Pytorch_Tutorial.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "gpuClass": "standard",
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "d55373061520403c96d61ffdad06993a": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "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_bb4bf1afa52546faa33ed0a48d5465a0",
              "IPY_MODEL_3744035230bb46e6b30dfe79386368ba",
              "IPY_MODEL_c13a23b2a235491a8686aa83dad15aeb"
            ],
            "layout": "IPY_MODEL_2f4c0514d0314351a1c81815569b829c"
          }
        },
        "bb4bf1afa52546faa33ed0a48d5465a0": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "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_236501633e0c4e128485a479533c524c",
            "placeholder": "​",
            "style": "IPY_MODEL_a34f89e6c3b74ef09a72a0d2cc4c0100",
            "value": "100%"
          }
        },
        "3744035230bb46e6b30dfe79386368ba": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "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_bbc2fd671a6541e3bc6542f2497d88a8",
            "max": 9912422,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_0599dcc80a0840a3bf4608970b612f49",
            "value": 9912422
          }
        },
        "c13a23b2a235491a8686aa83dad15aeb": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "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_ebd5b7a5cdfa4643832ffc14cfff7d3d",
            "placeholder": "​",
            "style": "IPY_MODEL_d3113b65e3634052a0cb8498cb5e5428",
            "value": " 9912422/9912422 [00:00&lt;00:00, 101796711.82it/s]"
          }
        },
        "2f4c0514d0314351a1c81815569b829c": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "236501633e0c4e128485a479533c524c": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "a34f89e6c3b74ef09a72a0d2cc4c0100": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "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": ""
          }
        },
        "bbc2fd671a6541e3bc6542f2497d88a8": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "0599dcc80a0840a3bf4608970b612f49": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "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": ""
          }
        },
        "ebd5b7a5cdfa4643832ffc14cfff7d3d": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "d3113b65e3634052a0cb8498cb5e5428": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "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": ""
          }
        },
        "1e3357a13e954cb794d9af14d7430149": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "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_2ef9ce1d8cf242eea15494074fa1d9b3",
              "IPY_MODEL_b5b81ae9091c4b97991c66e809b9d5b7",
              "IPY_MODEL_77c5f7a0074c43ba9db4d316c5af7fdf"
            ],
            "layout": "IPY_MODEL_5ab42f5d11e44cb6a3206392713b1227"
          }
        },
        "2ef9ce1d8cf242eea15494074fa1d9b3": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "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_be8f5d871fa94c849832c334112bb861",
            "placeholder": "​",
            "style": "IPY_MODEL_f28b61162f38421095653650f21d1ef3",
            "value": "100%"
          }
        },
        "b5b81ae9091c4b97991c66e809b9d5b7": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "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_d772cbe0a09a451bbb11f57eb51282c3",
            "max": 28881,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_f20b7645111148a5b5f6851a8179786f",
            "value": 28881
          }
        },
        "77c5f7a0074c43ba9db4d316c5af7fdf": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "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_3369290057a34f0c8e71d5f20f1b210c",
            "placeholder": "​",
            "style": "IPY_MODEL_b11ef81d248246169cb333d7eb14e50b",
            "value": " 28881/28881 [00:00&lt;00:00, 1115378.61it/s]"
          }
        },
        "5ab42f5d11e44cb6a3206392713b1227": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "be8f5d871fa94c849832c334112bb861": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "f28b61162f38421095653650f21d1ef3": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "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": ""
          }
        },
        "d772cbe0a09a451bbb11f57eb51282c3": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "f20b7645111148a5b5f6851a8179786f": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "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": ""
          }
        },
        "3369290057a34f0c8e71d5f20f1b210c": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "b11ef81d248246169cb333d7eb14e50b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "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": ""
          }
        },
        "8069d06f822e4c089d93fd3630916436": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "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_f973f0bb77a24150b05f8f8691938235",
              "IPY_MODEL_b1f76cf5b4294f0e9f017e4fa29725f1",
              "IPY_MODEL_c903a51b4d734ec2b6eacffeaedefe87"
            ],
            "layout": "IPY_MODEL_763fdefc59574b8885dbad9bd741a273"
          }
        },
        "f973f0bb77a24150b05f8f8691938235": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "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_85fff734d85b422787d6f8afb059531a",
            "placeholder": "​",
            "style": "IPY_MODEL_71ef36151be742ab85a52ad965855c3b",
            "value": "100%"
          }
        },
        "b1f76cf5b4294f0e9f017e4fa29725f1": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "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_22c042aa8bd2407b8f39eba1f4a24647",
            "max": 1648877,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_c7fc9bae0ba243aca087c138f9803831",
            "value": 1648877
          }
        },
        "c903a51b4d734ec2b6eacffeaedefe87": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "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_a3c66bf1da414cd3865fec61e9198150",
            "placeholder": "​",
            "style": "IPY_MODEL_f17fa29c78f7474eb6cc3a4c28ef2688",
            "value": " 1648877/1648877 [00:00&lt;00:00, 23151364.62it/s]"
          }
        },
        "763fdefc59574b8885dbad9bd741a273": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "85fff734d85b422787d6f8afb059531a": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "71ef36151be742ab85a52ad965855c3b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "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": ""
          }
        },
        "22c042aa8bd2407b8f39eba1f4a24647": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "c7fc9bae0ba243aca087c138f9803831": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "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": ""
          }
        },
        "a3c66bf1da414cd3865fec61e9198150": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "f17fa29c78f7474eb6cc3a4c28ef2688": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "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": ""
          }
        },
        "6c1402291ee6457395f9c57c7a1b1879": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "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_a3f206bbdca5419d8fa0eaaa857d6141",
              "IPY_MODEL_4a76435cdf3f4043994030db53307ef2",
              "IPY_MODEL_2a1aa0c194d34fa0acfc9db5c71ad7e6"
            ],
            "layout": "IPY_MODEL_df14db4dc27648c6b539218b24fe3da4"
          }
        },
        "a3f206bbdca5419d8fa0eaaa857d6141": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "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_31c6b15d5661455da031599655e66470",
            "placeholder": "​",
            "style": "IPY_MODEL_60915470b1804b5bbf5fe528509f909d",
            "value": "100%"
          }
        },
        "4a76435cdf3f4043994030db53307ef2": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "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_68be7d204c6b446f8f22e235201cb64b",
            "max": 4542,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_faf53a2d87114f66bdd181ef52bfddc2",
            "value": 4542
          }
        },
