_Minor1.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "name": "Recognition_Minor1.ipynb",
      "provenance": [],
      "collapsed_sections": [
        "DVR1FiuJ-BzX"
      ]
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "58d4c8ff778f40e98daf7d28829c82f3": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_53c34d36394d44e1956ed7e1614c0174",
              "IPY_MODEL_98649b65f60f460782081f3ef3292268"
            ],
            "layout": "IPY_MODEL_61cf1825adc34bbda210811d7244af86"
          }
        },
        "53c34d36394d44e1956ed7e1614c0174": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_987b90badeb04807aca7a58a8bb5c13f",
            "max": 1,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_3b8ff4553cd24ab9bd445062938f0db9",
            "value": 0
          }
        },
        "98649b65f60f460782081f3ef3292268": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_24fb2474b2d44a05900dde554f3fb7df",
            "placeholder": "​",
            "style": "IPY_MODEL_fd4250fd31f14b3ea76e1767536bbf40",
            "value": " 0/? [00:06&lt;?, ?it/s]"
          }
        },
        "61cf1825adc34bbda210811d7244af86": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "987b90badeb04807aca7a58a8bb5c13f": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "3b8ff4553cd24ab9bd445062938f0db9": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": "initial"
          }
        },
        "24fb2474b2d44a05900dde554f3fb7df": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "fd4250fd31f14b3ea76e1767536bbf40": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "37cc23646d794ee0b1040a8741802298": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_7f2e54d70c9d4ecd948cf8713fb6951e",
              "IPY_MODEL_3879e40007fa4a6cbf177779f476fe65"
            ],
            "layout": "IPY_MODEL_dc614e9f2acb4d128eb1b1c2f27da9c4"
          }
        },
        "7f2e54d70c9d4ecd948cf8713fb6951e": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_5ef2d3b1b588458ebafad185361f3915",
            "max": 1,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_aeeee26de1884eb4905fe853a0467089",
            "value": 1
          }
        },
        "3879e40007fa4a6cbf177779f476fe65": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_7c7364e7b0064591b995b0bd9f8c7493",
            "placeholder": "​",
            "style": "IPY_MODEL_cbff30530a924deebe17a83aca9253e3",
            "value": " 170500096/? [00:06&lt;00:00, 26796305.17it/s]"
          }
        },
        "dc614e9f2acb4d128eb1b1c2f27da9c4": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "5ef2d3b1b588458ebafad185361f3915": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "aeeee26de1884eb4905fe853a0467089": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": "initial"
          }
        },
        "7c7364e7b0064591b995b0bd9f8c7493": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "cbff30530a924deebe17a83aca9253e3": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "09becf37cc1242e9a80cb13445613e63": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_1d66ddc967e6482b9a109f406f6bc5b5",
              "IPY_MODEL_63c13e4e28e248f08e58a99fffc5a416"
            ],
            "layout": "IPY_MODEL_7714efcf6b734b95992734487b9665f4"
          }
        },
        "1d66ddc967e6482b9a109f406f6bc5b5": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "100%",
            "description_tooltip": null,
            "layout": "IPY_MODEL_7382230a09e44481bf4d7e7219fc4eff",
            "max": 553507836,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_4c0c668a465148c981b5c1ed588d20ee",
            "value": 553507836
          }
        },
        "63c13e4e28e248f08e58a99fffc5a416": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_30ea63b4a1614abe96b838a9024f067d",
            "placeholder": "​",
            "style": "IPY_MODEL_ca67a6e01cfd4efd8cf0ce93a78c444b",
            "value": " 528M/528M [00:07&lt;00:00, 77.9MB/s]"
          }
        },
        "7714efcf6b734b95992734487b9665f4": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "7382230a09e44481bf4d7e7219fc4eff": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "4c0c668a465148c981b5c1ed588d20ee": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": "initial"
          }
        },
        "30ea63b4a1614abe96b838a9024f067d": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "ca67a6e01cfd4efd8cf0ce93a78c444b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        }
      }
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6KNf9CD2ytqH"
      },
      "source": [
        "## Outline"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BDKI47gsW1Ii"
      },
      "source": [
        "1. Loading datasets - Transforming images\n",
        "2. VGG-16 with modification to network head\n",