        "2a1aa0c194d34fa0acfc9db5c71ad7e6": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "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_242ca09ef3ca4384a8d1e647127bd99f",
            "placeholder": "​",
            "style": "IPY_MODEL_8e171fb98f8244d7aed3f27657b80992",
            "value": " 4542/4542 [00:00&lt;00:00, 216796.16it/s]"
          }
        },
        "df14db4dc27648c6b539218b24fe3da4": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "31c6b15d5661455da031599655e66470": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "60915470b1804b5bbf5fe528509f909d": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "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": ""
          }
        },
        "68be7d204c6b446f8f22e235201cb64b": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "faf53a2d87114f66bdd181ef52bfddc2": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "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": ""
          }
        },
        "242ca09ef3ca4384a8d1e647127bd99f": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "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
          }
        },
        "8e171fb98f8244d7aed3f27657b80992": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "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",
      "source": [
        "# Pytorch Tutorial \n",
        "Prepared by: Shayan Vassef, Email: sh.vassef\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "![Pytorch](https://drive.google.com/uc?export=view&id=14HmbCV22D052ywxAyu5j3vu1DUnrlqLj)"
      ],
      "metadata": {
        "id": "1Ay1mOL2bchU"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install torch"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BbSUxc9VljbM",
        "outputId": "86cdb749-77ce-4336-f711-a20a5dd1ef84"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: torch in /usr/local/lib/python3.8/dist-packages (1.13.0+cu116)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.8/dist-packages (from torch) (4.4.0)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Pytorch Basics**"
      ],
      "metadata": {
        "id": "MvgcgpigQNiG"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "qm97bTAQEQpg"
      },
      "outputs": [],
      "source": [
        "import torch"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.tensor([1,2,3])\n",
        "x"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rBexPvQdESJ6",
        "outputId": "0bbaa23b-f377-4d21-b446-8ccb9a3c58d4"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([1, 2, 3])"
            ]
          },
          "metadata": {},
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.rand(3)\n",
        "x"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nUBFm9T-EbCG",
        "outputId": "cb8f61b9-2d33-4f26-bee3-3a0f19bd48cc"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([0.4595, 0.7038, 0.1954])"
            ]
          },
          "metadata": {},
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.rand(3,5,dtype=torch.float16)\n",
        "x"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gd_d9mCREi57",
        "outputId": "97eff3f4-d973-436d-dee3-a86f3b3c371e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[0.3813, 0.1470, 0.7236, 0.1143, 0.7983],\n",
              "        [0.9834, 0.0015, 0.8564, 0.5391, 0.5967],\n",
              "        [0.8579, 0.0381, 0.7275, 0.2676, 0.9385]], dtype=torch.float16)"
            ]
          },
          "metadata": {},
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from torch._C import dtype\n",
        "import numpy as np\n",
        "arr = np.random.randn(3,4)\n",
        "x = torch.from_numpy(arr)\n",
        "x"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "siZp2XntEnlX",
        "outputId": "6bd268c5-0998-4ad9-cdbc-c6171f40962f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[-1.0008, -1.0003,  0.4646, -0.4147],\n",
              "        [-1.6396, -0.5620,  0.0696,  1.3053],\n",
              "        [-2.6480, -0.6226,  1.0265,  1.7258]], dtype=torch.float64)"
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x.type(torch.float16) "
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7QdGFOe2Ex-1",
        "outputId": "332f3e19-b9ff-42c6-e41d-479870f63ea7"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[-1.0010, -1.0000,  0.4646, -0.4148],\n",
              "        [-1.6396, -0.5620,  0.0696,  1.3057],\n",
              "        [-2.6484, -0.6226,  1.0264,  1.7256]], dtype=torch.float16)"
            ]
          },
          "metadata": {},
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.rand(3)\n",
        "print(x)\n",
        "y = torch.ones(3)\n",
        "x.add(y)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qMxN4zwBFTmf",
        "outputId": "14213ca5-9ded-4e0e-823e-30334a32f6b3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([0.7126, 0.3163, 0.3549])\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([1.7126, 1.3163, 1.3549])"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x.add_(y)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xiMB9nnNGqD-",
        "outputId": "f3633e76-d80f-42a2-ae37-cd12065c5840"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([1.7126, 1.3163, 1.3549])"
            ]
          },
          "metadata": {},
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x =  torch.empty(3,5)\n",
        "print(x)\n",
        "x.view(-1,3)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ZA31D4qDGwx2",
        "outputId": "d42675ce-98f0-4ab1-baeb-6f76f3f02edf"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[4.2350e-34, 0.0000e+00, 3.5032e-44, 0.0000e+00,        nan],\n",
            "        [1.6775e+00, 1.1578e+27, 7.1463e+22, 4.6241e+30, 1.0552e+24],\n",
            "        [5.5757e-02, 1.8728e+31, 5.9571e-02, 7.0374e+22, 2.7558e-09]])\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[4.2350e-34, 0.0000e+00, 3.5032e-44],\n",
              "        [0.0000e+00,        nan, 1.6775e+00],\n",
              "        [1.1578e+27, 7.1463e+22, 4.6241e+30],\n",
              "        [1.0552e+24, 5.5757e-02, 1.8728e+31],\n",
              "        [5.9571e-02, 7.0374e+22, 2.7558e-09]])"
            ]
          },
          "metadata": {},
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# first, change the runtime to gpu\n",
        "if torch.cuda.is_available():\n",
        "  device = 'cuda' #torch.device(\"cuda\")\n",
        "  x = torch.ones(5, device = device)\n",
        "  # or\n",
        "  y = torch.rand(5).to(device)\n",
        "  z = x + y\n",
        "  print(z)\n",
        "  # return back to cpu\n",
        "  print(z.to(\"cpu\").numpy()) # produce an error\n",
        "  "
      ],
      "metadata": {
        "id": "v_8XkVWaG3nY",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f45790d3-d2af-42be-caae-0c1b15b2e637"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([1.2063, 1.1087, 1.1052, 1.3392, 1.5358], device='cuda:0')\n",
            "[1.2062893 1.1086557 1.1052477 1.3392262 1.5358253]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Gradient**"
      ],
      "metadata": {
        "id": "qH9Tb-3aQS0V"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Vector differentiation"
      ],
      "metadata": {
        "id": "BXJgB9o9GJHS"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "![part1](https://drive.google.com/uc?export=view&id=1u2dcTLn2v5kNH4fy0sK910GLUPiWTUgL)\n",
        "\n",
        "![part2](https://drive.google.com/uc?export=view&id=1Ynr_vXxLqsB5g6f0kHn40eXsvvkmsNr0)"
      ],
      "metadata": {
        "id": "KsyGYDrtBuVV"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 1*1 tensor"
      ],
      "metadata": {
        "id": "ncKhaosTRosm"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Z = z(1,3) * I(3,1)\n",
        "D Z / D x = D (z(1,3) * I(3,1))/ D x = D z / D x * I"
      ],
      "metadata": {
        "id": "wvGgMStErglq"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.rand(3,requires_grad=True)\n",
        "print(x)\n",
        "y = x + 2\n",
        "z= 2 * y**2\n",
        "z = z.mean() \n",
        "z"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xi2Y8EAmQYaI",
        "outputId": "c05f3f90-0871-40e6-a51f-34c70359a56a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([0.5348, 0.8765, 0.3493], requires_grad=True)\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor(13.4789, grad_fn=<MeanBackward0>)"
            ]
          },
          "metadata": {},
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "z.backward() # Calculate dz/dx"
      ],
      "metadata": {
        "id": "-xQ0xf89RDHh"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x.grad"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Uoox5HE-RHuj",
        "outputId": "6fe19f7f-0531-4fa5-913b-df0979d4bdfb"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([3.3797, 3.8353, 3.1324])"
            ]
          },
          "metadata": {},
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.rand(3,requires_grad=True)\n",
        "y = x + 2\n",
        "z= 2 * y**2\n",
        "\n",
        "z.backward()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 345
        },
        "id": "5zJMGjx8RKow",
        "outputId": "ac93e190-3d3a-4566-e5b5-0b8171c952cc"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "error",
          "ename": "RuntimeError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-14-c770a8d8474d>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0mz\u001b[0m\u001b[0;34m=\u001b[0m \u001b[0;36m2\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 5\u001b[0;31m \u001b[0mz\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;32m/usr/local/lib/python3.8/dist-packages/torch/_tensor.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(self, gradient, retain_graph, create_graph, inputs)\u001b[0m\n\u001b[1;32m    485\u001b[0m                 \u001b[0minputs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    486\u001b[0m             )\n\u001b[0;32m--> 487\u001b[0;31m         torch.autograd.backward(\n\u001b[0m\u001b[1;32m    488\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgradient\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    489\u001b[0m         )\n",
            "\u001b[0;32m/usr/local/lib/python3.8/dist-packages/torch/autograd/__init__.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(tensors, grad_tensors, retain_graph, create_graph, grad_variables, inputs)\u001b[0m\n\u001b[1;32m    188\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    189\u001b[0m     \u001b[0mgrad_tensors_\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_tensor_or_tensors_to_tuple\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgrad_tensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensors\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 190\u001b[0;31m     \u001b[0mgrad_tensors_\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_make_grads\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad_tensors_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mis_grads_batched\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    191\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mretain_graph\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    192\u001b[0m         \u001b[0mretain_graph\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.8/dist-packages/torch/autograd/__init__.py\u001b[0m in \u001b[0;36m_make_grads\u001b[0;34m(outputs, grads, is_grads_batched)\u001b[0m\n\u001b[1;32m     83\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mout\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrequires_grad\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     84\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0mout\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnumel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 85\u001b[0;31m                     \u001b[0;32mraise\u001b[0m \u001b[0mRuntimeError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"grad can be implicitly created only for scalar outputs\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     86\u001b[0m                 \u001b[0mnew_grads\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mones_like\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mout\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmemory_format\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpreserve_format\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     87\u001b[0m             \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mRuntimeError\u001b[0m: grad can be implicitly created only for scalar outputs"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 1*n tensor"
      ],
      "metadata": {
        "id": "UBPoRApcR4pu"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.ones(2,3,requires_grad=True)\n",
        "y = x + 2\n",
        "z = 2 * y**2 \n",
        "\n",
        "v = torch.tensor([[0.1 , 0.5 , 0.3 ],[0.1,0.5,0.3]]) # weighted sum (f =  z * v.t) \n",
        "z.backward(v)\n",
        "\n",
        "x.grad"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "k5JRRzcbRzNb",
        "outputId": "bcbaec32-23e7-4540-b02b-68d29535abab"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[1.2000, 6.0000, 3.6000],\n",
              "        [1.2000, 6.0000, 3.6000]])"
            ]
          },
          "metadata": {},
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## prevent pytorch from tracking gradient\n",
        "  * x.requires_grad = False\n",
        "  * x.detach()\n",
        "  * with torch.no_grad()\n",
        "  "
      ],
      "metadata": {
        "id": "FENaGb2MTFln"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.rand(3,requires_grad=True)\n",
        "print(x)\n",
        "x.requires_grad_(False) # No longer tracking"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TNKvGlnvSvBh",
        "outputId": "094105f8-5ef0-4257-cac7-dc687cead898"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([0.8616, 0.9201, 0.4289], requires_grad=True)\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([0.8616, 0.9201, 0.4289])"
            ]
          },
          "metadata": {},
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.rand(3,requires_grad=True)\n",
        "print(x)\n",
        "y = x.detach() # No longer tracking\n",
        "print(y)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "d5Ch1_KeUGX4",
        "outputId": "8daaaf90-d6a2-4266-c398-3d3c4499505b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([0.4289, 0.2454, 0.7642], requires_grad=True)\n",
            "tensor([0.4289, 0.2454, 0.7642])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.rand(3,requires_grad=True)\n",
        "print(x)\n",
        "with torch.no_grad():\n",
        "  y = x + 3\n",
        "print(y)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Gv-6gtjuUZVl",
        "outputId": "a6496cbc-e9b5-4d69-d789-217771a6cb6a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([0.1024, 0.5840, 0.7556], requires_grad=True)\n",
            "tensor([3.1024, 3.5840, 3.7556])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## .grad will sum the previously calculated gradient to the actual ones"
      ],
      "metadata": {
        "id": "Bbva4hIJVjhv"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "w = torch.ones(4 , requires_grad = True)\n",
        "for noe in range(3):\n",
        "  y = (w*3).sum()\n",
        "  y.backward()\n",
        "  print(w.grad)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nZtmqyQgVu7t",
        "outputId": "e2449caa-635d-478e-a408-60d70704adaa"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([3., 3., 3., 3.])\n",
            "tensor([6., 6., 6., 6.])\n",
            "tensor([9., 9., 9., 9.])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "w = torch.ones(4 , requires_grad = True)\n",
        "for noe in range(3):\n",
        "  y = (w*3).sum()\n",
        "  y.backward()\n",
        "  print(w.grad)\n",
        "  w.grad.zero_()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "am-zT6jtWH63",
        "outputId": "c344dd48-6255-4e36-da8d-36fbe253666b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([3., 3., 3., 3.])\n",
            "tensor([3., 3., 3., 3.])\n",
            "tensor([3., 3., 3., 3.])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Backprobagation**"
      ],
      "metadata": {
        "id": "DUXOe3R9a7mO"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.tensor(1,dtype=torch.float32)\n",
        "\n",
        "y = torch.tensor(2,dtype=torch.float32)\n",
        "\n",
        "w = torch.tensor(1,dtype=torch.float32,requires_grad=True) \n",
        "\n",
        "yhat = w * x\n",
        "loss = (y - yhat)**2\n",
        "print(loss)\n",
        "\n",
        "loss.backward() # d loss / d w\n",
        "print(w.grad)\n",
        "\n",
        "## next, update weights and then start forwarding , ..."