        "3. Using pre-trained models\n",
        "4. Storing intermediate models\n",
        "5. Inception v3"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rbZpqiiiylAg"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "import torch\n",
        "import torchvision\n",
        "import torchvision.transforms as transforms\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1NBprN3Ry2NE",
        "outputId": "3ff172ce-ae0a-4cc3-9a44-dd24c08939f9"
      },
      "source": [
        "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
        "print(device)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "cuda:0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hGzpz3oE1Ikz"
      },
      "source": [
        "## Dataset, transforms, and visualisation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y9c57fNA5Wsi"
      },
      "source": [
        "transform_train = transforms.Compose([\n",
        "    transforms.RandomResizedCrop(224), \n",
        "    transforms.ToTensor(),\n",
        "    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),\n",
        "    ])\n",
        "\n",
        "transform_test = transforms.Compose([\n",
        "    transforms.RandomResizedCrop(224), \n",
        "    transforms.ToTensor(),\n",
        "    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),\n",
        "    ])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 148,
          "referenced_widgets": [
            "58d4c8ff778f40e98daf7d28829c82f3",
            "53c34d36394d44e1956ed7e1614c0174",
            "98649b65f60f460782081f3ef3292268",
            "61cf1825adc34bbda210811d7244af86",
            "987b90badeb04807aca7a58a8bb5c13f",
            "3b8ff4553cd24ab9bd445062938f0db9",
            "24fb2474b2d44a05900dde554f3fb7df",
            "fd4250fd31f14b3ea76e1767536bbf40",
            "37cc23646d794ee0b1040a8741802298",
            "7f2e54d70c9d4ecd948cf8713fb6951e",
            "3879e40007fa4a6cbf177779f476fe65",
            "dc614e9f2acb4d128eb1b1c2f27da9c4",
            "5ef2d3b1b588458ebafad185361f3915",
            "aeeee26de1884eb4905fe853a0467089",
            "7c7364e7b0064591b995b0bd9f8c7493",
            "cbff30530a924deebe17a83aca9253e3"
          ]
        },
        "id": "fvsHorMPzISb",
        "outputId": "fe14d5cd-ccaf-41d8-d33a-b9d5e7404a42"
      },
      "source": [
        "trainset = torchvision.datasets.CIFAR10(root='./data', train=True, \n",
        "                                        download=True, \n",
        "                                        transform=transform_train)\n",
        "testset = torchvision.datasets.CIFAR10(root='./data', train=False, \n",
        "                                        download=True, \n",
        "                                        transform=transform_test)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to ./data/cifar-10-python.tar.gz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "58d4c8ff778f40e98daf7d28829c82f3",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=1.0, bar_style='info', max=1.0), HTML(value='')))"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Failed download. Trying https -> http instead. Downloading http://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to ./data/cifar-10-python.tar.gz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "37cc23646d794ee0b1040a8741802298",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=1.0, bar_style='info', max=1.0), HTML(value='')))"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Extracting ./data/cifar-10-python.tar.gz to ./data\n",
            "Files already downloaded and verified\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "V0dOpKtiAG4s"
      },
      "source": [
        "num_classes = 10"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xTNj3LQY4eTS"
      },
      "source": [
        "batch_size = 4"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KfLwRIXH08tg"
      },
      "source": [
        "trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size, shuffle=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KEOz-75x1NGu",
        "outputId": "a8768e6a-2789-42c0-a8ea-d54429b06f68"
      },
      "source": [
        "dataiter = iter(trainloader)\n",
        "images, labels = dataiter.next()\n",
        "\n",
        "print(images.shape)\n",
        "\n",
        "print(images[1].shape)\n",
        "print(labels[1].item())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([4, 3, 224, 224])\n",
            "torch.Size([3, 224, 224])\n",
            "7\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5oBeIwYC1N3c"
      },
      "source": [
        "def imshow(img, title):\n",
        "    npimg = img.numpy() / 2 + 0.5\n",
        "    plt.figure(figsize=(batch_size, 1))\n",
        "    plt.axis('off')\n",