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lHbZ2_lkWQVF",
        "outputId": "73c2d4d6-08d6-4951-a7ce-6c9822e69740"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor(1., grad_fn=<PowBackward0>)\n",
            "tensor(-2.)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Implementing all in pytorch"
      ],
      "metadata": {
        "id": "npqbyrCuISU1"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "\n",
        "x = torch.tensor([1,2,3,4] , dtype = torch.float32).view(4,-1)\n",
        "y = torch.tensor([2,4,6,8] , dtype = torch.float32).view(4,-1)\n",
        "x_test = torch.tensor([5],dtype=torch.float32) \n",
        "\n",
        "num_sam , num_feat = x.size() # (row, column)\n",
        "\n",
        "print(num_sam , num_feat) \n",
        "\n",
        "input_size = num_feat # refer to input dimention\n",
        "output_size = num_feat # refer to output dimention\n",
        "\n",
        "model = nn.Linear(input_size , output_size) \n",
        "\n",
        "print(f\"Model's prediction before training for x = 5: {model(x_test).item():.3f}\\n\")\n",
        "\n",
        "lr = 0.01\n",
        "epochs = 100\n",
        "   \n",
        "loss = nn.MSELoss() # creat an instance of MSE class\n",
        "optimizer = torch.optim.SGD(model.parameters(),lr = lr)\n",
        "\n",
        "for epoch in range(epochs):\n",
        "  # forwarding\n",
        "  y_pred = model(x)\n",
        "  # loss\n",
        "  l = loss(y,y_pred)\n",
        "  # backward\n",
        "  l.backward()\n",
        "  #optimizer => updating weights\n",
        "  optimizer.step()\n",
        "  #reset gradient\n",
        "  optimizer.zero_grad()\n",
        "\n",
        "  if epoch % 10 ==0 :\n",
        "    w , b = model.parameters()\n",
        "    print(f\"Epoch : {epoch} , w : {w[0][0].item()} , loss : {l}\\n\")\n",
        "\n",
        "print(f'Model prediction aftre training for an arbitrary input : {x_test.item()} is {model(x_test).item():.3f}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JOBiC19Ge-4I",
        "outputId": "a5c442ee-2cf1-450e-df07-59f887f35005"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "4 1\n",
            "Model's prediction before training for x = 5: 4.312\n",
            "\n",
            "Epoch : 0 , w : 1.1244351863861084 , loss : 10.70458984375\n",
            "\n",
            "Epoch : 10 , w : 1.8731905221939087 , loss : 0.27735230326652527\n",
            "\n",
            "Epoch : 20 , w : 1.9932128190994263 , loss : 0.007551664020866156\n",
            "\n",
            "Epoch : 30 , w : 2.0121161937713623 , loss : 0.0005494178622029722\n",
            "\n",
            "Epoch : 40 , w : 2.0147664546966553 , loss : 0.00034765468444675207\n",
            "\n",
            "Epoch : 50 , w : 2.0148143768310547 , loss : 0.0003230239963158965\n",
            "\n",
            "Epoch : 60 , w : 2.0144546031951904 , loss : 0.00030410438193939626\n",
            "\n",
            "Epoch : 70 , w : 2.014039993286133 , loss : 0.00028640052187256515\n",
            "\n",
            "Epoch : 80 , w : 2.013627290725708 , loss : 0.0002697323798201978\n",
            "\n",
            "Epoch : 90 , w : 2.0132250785827637 , loss : 0.0002540323475841433\n",
            "\n",
            "Model prediction aftre training for an arbitrary input : 5.0 is 10.027\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(model(x_test))\n",
        "print(model(x_test).detach())\n",
        "print(model(x_test).item())"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QET2t4QtzcYw",
        "outputId": "19d6847a-00cc-4bea-ac76-7cf413fedee0"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([10.0265], grad_fn=<AddBackward0>)\n",
            "tensor([10.0265])\n",
            "10.02651596069336\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Define by nn.Module"
      ],
      "metadata": {
        "id": "cbBbfGCOSCPW"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Class LinearRegression:\n",
        "  def __init__(self, input_size, output_size, ...):\n",
        "    self. input_size = input_size\n",
        "    ....\n",
        "  def Att.1 (...):\n",
        "\n",
        "\n",
        "Class Virus:\n",
        "  .....\n",
        "\n",
        "Class Corona_Virus(Virus):\n",
        "  .....\n"
      ],
      "metadata": {
        "id": "IvqXiiF95VHQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "\n",
        "x = torch.tensor([1,2,3,4] , dtype = torch.float32).view(4,-1)\n",
        "y = torch.tensor([2,4,6,8] , dtype = torch.float32).view(4,-1)\n",
        "x_test = torch.tensor([5],dtype=torch.float32)\n",
        "\n",
        "num_sam , num_feat = x.size()\n",
        "\n",
        "print(num_sam , num_feat)\n",
        "\n",
        "input_size = num_feat # refer to inpt dimention\n",
        "output_size = num_feat # refer to output dimention\n",
        "\n",
        "class LinearRegression(nn.Module):\n",
        "  def __init__(self , In_size , O_size ):\n",
        "    super(LinearRegression, self).__init__() # super().forward # nn.Module.forward()\n",
        "    # define layes\n",
        "    self.lin = nn.Linear(In_size , O_size)\n",
        "\n",
        "  def forward(self , x):\n",
        "    return self.lin(x)\n",
        "\n",
        "model = LinearRegression(input_size , output_size)\n",
        "print(f\"Model's prediction before training for x = 5: {model(x_test).item():.3f}\\n\")\n",
        "\n",
        "lr = 0.01\n",
        "epochs = 100\n",
        "\n",
        "loss = nn.MSELoss() # creat an instance of MSE class\n",
        "optimizer = torch.optim.SGD(model.parameters(),lr = lr)\n",
        "\n",
        "for epoch in range(epochs):\n",
        "  # forwarding\n",
        "  y_pred = model(x)\n",
        "  # loss\n",
        "  l = loss(y,y_pred)\n",
        "  # backward\n",
        "  l.backward()\n",
        "  #optimizer => updating weights\n",
        "  optimizer.step()\n",
        "  #reset gradient\n",
        "  optimizer.zero_grad()\n",
        "\n",
        "  if epoch % 10 ==0 :\n",
        "    w , b = model.parameters()\n",
        "    print(f\"Epoch : {epoch} , w : {w[0][0].item()} , loss : {l}\\n\")\n",
        "\n",
        "print(f'Model prediction aftre training for an arbitrary input : {x_test.item()} is {model(x_test).item():.3f}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_ZRY5o2lH3RH",
        "outputId": "c9ebb941-1641-4c6c-ce20-27414873812d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "4 1\n",
            "Model's prediction before training for x = 5: -2.654\n",
            "\n",
            "Epoch : 0 , w : -0.30139246582984924 , loss : 44.71574783325195\n",
            "\n",
            "Epoch : 10 , w : 1.2384709119796753 , loss : 1.4677343368530273\n",
            "\n",
            "Epoch : 20 , w : 1.4977164268493652 , loss : 0.33071279525756836\n",
            "\n",
            "Epoch : 30 , w : 1.550634503364563 , loss : 0.2842564582824707\n",
            "\n",
            "Epoch : 40 , w : 1.5700336694717407 , loss : 0.26700717210769653\n",
            "\n",
            "Epoch : 50 , w : 1.5837194919586182 , loss : 0.2514478862285614\n",
            "\n",
            "Epoch : 60 , w : 1.5961744785308838 , loss : 0.23681189119815826\n",
            "\n",
            "Epoch : 70 , w : 1.608128547668457 , loss : 0.2230280339717865\n",
            "\n",
            "Epoch : 80 , w : 1.6197081804275513 , loss : 0.2100467085838318\n",
            "\n",
            "Epoch : 90 , w : 1.6309422254562378 , loss : 0.19782082736492157\n",
            "\n",
            "Model prediction aftre training for an arbitrary input : 5.0 is 9.260\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## **Logistic regression**"
      ],
      "metadata": {
        "id": "W_WZfiVVKaX8"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import numpy as np\n",
        "from sklearn import datasets\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "from sklearn.model_selection import train_test_split"
      ],
      "metadata": {
        "id": "bHC6Xbm5Kctw"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "  * 0 : prepare the data\n",
        "  * 1 : model\n",
        "  * 2 : loss , optimizer\n",
        "  * 3 : training loop\n",
        "  "
      ],
      "metadata": {
        "id": "Ctaotx6hMDtG"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#step 0\n",
        "bc = datasets.load_breast_cancer()\n",
        "x , y = bc.data , bc.target\n",
        "\n",
        "n_samples , n_features = x.shape\n",
        "print(n_samples , n_features)\n",
        "\n",
        "x_train , x_test , y_train , y_test = train_test_split(x , y , test_size=0.25 , random_state=111)\n",
        "\n",
        "sc = StandardScaler()\n",
        "x_train = sc.fit_transform(x_train)\n",
        "x_test = sc.transform(x_test)\n",
        "\n",
        "x_train = torch.from_numpy(x_train.astype(np.float32))\n",
        "y_train = torch.from_numpy(y_train.astype(np.float32))\n",
        "x_test = torch.from_numpy(x_test.astype(np.float32))\n",
        "y_test = torch.from_numpy(y_test.astype(np.float32))\n",
        "\n",
        "y_train = y_train.view(y_train.shape[0],1) # Converting to one column\n",
        "y_test = y_test.view(y_test.shape[0],1) # Converting to one column\n",
        "# step 1, model\n",
        "# f = w.x , apply sigmoid at the end\n",
        "class LogisticRegression(nn.Module):\n",
        "  def __init__(self,n_input_features):\n",
        "    super(LogisticRegression , self).__init__()\n",
        "    self.lin = nn.Linear(n_input_features , 1)\n",
        "  def forward(self , x):\n",
        "    return torch.sigmoid(self.lin(x))\n",
        "\n",
        "model = LogisticRegression(n_features)\n",
        "\n",
        "# 2 : define loss and optimizer\n",
        "loss = nn.BCELoss()\n",
        "optimizer = torch.optim.SGD(model.parameters() , lr = 0.01)\n",
        "# 3 : train the model\n",
        "n_epcohs = 100\n",
        "for epoch in range(n_epcohs):\n",
        "  y_train_pred  = model(x_train)\n",
        "  # define loss\n",
        "  l = loss(y_train_pred , y_train)\n",
        "  # backprog\n",
        "  l.backward()\n",
        "  #update gd\n",
        "  optimizer.step()\n",
        "\n",
        "  optimizer.zero_grad()\n",
        "\n",
        "  with torch.no_grad():\n",
        "    y_test_pred = model(x_test)\n",
        "    y_np = y_test_pred.numpy() # y_test_pred_cls = y_test_pred.round()\n",
        "    label = (y_np >= 0.5).astype(np.float32) \n",
        "    acc = np.round(np.sum((label == y_test.numpy())/label.shape[0])*100,2) # acc = y_test_pred_cls.eq(y_test).sum()/float(y_test.shape[0])*100\n",
        "  if (epoch+1) % 10 ==0:\n",
        "    print(f\"Training for => Epoch : {epoch+1} ,  loss : {l.item():.4f}\")\n",
        "    print(f\"The Accuracy on the test data on Epoch {epoch +1 } is : {acc}%\\n\")\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5DB0UUfiLxJS",
        "outputId": "832074ee-a28b-4d3f-b594-69cd6aa645db"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "569 30\n",
            "Training for => Epoch : 10 ,  loss : 0.6142\n",
            "The Accuracy on the test data on Epoch 10 is : 67.83%\n",
            "\n",
            "Training for => Epoch : 20 ,  loss : 0.5109\n",
            "The Accuracy on the test data on Epoch 20 is : 85.31%\n",
            "\n",
            "Training for => Epoch : 30 ,  loss : 0.4447\n",
            "The Accuracy on the test data on Epoch 30 is : 88.11%\n",
            "\n",
            "Training for => Epoch : 40 ,  loss : 0.3984\n",
            "The Accuracy on the test data on Epoch 40 is : 89.51%\n",
            "\n",
            "Training for => Epoch : 50 ,  loss : 0.3642\n",
            "The Accuracy on the test data on Epoch 50 is : 90.21%\n",
            "\n",
            "Training for => Epoch : 60 ,  loss : 0.3376\n",
            "The Accuracy on the test data on Epoch 60 is : 91.61%\n",
            "\n",
            "Training for => Epoch : 70 ,  loss : 0.3163\n",
            "The Accuracy on the test data on Epoch 70 is : 92.31%\n",
            "\n",
            "Training for => Epoch : 80 ,  loss : 0.2988\n",
            "The Accuracy on the test data on Epoch 80 is : 92.31%\n",
            "\n",
            "Training for => Epoch : 90 ,  loss : 0.2840\n",
            "The Accuracy on the test data on Epoch 90 is : 93.01%\n",
            "\n",
            "Training for => Epoch : 100 ,  loss : 0.2714\n",
            "The Accuracy on the test data on Epoch 100 is : 93.01%\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Dataset and DataLoader**"
      ],
      "metadata": {
        "id": "lT1NT-_B2I7U"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Dataset"
      ],
      "metadata": {
        "id": "toEjkX1P8p8U"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torchvision\n",
        "from torch.utils.data import Dataset , DataLoader\n",
        "import numpy as np\n",
        "import math\n"
      ],
      "metadata": {
        "id": "AzrUZTmul706"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from sklearn.datasets import load_wine\n",
        "data = load_wine()"
      ],
      "metadata": {
        "id": "pYRXuo97RknX"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(data.data.shape)\n",
        "print(data.target.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "A_-m61nLRnXI",
        "outputId": "8f8d85b6-1bd5-4927-fe68-2f119e22debe"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(178, 13)\n",
            "(178,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "type(data.data)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zmhcKaEWXca9",
        "outputId": "4e02e0f3-23b9-44eb-ade4-607e2c04bcc3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "numpy.ndarray"
            ]
          },
          "metadata": {},
          "execution_count": 30
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class WineDataset(Dataset):\n",
        "  def __init__(self):\n",
        "    #data loading\n",
        "    #xy = np.loadtxt('wine.csv' , delimiter= \",\" , dtype = np.float32 , skiprows=1) # ignore the header\n",
        "    data = load_wine()\n",
        "    self.x = torch.from_numpy(data.data)\n",
        "    self.y = torch.from_numpy(data.target) # [0] => reshaped already to accepted format\n",
        "    self.n_samples = data.data.shape[0]\n",
        "  def __getitem__(self, index):\n",
        "    #indexing\n",
        "      return self.x[index] , self.y[index]\n",
        "  def __len__(self):\n",
        "    return self.n_samples\n",
        "\n",
        "\n",
        "dataset = WineDataset()\n",
        "first_data = dataset[0]\n",
        "features , labels = first_data\n",
        "print(features , labels)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OJr5mfZY4O0F",
        "outputId": "fbf2cc09-0f20-4f66-d1f8-44b26559c363"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([1.4230e+01, 1.7100e+00, 2.4300e+00, 1.5600e+01, 1.2700e+02, 2.8000e+00,\n",
            "        3.0600e+00, 2.8000e-01, 2.2900e+00, 5.6400e+00, 1.0400e+00, 3.9200e+00,\n",
            "        1.0650e+03], dtype=torch.float64) tensor(0)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## DataLoader"
      ],
      "metadata": {
        "id": "ZOTH_quv8s6D"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "bt = 5\n",
        "dataloader = DataLoader(dataset , batch_size = bt ,shuffle = True , num_workers = 2)\n",
        "dataiter = iter(dataloader)\n",
        "data = next(dataiter)\n",
        "features , labels  = data\n",
        "print(features , \"\\n\" , labels )"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Zy6J1XZi6aUN",
        "outputId": "bd04c7f4-7e60-4558-c97c-eee45a16f826"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[1.3840e+01, 4.1200e+00, 2.3800e+00, 1.9500e+01, 8.9000e+01, 1.8000e+00,\n",
            "         8.3000e-01, 4.8000e-01, 1.5600e+00, 9.0100e+00, 5.7000e-01, 1.6400e+00,\n",
            "         4.8000e+02],\n",
            "        [1.2420e+01, 1.6100e+00, 2.1900e+00, 2.2500e+01, 1.0800e+02, 2.0000e+00,\n",