        "    plt.imshow(np.transpose(npimg, (1, 2, 0)))\n",
        "    plt.title(title)\n",
        "    plt.show()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cfk-SYLY1Sbl"
      },
      "source": [
        "def show_batch_images(dataloader):\n",
        "    images, labels = next(iter(dataloader))\n",
        "    img = torchvision.utils.make_grid(images)\n",
        "    imshow(img, title=[str(x.item()) for x in labels])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 386
        },
        "id": "l_Zi_s3p1htN",
        "outputId": "24cd8e88-eda5-4dd1-fa64-2950888c6a73"
      },
      "source": [
        "for i in range(4):\n",
        "    show_batch_images(trainloader)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 288x72 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 288x72 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 288x72 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 288x72 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1E-pnTG97DDz"
      },
      "source": [
        "## Creating VGG-16"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_QMSp0gEJLow"
      },
      "source": [
        "https://pytorch.org/docs/master/_modules/torchvision/models/vgg.html"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ra33PbJS28P3"
      },
      "source": [
        "from torchvision import models"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wvte5SSA7G7r"
      },
      "source": [
        "vgg = models.vgg16_bn()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7w-p73Tz9aZ_",
        "outputId": "cb32edd4-9897-494f-d583-bc148c4f4106"
      },
      "source": [
        "print(vgg)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "VGG(\n",
            "  (features): Sequential(\n",
            "    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (2): ReLU(inplace=True)\n",
            "    (3): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (4): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (7): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (8): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (9): ReLU(inplace=True)\n",
            "    (10): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (11): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (12): ReLU(inplace=True)\n",
            "    (13): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (14): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (15): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (16): ReLU(inplace=True)\n",
            "    (17): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (18): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (19): ReLU(inplace=True)\n",
            "    (20): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (21): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (22): ReLU(inplace=True)\n",
            "    (23): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (24): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (25): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (26): ReLU(inplace=True)\n",
            "    (27): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (28): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (29): ReLU(inplace=True)\n",
            "    (30): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (31): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (32): ReLU(inplace=True)\n",
            "    (33): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (34): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (35): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (36): ReLU(inplace=True)\n",
            "    (37): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (38): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (39): ReLU(inplace=True)\n",
            "    (40): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (41): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (42): ReLU(inplace=True)\n",
            "    (43): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  )\n",
            "  (avgpool): AdaptiveAvgPool2d(output_size=(7, 7))\n",
            "  (classifier): Sequential(\n",
            "    (0): Linear(in_features=25088, out_features=4096, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Dropout(p=0.5, inplace=False)\n",
            "    (3): Linear(in_features=4096, out_features=4096, bias=True)\n",
            "    (4): ReLU(inplace=True)\n",
            "    (5): Dropout(p=0.5, inplace=False)\n",
            "    (6): Linear(in_features=4096, out_features=1000, bias=True)\n",
            "  )\n",
            ")\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "N1bQdxQz-Dil",
        "outputId": "f58a9133-1f06-4e98-bb1b-f54a485bdf44"
      },
      "source": [
        "print(vgg.features[0])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CX9PohSB-1Dx",
        "outputId": "8d4acd26-061a-4ff1-b9ae-7dc56fe7e2bf"
      },
      "source": [
        "print(vgg.classifier[6])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Linear(in_features=4096, out_features=1000, bias=True)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "koDRbWi8_ApT",
        "outputId": "8862747d-602f-4545-c6e7-398aaf335cc2"
      },
      "source": [
        "final_in_features = vgg.classifier[6].in_features\n",
        "mod_classifier = list(vgg.classifier.children())[:-1]\n",
        "mod_classifier.extend([nn.Linear(final_in_features, num_classes)])\n",
        "print(mod_classifier)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[Linear(in_features=25088, out_features=4096, bias=True), ReLU(inplace=True), Dropout(p=0.5, inplace=False), Linear(in_features=4096, out_features=4096, bias=True), ReLU(inplace=True), Dropout(p=0.5, inplace=False), Linear(in_features=4096, out_features=10, bias=True)]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_vHjXi1j_glv",
        "outputId": "9092e74c-156d-491c-9ee9-cbe752acdea4"
      },
      "source": [
        "vgg.classifier = nn.Sequential(*mod_classifier)\n",
        "print(vgg)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "VGG(\n",
            "  (features): Sequential(\n",
            "    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (2): ReLU(inplace=True)\n",
            "    (3): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (4): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (7): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (8): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (9): ReLU(inplace=True)\n",
            "    (10): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (11): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (12): ReLU(inplace=True)\n",
            "    (13): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (14): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (15): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (16): ReLU(inplace=True)\n",
            "    (17): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (18): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (19): ReLU(inplace=True)\n",
            "    (20): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (21): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (22): ReLU(inplace=True)\n",
            "    (23): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (24): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (25): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (26): ReLU(inplace=True)\n",
            "    (27): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (28): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (29): ReLU(inplace=True)\n",
            "    (30): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (31): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (32): ReLU(inplace=True)\n",
            "    (33): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (34): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (35): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (36): ReLU(inplace=True)\n",
            "    (37): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (38): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (39): ReLU(inplace=True)\n",
            "    (40): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (41): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (42): ReLU(inplace=True)\n",
            "    (43): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  )\n",
            "  (avgpool): AdaptiveAvgPool2d(output_size=(7, 7))\n",
            "  (classifier): Sequential(\n",
            "    (0): Linear(in_features=25088, out_features=4096, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Dropout(p=0.5, inplace=False)\n",
            "    (3): Linear(in_features=4096, out_features=4096, bias=True)\n",
            "    (4): ReLU(inplace=True)\n",
            "    (5): Dropout(p=0.5, inplace=False)\n",
            "    (6): Linear(in_features=4096, out_features=10, bias=True)\n",
            "  )\n",
            ")\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DVR1FiuJ-BzX"
      },
      "source": [
        "### Train CIFAR10"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pKnlGE1q7JtN"
      },
      "source": [
        "batch_size = 16\n",
        "\n",
        "trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size, shuffle=True)\n",
        "testloader = torch.utils.data.DataLoader(testset, batch_size=batch_size, shuffle=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MtRHmDs_BvZG"
      },
      "source": [
        "def evaluation(dataloader, model):\n",
        "    total, correct = 0, 0\n",
        "    for data in dataloader:\n",
        "        inputs, labels = data\n",