            "         2.0900e+00, 3.4000e-01, 1.6100e+00, 2.0600e+00, 1.0600e+00, 2.9600e+00,\n",
            "         3.4500e+02],\n",
            "        [1.4830e+01, 1.6400e+00, 2.1700e+00, 1.4000e+01, 9.7000e+01, 2.8000e+00,\n",
            "         2.9800e+00, 2.9000e-01, 1.9800e+00, 5.2000e+00, 1.0800e+00, 2.8500e+00,\n",
            "         1.0450e+03],\n",
            "        [1.4750e+01, 1.7300e+00, 2.3900e+00, 1.1400e+01, 9.1000e+01, 3.1000e+00,\n",
            "         3.6900e+00, 4.3000e-01, 2.8100e+00, 5.4000e+00, 1.2500e+00, 2.7300e+00,\n",
            "         1.1500e+03],\n",
            "        [1.4370e+01, 1.9500e+00, 2.5000e+00, 1.6800e+01, 1.1300e+02, 3.8500e+00,\n",
            "         3.4900e+00, 2.4000e-01, 2.1800e+00, 7.8000e+00, 8.6000e-01, 3.4500e+00,\n",
            "         1.4800e+03]], dtype=torch.float64) \n",
            " tensor([2, 1, 0, 0, 0])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Now creat a training loop\n",
        "n_epochs = 10\n",
        "total_samples = len(dataset)\n",
        "n_iterations = math.ceil(total_samples / bt)-1\n",
        "print(total_samples , n_iterations ) \n",
        "\n",
        "for epoch in range(n_epochs):\n",
        "  for i , (inputs , labels) in enumerate(dataloader):\n",
        "    #forward , backward , updating\n",
        "    if (i+1) % 5 ==0:\n",
        "      print(f\"epoch : {epoch+1}/{n_epochs} , step : {i+1}/{n_iterations} , inputs shape : {inputs.shape}\")"
      ],
      "metadata": {
        "id": "qN8RxVmC9cES",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f2ff4a33-0179-48b7-edc4-c75d160475df"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "178 35\n",
            "epoch : 1/10 , step : 5/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 1/10 , step : 10/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 1/10 , step : 15/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 1/10 , step : 20/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 1/10 , step : 25/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 1/10 , step : 30/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 1/10 , step : 35/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 2/10 , step : 5/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 2/10 , step : 10/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 2/10 , step : 15/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 2/10 , step : 20/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 2/10 , step : 25/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 2/10 , step : 30/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 2/10 , step : 35/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 3/10 , step : 5/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 3/10 , step : 10/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 3/10 , step : 15/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 3/10 , step : 20/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 3/10 , step : 25/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 3/10 , step : 30/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 3/10 , step : 35/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 4/10 , step : 5/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 4/10 , step : 10/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 4/10 , step : 15/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 4/10 , step : 20/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 4/10 , step : 25/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 4/10 , step : 30/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 4/10 , step : 35/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 5/10 , step : 5/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 5/10 , step : 10/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 5/10 , step : 15/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 5/10 , step : 20/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 5/10 , step : 25/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 5/10 , step : 30/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 5/10 , step : 35/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 6/10 , step : 5/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 6/10 , step : 10/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 6/10 , step : 15/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 6/10 , step : 20/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 6/10 , step : 25/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 6/10 , step : 30/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 6/10 , step : 35/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 7/10 , step : 5/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 7/10 , step : 10/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 7/10 , step : 15/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 7/10 , step : 20/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 7/10 , step : 25/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 7/10 , step : 30/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 7/10 , step : 35/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 8/10 , step : 5/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 8/10 , step : 10/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 8/10 , step : 15/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 8/10 , step : 20/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 8/10 , step : 25/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 8/10 , step : 30/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 8/10 , step : 35/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 9/10 , step : 5/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 9/10 , step : 10/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 9/10 , step : 15/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 9/10 , step : 20/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 9/10 , step : 25/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 9/10 , step : 30/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 9/10 , step : 35/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 10/10 , step : 5/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 10/10 , step : 10/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 10/10 , step : 15/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 10/10 , step : 20/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 10/10 , step : 25/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 10/10 , step : 30/35 , inputs shape : torch.Size([5, 13])\n",
            "epoch : 10/10 , step : 35/35 , inputs shape : torch.Size([5, 13])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Famous dataset\n",
        "#torchvision.datasets.MNIST()"
      ],
      "metadata": {
        "id": "XhbGjrf9LL1g"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# define a dataset class\n",
        "class WineDataset(Dataset):\n",
        "  def __init__(self ,  transform = None):\n",
        "    #data loading\n",
        "    # xy = np.loadtxt('wine.csv' , delimiter= \",\" , dtype = np.float32 , skiprows=1) # ignore the header\n",
        "\n",
        "    # self.x = xy[:,1:]\n",
        "    # self.y = xy[:,[0]] # [0] => reshaped already to accepted format\n",
        "    data = load_wine()\n",
        "    self.x = data.data\n",
        "    self.y = data.target.reshape(-1,1)# [0] => reshaped already to accepted format\n",
        "    self.n_samples = data.data.shape[0]\n",
        "  \n",
        "    #self.n_samples = xy.shape[0]\n",
        "\n",
        "    self.transform = transform\n",
        "\n",
        "  def __getitem__(self, index):\n",
        "    #indexing\n",
        "      sample = self.x[index] , self.y[index]\n",
        "\n",
        "      if self.transform :\n",
        "        sample = self.transform(sample)\n",
        "\n",
        "      return sample \n",
        "      \n",
        "  def __len__(self):\n",
        "    return self.n_samples"
      ],
      "metadata": {
        "id": "Q5qsJDPBL51p"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Create \"ToTensor\" transform, to transform the data from Numpy into Tensor"