        "        inputs, labels = inputs.to(device), labels.to(device)\n",
        "        outputs = model(inputs)\n",
        "        _, pred = torch.max(outputs.data, 1)\n",
        "        total += labels.size(0)\n",
        "        correct += (pred == labels).sum().item()\n",
        "    return 100 * correct / total"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "htVdEliECDsz"
      },
      "source": [
        "vgg = vgg.to(device)\n",
        "loss_fn = nn.CrossEntropyLoss()\n",
        "opt = optim.SGD(vgg.parameters(), lr=0.05)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 823
        },
        "id": "xSWnZJxjBbwW",
        "outputId": "e61a1a90-7a87-42ba-c7f4-1e825f9c914f"
      },
      "source": [
        "loss_epoch_arr = []\n",
        "max_epochs = 1\n",
        "\n",
        "n_iters = np.ceil(50000/batch_size)\n",
        "\n",
        "for epoch in range(max_epochs):\n",
        "\n",
        "    for i, data in enumerate(trainloader, 0):\n",
        "\n",
        "        inputs, labels = data\n",
        "        inputs, labels = inputs.to(device), labels.to(device)\n",
        "\n",
        "        opt.zero_grad()\n",
        "\n",
        "        outputs = vgg(inputs)\n",
        "        loss = loss_fn(outputs, labels)\n",
        "        loss.backward()\n",
        "        opt.step()\n",
        "        \n",
        "        del inputs, labels, outputs\n",
        "        torch.cuda.empty_cache()\n",
        "        \n",
        "        if i % 100 == 0:\n",
        "            print('Iteration: %d/%d, Loss: %0.2f' % (i, n_iters, loss.item()))\n",
        "        \n",
        "    loss_epoch_arr.append(loss.item())\n",
        "        \n",
        "    print('Epoch: %d/%d, Test acc: %0.2f, Train acc: %0.2f' % (\n",
        "        epoch, max_epochs, \n",
        "        evaluation(testloader, vgg), evaluation(trainloader, vgg)))\n",
        "    \n",
        "plt.plot(loss_epoch_arr)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Iteration: 0/3125, Loss: 2.56\n",
            "Iteration: 100/3125, Loss: 2.67\n",
            "Iteration: 200/3125, Loss: 2.39\n",
            "Iteration: 300/3125, Loss: 2.39\n",
            "Iteration: 400/3125, Loss: 2.21\n",
            "Iteration: 500/3125, Loss: 2.28\n",
            "Iteration: 600/3125, Loss: 2.13\n",
            "Iteration: 700/3125, Loss: 2.66\n",
            "Iteration: 800/3125, Loss: 2.61\n",
            "Iteration: 900/3125, Loss: 2.38\n",
            "Iteration: 1000/3125, Loss: 2.19\n",
            "Iteration: 1100/3125, Loss: 2.03\n",
            "Iteration: 1200/3125, Loss: 2.31\n",
            "Iteration: 1300/3125, Loss: 2.08\n",
            "Iteration: 1400/3125, Loss: 2.15\n",
            "Iteration: 1500/3125, Loss: 2.05\n",
            "Iteration: 1600/3125, Loss: 2.03\n",
            "Iteration: 1700/3125, Loss: 2.09\n",
            "Iteration: 1800/3125, Loss: 2.01\n",
            "Iteration: 1900/3125, Loss: 2.43\n",
            "Iteration: 2000/3125, Loss: 1.97\n",
            "Iteration: 2100/3125, Loss: 2.16\n",
            "Iteration: 2200/3125, Loss: 1.94\n",
            "Iteration: 2300/3125, Loss: 2.06\n",
            "Iteration: 2400/3125, Loss: 1.68\n",
            "Iteration: 2500/3125, Loss: 1.91\n",
            "Iteration: 2600/3125, Loss: 1.78\n",
            "Iteration: 2700/3125, Loss: 1.92\n",
            "Iteration: 2800/3125, Loss: 1.97\n",
            "Iteration: 2900/3125, Loss: 1.74\n",
            "Iteration: 3000/3125, Loss: 1.79\n",
            "Iteration: 3100/3125, Loss: 1.82\n",
            "Epoch: 0/1, Test acc: 27.44, Train acc: 26.59\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Bh4gREC-IGFB"
      },
      "source": [
        "### Freeze layers of Convolutional Operations"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3aD1lt1qcJOU"
      },
      "source": [
        "batch_size = 16\n",
        "\n",
        "trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size, shuffle=True)\n",
        "testloader = torch.utils.data.DataLoader(testset, batch_size=batch_size, shuffle=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 82,
          "referenced_widgets": [
            "09becf37cc1242e9a80cb13445613e63",
            "1d66ddc967e6482b9a109f406f6bc5b5",
            "63c13e4e28e248f08e58a99fffc5a416",
            "7714efcf6b734b95992734487b9665f4",
            "7382230a09e44481bf4d7e7219fc4eff",
            "4c0c668a465148c981b5c1ed588d20ee",
            "30ea63b4a1614abe96b838a9024f067d",
            "ca67a6e01cfd4efd8cf0ce93a78c444b"
          ]
        },
        "id": "FzniE5sKKlnY",
        "outputId": "95678ff0-1a5a-4672-be3f-04036a007b2b"
      },
      "source": [
        "vgg = models.vgg16_bn(pretrained=True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading: \"https://download.pytorch.org/models/vgg16_bn-6c64b313.pth\" to /root/.cache/torch/hub/checkpoints/vgg16_bn-6c64b313.pth\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "09becf37cc1242e9a80cb13445613e63",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, max=553507836.0), HTML(value='')))"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MzaHEHaczpWS"
      },
      "source": [
        "for param in vgg.parameters():\n",
        "    param.requires_grad = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YFAhAvahzrRh"
      },
      "source": [
        "final_in_features = vgg.classifier[6].in_features\n",
        "vgg.classifier[6] = nn.Linear(final_in_features, num_classes)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uqXCJEiK0Z2V",
        "outputId": "c3373724-8609-453f-9eaf-985c269476c8"
      },
      "source": [
        "for param in vgg.parameters():\n",
        "    if param.requires_grad:\n",
        "        print(param.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([10, 4096])\n",
            "torch.Size([10])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lH-99npmPRUk"
      },
      "source": [
        "vgg = vgg.to(device)\n",
        "loss_fn = nn.CrossEntropyLoss()\n",
        "opt = optim.SGD(vgg.parameters(), lr=0.05)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 819
        },
        "id": "5HuWwxaxIMTV",
        "outputId": "a27ea641-26bf-4cac-a640-892f9dd09cfb"
      },
      "source": [
        "loss_epoch_arr = []\n",
        "max_epochs = 1\n",
        "\n",
        "n_iters = np.ceil(50000/batch_size)\n",
        "\n",
        "for epoch in range(max_epochs):\n",
        "\n",
        "    for i, data in enumerate(trainloader, 0):\n",
        "\n",
        "        inputs, labels = data\n",
        "        inputs, labels = inputs.to(device), labels.to(device)\n",
        "\n",
        "        opt.zero_grad()\n",
        "\n",
        "        outputs = vgg(inputs)\n",
        "        loss = loss_fn(outputs, labels)\n",
        "        loss.backward()\n",
        "        opt.step()\n",
        "        \n",
        "        if i % 100 == 0:\n",
        "            print('Iteration: %d/%d, Loss: %0.2f' % (i, n_iters, loss.item()))\n",
        "            \n",
        "        del inputs, labels, outputs\n",
        "        torch.cuda.empty_cache()\n",
        "        \n",
        "    loss_epoch_arr.append(loss.item())\n",
        "        \n",
        "    print('Epoch: %d/%d, Test acc: %0.2f, Train acc: %0.2f' % (\n",
        "        epoch, max_epochs, \n",
        "        evaluation(testloader, vgg), evaluation(trainloader, vgg)))\n",
        "    \n",
        "    \n",
        "plt.plot(loss_epoch_arr)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Iteration: 0/3125, Loss: 2.35\n",
            "Iteration: 100/3125, Loss: 2.47\n",
            "Iteration: 200/3125, Loss: 1.53\n",
            "Iteration: 300/3125, Loss: 1.85\n",
            "Iteration: 400/3125, Loss: 1.91\n",
            "Iteration: 500/3125, Loss: 1.89\n",
            "Iteration: 600/3125, Loss: 1.75\n",
            "Iteration: 700/3125, Loss: 1.10\n",
            "Iteration: 800/3125, Loss: 1.20\n",
            "Iteration: 900/3125, Loss: 2.74\n",
            "Iteration: 1000/3125, Loss: 1.48\n",
            "Iteration: 1100/3125, Loss: 1.38\n",
            "Iteration: 1200/3125, Loss: 1.59\n",
            "Iteration: 1300/3125, Loss: 2.73\n",
            "Iteration: 1400/3125, Loss: 2.13\n",
            "Iteration: 1500/3125, Loss: 2.72\n",
            "Iteration: 1600/3125, Loss: 2.55\n",
            "Iteration: 1700/3125, Loss: 1.29\n",
            "Iteration: 1800/3125, Loss: 1.28\n",
            "Iteration: 1900/3125, Loss: 2.23\n",
            "Iteration: 2000/3125, Loss: 1.88\n",
            "Iteration: 2100/3125, Loss: 2.93\n",
            "Iteration: 2200/3125, Loss: 1.86\n",
            "Iteration: 2300/3125, Loss: 1.22\n",
            "Iteration: 2400/3125, Loss: 2.61\n",
            "Iteration: 2500/3125, Loss: 1.08\n",
            "Iteration: 2600/3125, Loss: 3.16\n",
            "Iteration: 2700/3125, Loss: 1.88\n",
            "Iteration: 2800/3125, Loss: 2.60\n",
            "Iteration: 2900/3125, Loss: 1.55\n",
            "Iteration: 3000/3125, Loss: 1.97\n",
            "Iteration: 3100/3125, Loss: 2.02\n",
            "Epoch: 0/1, Test acc: 38.94, Train acc: 39.14\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AeacSwlg5p2r"
      },
      "source": [
        "### With model copies"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4-dZMeUTpAxC"
      },
      "source": [
        "import copy"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "background_save": true,
          "base_uri": "https://localhost:8080/"
        },