      ],
      "metadata": {
        "id": "YUmD3fqPY61b"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "class ToTensor:\n",
        "  def __call__(self, sample) :\n",
        "      inputs , targets = sample\n",
        "      return torch.from_numpy(inputs) , torch.from_numpy(targets)\n",
        "\n",
        "\n",
        "dataset = WineDataset() \n",
        "first_data = dataset[0]\n",
        "features_ntf , labels_ntf = first_data\n",
        "print(features_ntf , \"\\n\" , labels_ntf)\n",
        "\n",
        "\n",
        "dataset = WineDataset(transform = ToTensor()) \n",
        "first_data = dataset[0]\n",
        "features_tf , labels_tf = first_data\n",
        "print(features_tf , labels_tf)    "
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "udkBl_OjVuwh",
        "outputId": "056ed5eb-efb9-4c00-9f99-d67ded326d76"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[1.423e+01 1.710e+00 2.430e+00 1.560e+01 1.270e+02 2.800e+00 3.060e+00\n",
            " 2.800e-01 2.290e+00 5.640e+00 1.040e+00 3.920e+00 1.065e+03] \n",
            " [0]\n",
            "tensor([1.4230e+01, 1.7100e+00, 2.4300e+00, 1.5600e+01, 1.2700e+02, 2.8000e+00,\n",
            "        3.0600e+00, 2.8000e-01, 2.2900e+00, 5.6400e+00, 1.0400e+00, 3.9200e+00,\n",
            "        1.0650e+03], dtype=torch.float64) tensor([0])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Creat MulTransform, to scale input data "
      ],
      "metadata": {
        "id": "Zg88YiKkZR7p"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "class MulTransform:\n",
        "  def __init__(self, scale):\n",
        "    self.scale = scale \n",
        "  def __call__(self , sample):\n",
        "    inputs , targets = sample\n",
        "    inputs *= self.scale\n",
        "    return inputs , targets"
      ],
      "metadata": {
        "id": "599VoIvNXvAy"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# apply the both constructed transforms\n",
        "composed = torchvision.transforms.Compose([ToTensor() , MulTransform(3)])\n",
        "dataset = WineDataset(transform = composed)\n",
        "first_data = dataset[0]\n",
        "features_cmp , labels_cmp = first_data\n",
        "print(features_tf , \"\\n\" , labels_tf , \"\\n\")\n",
        "print(features_cmp , \"\\n\" , labels_cmp)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "vuXC0nMDaP3X",
        "outputId": "e3078610-1ed2-4537-eb74-27a2c404a606"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([1.4230e+01, 1.7100e+00, 2.4300e+00, 1.5600e+01, 1.2700e+02, 2.8000e+00,\n",
            "        3.0600e+00, 2.8000e-01, 2.2900e+00, 5.6400e+00, 1.0400e+00, 3.9200e+00,\n",
            "        1.0650e+03], dtype=torch.float64) \n",
            " tensor([0]) \n",
            "\n",
            "tensor([4.2690e+01, 5.1300e+00, 7.2900e+00, 4.6800e+01, 3.8100e+02, 8.4000e+00,\n",
            "        9.1800e+00, 8.4000e-01, 6.8700e+00, 1.6920e+01, 3.1200e+00, 1.1760e+01,\n",
            "        3.1950e+03], dtype=torch.float64) \n",
            " tensor([0])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **soft max , cross entropy**"
      ],
      "metadata": {
        "id": "nvKtMNLpcMAb"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Implementing softmax "
      ],
      "metadata": {
        "id": "SpQxOPuzrw6I"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### by numpy"
      ],
      "metadata": {
        "id": "ZJnB7IFVtaQh"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "\n",
        "def softmax(x):\n",
        "  return np.exp(x)/np.sum(np.exp(x))\n",
        "\n",
        "x = [0.1 , 1 , 2]\n",
        "print(softmax(x))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ueYCaMyibZbs",
        "outputId": "299c6b07-5bb4-4edd-c64c-1df881c97a1b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[0.09856589 0.24243297 0.65900114]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "###  by pytorch"
      ],
      "metadata": {
        "id": "au0HB_ZvsgYK"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "\n",
        "x = torch.tensor([0.1 , 1 , 2])\n",
        "print(torch.softmax(x , dim = 0))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "poPLft1Xsf1J",
        "outputId": "95017a2a-fe2d-4d01-e2f2-9ef84483947c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([0.0986, 0.2424, 0.6590])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Implementing Cross Entropy"
      ],
      "metadata": {
        "id": "jHG4g4EEtffJ"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### by numpy"
      ],
      "metadata": {
        "id": "8in3U-MytuLx"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def cross_entropy(actual , predicted):\n",
        "  # actual must be one-hot encoded\n",
        "  return -np.sum(actual * np.log(predicted))\n",
        "\n",
        "# class 1 => y = [1 0 0]\n",
        "# class 2 => y = [0 1 0]\n",
        "# class 3 => y = [0 0 1]\n",
        "\n",
        "# good prediction => y^ = [0.7 ,0.2 ,0.1]\n",
        "# bad  prediction => y^ = [0.1 ,0.5 ,0.4]\n",
        "y = [1, 0, 0] # Class 0(1)\n",
        "y_g_pred = [0.7 ,0.2 ,0.1]\n",
        "y_b_pred = [0.1 ,0.5 ,0.4]\n",
        "print(f\"The good prediction cross entropy loss is {cross_entropy(y , y_g_pred):.2f}\\n\")\n",
        "print(f\"The bad prediction cross entropy loss is {cross_entropy(y , y_b_pred):.2f}\\n\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OKhnMf4SsQ4b",
        "outputId": "2ca51515-1e4a-4788-aeea-b44c35807265"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "The good prediction cross entropy loss is 0.36\n",
            "\n",
            "The bad prediction cross entropy loss is 2.30\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### by pytorch\n",
        "### some important notes concerning with nn.CrossEntropyLoss from pyorch:\n",
        "  * pytorch will apply nn.LogSoftmax() + nn.NNLLoss(negative log-likelihood loss)\n",
        "  * The true label(y) should'nt be defined as one-hot, should be the coresponding label\n",
        "  * The predicted y, is the model raw output. Ex : [0.1 , 2 , 3]"
      ],
      "metadata": {
        "id": "QvSD-fhUwKJJ"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Single sample"
      ],
      "metadata": {
        "id": "6nWLWJ0VVlnQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "loss = nn.CrossEntropyLoss()\n",
        "y = torch.tensor([1])\n",
        "y_pred_good = torch.tensor([[0.2, 4, 1]])\n",
        "y_pred_bad = torch.tensor([[3 , 1 , 0.5]])\n",
        "\n",
        "l1 = loss(y_pred_good , y)\n",
        "l2 = loss(y_pred_bad , y)\n",
        "print(f\" The good prediction loss : {l1.item():.2f}\\n\")\n",
        "print(f\" The bad prediction loss : {l2.item():.2f}\")\n",
        "\n",
        "_,pred_label1 = torch.max(y_pred_good , 1)\n",
        "_,pred_label2 = torch.max(y_pred_bad , 1)\n",
        "print(\"\\n\" , pred_label1 , pred_label2)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "m3DQo84-uXbb",
        "outputId": "5afdf870-c31f-4de4-c2ab-3f6e1cc57a02"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            " The good prediction loss : 0.07\n",
            "\n",
            " The bad prediction loss : 2.20\n",
            "\n",
            " tensor([1]) tensor([0])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Multiple Samples"
      ],
      "metadata": {
        "id": "0xsSYl3TVoO_"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "loss = nn.CrossEntropyLoss()\n",
        "y = torch.tensor([1 , 2 , 0])\n",
        "y_pred_good = torch.tensor([[2, 4, 1] , [0.1, 0.1, 2], [3.5, 1, 0.5]]) # 3*3\n",
        "y_pred_bad = torch.tensor([[1, 1, 0.5] , [5, 2.5, 0.3] , [0.4, 0.5, 3]]) # 3*3\n",
        "\n",
        "l1 = loss(y_pred_good , y)\n",
        "l2 = loss(y_pred_bad , y)\n",
        "print(f\" The good prediction loss : {l1.item():.2f}\\n\")\n",
        "print(f\" The bad prediction loss : {l2.item():.2f}\")\n",
        "\n",
        "_,pred_label1 = torch.max(y_pred_good , 1)\n",
        "_,pred_label2 = torch.max(y_pred_bad , 1)\n",