        "id": "Wf0qM_-PlXl-",
        "outputId": "1bc8d594-9c2f-4839-9205-01cc402880e3"
      },
      "source": [
        "loss_epoch_arr = []\n",
        "max_epochs = 1\n",
        "\n",
        "min_loss = 1000\n",
        "\n",
        "n_iters = np.ceil(50000/batch_size)\n",
        "\n",
        "for epoch in range(max_epochs):\n",
        "\n",
        "    for i, data in enumerate(trainloader, 0):\n",
        "\n",
        "        inputs, labels = data\n",
        "        inputs, labels = inputs.to(device), labels.to(device)\n",
        "\n",
        "        opt.zero_grad()\n",
        "\n",
        "        outputs = vgg(inputs)\n",
        "        loss = loss_fn(outputs, labels)\n",
        "        loss.backward()\n",
        "        opt.step()\n",
        "        \n",
        "        if min_loss > loss.item():\n",
        "            min_loss = loss.item()\n",
        "            best_model = copy.deepcopy(vgg.state_dict())\n",
        "            print('Min loss %0.2f' % min_loss)\n",
        "        \n",
        "        if i % 100 == 0:\n",
        "            print('Iteration: %d/%d, Loss: %0.2f' % (i, n_iters, loss.item()))\n",
        "            \n",
        "        del inputs, labels, outputs\n",
        "        torch.cuda.empty_cache()\n",
        "        \n",
        "    loss_epoch_arr.append(loss.item())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Min loss 2.90\n",
            "Iteration: 0/3125, Loss: 2.90\n",
            "Min loss 1.86\n",
            "Min loss 1.62\n",
            "Min loss 1.53\n",
            "Min loss 1.09\n",
            "Min loss 0.84\n",
            "Iteration: 100/3125, Loss: 2.94\n",
            "Iteration: 200/3125, Loss: 4.09\n",
            "Min loss 0.73\n",
            "Min loss 0.68\n",
            "Iteration: 300/3125, Loss: 2.12\n",
            "Iteration: 400/3125, Loss: 1.82\n",
            "Iteration: 500/3125, Loss: 2.12\n",
            "Iteration: 600/3125, Loss: 2.95\n",
            "Iteration: 700/3125, Loss: 3.27\n",
            "Iteration: 800/3125, Loss: 2.00\n",
            "Iteration: 900/3125, Loss: 2.71\n",
            "Iteration: 1000/3125, Loss: 2.83\n",
            "Min loss 0.46\n",
            "Iteration: 1100/3125, Loss: 1.10\n",
            "Iteration: 1200/3125, Loss: 2.80\n",
            "Iteration: 1300/3125, Loss: 2.22\n",
            "Iteration: 1400/3125, Loss: 2.40\n",
            "Iteration: 1500/3125, Loss: 2.33\n",
            "Iteration: 1600/3125, Loss: 2.57\n",
            "Iteration: 1700/3125, Loss: 1.71\n",
            "Iteration: 1800/3125, Loss: 2.09\n",
            "Iteration: 1900/3125, Loss: 2.50\n",
            "Iteration: 2000/3125, Loss: 2.26\n",
            "Iteration: 2100/3125, Loss: 3.77\n",
            "Iteration: 2200/3125, Loss: 1.73\n",
            "Iteration: 2300/3125, Loss: 1.93\n",
            "Iteration: 2400/3125, Loss: 2.46\n",
            "Iteration: 2500/3125, Loss: 1.91\n",
            "Iteration: 2600/3125, Loss: 3.11\n",
            "Iteration: 2700/3125, Loss: 2.68\n",
            "Iteration: 2800/3125, Loss: 1.66\n",
            "Iteration: 2900/3125, Loss: 1.83\n",
            "Iteration: 3000/3125, Loss: 1.14\n",
            "Iteration: 3100/3125, Loss: 2.89\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "background_save": true
        },
        "id": "_2GAOmBytwZ9",
        "outputId": "726557c7-7919-4fa2-97b8-fb9234d95ec0"
      },
      "source": [
        "vgg.load_state_dict(best_model)\n",
        "print(evaluation(trainloader, vgg), evaluation(testloader, vgg))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "46.684 46.22\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "asFT3q7vdDbd"
      },
      "source": [
        "## Inception Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "k-gBu2zCGWE8"
      },
      "source": [
        "https://pytorch.org/docs/master/_modules/torchvision/models/inception.html"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rpAr08aGIOle"
      },
      "source": [
        "inception = models.inception_v3(pretrained=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TM_ACRUHdMfi"
      },
      "source": [
        "print(inception)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Nbu0JobP6ea9"
      },
      "source": [
        "for param in inception.parameters():\n",
        "    param.requires_grad = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "j31it2qX5Nfs"
      },
      "source": [
        "aux_in_features = inception.AuxLogits.fc.in_features\n",
        "inception.AuxLogits.fc = nn.Linear(aux_in_features, num_classes)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7-M6nGOh6Xg2"
      },
      "source": [
        "for param in inception.parameters():\n",
        "    if param.requires_grad:\n",
        "        print(param.shape)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pOI1Au2A6l9X"
      },
      "source": [
        "in_features = inception.fc.in_features\n",
        "inception.fc = nn.Linear(in_features, num_classes)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sF5pPlUS7I-7"
      },
      "source": [
        "for param in inception.parameters():\n",
        "    if param.requires_grad:\n",
        "        print(param.shape)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vMSMZa_j7JXG"
      },
      "source": [
        "transform_train = transforms.Compose([\n",
        "    transforms.RandomResizedCrop(299), \n",
        "    transforms.ToTensor(),\n",
        "    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),\n",
        "    ])\n",
        "\n",
        "transform_test = transforms.Compose([\n",
        "    transforms.RandomResizedCrop(299), \n",
        "    transforms.ToTensor(),\n",
        "    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),\n",
        "    ])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kYVy8C0H7vL7"
      },
      "source": [
        "trainset = torchvision.datasets.CIFAR10(root='./data', train=True, \n",
        "                                        download=True, \n",
        "                                        transform=transform_train)\n",
        "testset = torchvision.datasets.CIFAR10(root='./data', train=False, \n",
        "                                        download=True, \n",
        "                                        transform=transform_test)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tkUGBe3x7zld"
      },
      "source": [
        "batch_size=16\n",
        "\n",
        "trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size, shuffle=True)\n",
        "testloader = torch.utils.data.DataLoader(testset, batch_size=batch_size, shuffle=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jeKDmSV98QfZ"
      },
      "source": [
        "inception = inception.to(device)\n",
        "loss_fn = nn.CrossEntropyLoss()\n",
        "opt = optim.SGD(inception.parameters(), lr=0.01)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ou_Q_JwK_fqn"
      },
      "source": [
        "def evaluation_inception(dataloader, model):\n",
        "    total, correct = 0, 0\n",
        "    for data in dataloader:\n",
        "        inputs, labels = data\n",
        "        inputs, labels = inputs.to(device), labels.to(device)\n",
        "        outputs, aux_outputs = model(inputs)\n",
        "        _, pred = torch.max(outputs.data, 1)\n",
        "        total += labels.size(0)\n",
        "        correct += (pred == labels).sum().item()\n",
        "    return 100 * correct / total"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VqmnOJqP75AK"
      },
      "source": [
        "loss_epoch_arr = []\n",
        "max_epochs = 1\n",
        "\n",
        "min_loss = 1000\n",
        "\n",
        "n_iters = np.ceil(50000/batch_size)\n",
        "\n",
        "for epoch in range(max_epochs):\n",
        "\n",
        "    for i, data in enumerate(trainloader, 0):\n",
        "\n",
        "        inputs, labels = data\n",
        "        inputs, labels = inputs.to(device), labels.to(device)\n",
        "\n",
        "        opt.zero_grad()\n",
        "\n",
        "        outputs, aux_outputs = inception(inputs)\n",
        "        loss = loss_fn(outputs, labels) + 0.3 * loss_fn(aux_outputs, labels)\n",
        "        loss.backward()\n",
        "        opt.step()\n",
        "        \n",
        "        if min_loss > loss.item():\n",
        "            min_loss = loss.item()\n",
        "            best_model = copy.deepcopy(inception.state_dict())\n",
        "            print('Min loss %0.2f' % min_loss)\n",
        "        \n",
        "        if i % 100 == 0:\n",
        "            print('Iteration: %d/%d, Loss: %0.2f' % (i, n_iters, loss.item()))\n",
        "            \n",
        "        del inputs, labels, outputs\n",
        "        torch.cuda.empty_cache()\n",
        "        \n",
        "    loss_epoch_arr.append(loss.item())\n",
        "        \n",
        "    print('Epoch: %d/%d, Test acc: %0.2f, Train acc: %0.2f' % (\n",
        "        epoch, max_epochs, \n",
        "        evaluation_inception(testloader, inception), \n",
        "        evaluation_inception(trainloader, inception)))\n",
        "    \n",
        "    \n",
        "plt.plot(loss_epoch_arr)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YfcUcxaL8T_U"
      },
      "source": [
        "inception.load_state_dict(best_model)\n",
        "print(evaluation_inception(trainloader, inception), evaluation_inception(testloader, inception))"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}