        "print(\"\\n\" , pred_label1 , pred_label2)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iAlRRyB-Trzn",
        "outputId": "a9c91b6a-99cb-4b7e-f3d9-e15bcca3e9a8"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            " The good prediction loss : 0.19\n",
            "\n",
            " The bad prediction loss : 2.83\n",
            "\n",
            " tensor([1, 2, 0]) tensor([0, 0, 2])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### CrossEntropyLoss(CEE) vs BinaryCrossEntropyLoss(BCE)\n",
        "  * CrossEntropyLoss can be applied to any classification task with arbitrary number of classes(2 and greater than 2 classes)\n",
        "  * Binary CrossEntropy can only be applied to binary classification task.\n",
        "  * CrossEntropyLoss employes softmax in it's own implementation.\n",
        "  * BinaryCrossEntropy only employes NNLLoss, so we should apply a sigmoid to convert the model output into probability, and then calculate the loss.  "
      ],
      "metadata": {
        "id": "Yfqlj4ZSXWrQ"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Activation Functions**"
      ],
      "metadata": {
        "id": "SBu2vahaZf9m"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F"
      ],
      "metadata": {
        "id": "lo0YeSqWWlo8"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Importing activations using nn\n",
        "  * First, create activation class, then apply it to a input tensor\n",
        "  * nn.ReLU() , nn.Sigmoid() , nn.Softmax() nn.TanH() , nn.LeakyReLU()\n",
        "  * Starts with capital letter.\n",
        "  * Create an instance using self in the __init__ section\n",
        "  * Ex. __init__ : self.relu = nn.ReLU() , forward : out = self.relu(...) "
      ],
      "metadata": {
        "id": "pkXtU7SPiXn-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(torch.tensor([[1 , 2 , 0.5]]).shape)\n",
        "print(torch.tensor([1 , 2 , 0.5]).shape)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Whc8VqB121fC",
        "outputId": "95069808-ed5e-45f4-9e76-cbc4ef618620"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([1, 3])\n",
            "torch.Size([3])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "Act = nn.Softmax(dim=1) # Create the instance\n",
        "\n",
        "print(Act(torch.tensor([[1 , 2 , 0.5]]))) # apply\n",
        "# or\n",
        "Act = nn.Softmax(dim=0)\n",
        "print(Act(torch.tensor([1 , 2 , 0.5])))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9YB_OP0CiS8n",
        "outputId": "a4bef2f9-83b4-4286-ce29-7651104e8b40"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[0.2312, 0.6285, 0.1402],\n",
            "        [0.0900, 0.2447, 0.6652]])\n",
            "tensor([0.2312, 0.6285, 0.1402])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Importing activations using torch\n",
        "  * Used directly in the forward section: Ex. out = torch.sigmoid(...)\n",
        "  * torch.softmax , troch.sigmoid , torch.relu , torch.tanh (Not include **leaky_relu**) => recommend to use nn.functional\n",
        "  * written in small letter."
      ],
      "metadata": {
        "id": "b7XF95uFjjtw"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.tensor([1, 2, 3])\n",
        "print(torch.sigmoid(x))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1HYCzriaim6l",
        "outputId": "084e450f-abc2-4247-9772-c2b91be45906"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([0.7311, 0.8808, 0.9526])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Importing activations using F \n",
        "  * input tensor should be a float32 type."
      ],
      "metadata": {
        "id": "GXXDVyZLner_"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# exclude leaky_relu\n",
        "x = torch.tensor([1, 2, 3])\n",
        "print(torch.leaky_relu(x))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 200
        },
        "id": "VUrTQ3DUkaNZ",
        "outputId": "201f8da6-ecdd-49c2-b247-e23f117ccdd1"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "error",
          "ename": "AttributeError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mAttributeError\u001b[0m                            Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-24-2da69de95eeb>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;31m# exclude leaky_relu\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0mx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m2\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m3\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 3\u001b[0;31m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mleaky_relu\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;31mAttributeError\u001b[0m: module 'torch' has no attribute 'leaky_relu'"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.tensor([-1, -0.5 , 2, 3],dtype=torch.float32)\n",
        "print(F.leaky_relu(x))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LUu_1j7FnsAd",
        "outputId": "059a188a-4a33-46a2-ec44-ff5b66c882ae"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([-0.0100, -0.0050,  2.0000,  3.0000])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Implementing a feed forward neuarl network**"
      ],
      "metadata": {
        "id": "2g1damBh0DW3"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torchvision\n",
        "import torchvision.transforms as transforms\n",
        "from torch.utils.data import DataLoader\n",
        "import matplotlib.pyplot as plt"
      ],
      "metadata": {
        "id": "sc9FW-pFoZKk"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# device config\n",
        "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')"
      ],
      "metadata": {
        "id": "dLq5ox1H0hi0"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# init\n",
        "input_size = 28 * 28\n",
        "hidden_size = 100\n",
        "n_classes = 10\n",
        "num_epochs = 2\n",
        "batch_size = 100\n",
        "learning_rate = 0.001 "
      ],
      "metadata": {
        "id": "JTBYFdmJ3OIY"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Load Dataset(MNIST)\n",
        "train_dataset = torchvision.datasets.MNIST(root = './data' , train = True , transform = transforms.ToTensor(), download = True )\n",
        "test_dataset = torchvision.datasets.MNIST(root = './data' , train = False , transform = transforms.ToTensor())"
      ],
      "metadata": {
        "id": "t96rTR1d4PWl",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 423,
          "referenced_widgets": [
            "d55373061520403c96d61ffdad06993a",
            "bb4bf1afa52546faa33ed0a48d5465a0",
            "3744035230bb46e6b30dfe79386368ba",
            "c13a23b2a235491a8686aa83dad15aeb",
            "2f4c0514d0314351a1c81815569b829c",
            "236501633e0c4e128485a479533c524c",
            "a34f89e6c3b74ef09a72a0d2cc4c0100",
            "bbc2fd671a6541e3bc6542f2497d88a8",
            "0599dcc80a0840a3bf4608970b612f49",
            "ebd5b7a5cdfa4643832ffc14cfff7d3d",
            "d3113b65e3634052a0cb8498cb5e5428",
            "1e3357a13e954cb794d9af14d7430149",
            "2ef9ce1d8cf242eea15494074fa1d9b3",
            "b5b81ae9091c4b97991c66e809b9d5b7",
            "77c5f7a0074c43ba9db4d316c5af7fdf",
            "5ab42f5d11e44cb6a3206392713b1227",
            "be8f5d871fa94c849832c334112bb861",
            "f28b61162f38421095653650f21d1ef3",
            "d772cbe0a09a451bbb11f57eb51282c3",
            "f20b7645111148a5b5f6851a8179786f",
            "3369290057a34f0c8e71d5f20f1b210c",
            "b11ef81d248246169cb333d7eb14e50b",
            "8069d06f822e4c089d93fd3630916436",
            "f973f0bb77a24150b05f8f8691938235",
            "b1f76cf5b4294f0e9f017e4fa29725f1",
            "c903a51b4d734ec2b6eacffeaedefe87",
            "763fdefc59574b8885dbad9bd741a273",
            "85fff734d85b422787d6f8afb059531a",
            "71ef36151be742ab85a52ad965855c3b",
            "22c042aa8bd2407b8f39eba1f4a24647",
            "c7fc9bae0ba243aca087c138f9803831",
            "a3c66bf1da414cd3865fec61e9198150",
            "f17fa29c78f7474eb6cc3a4c28ef2688",
            "6c1402291ee6457395f9c57c7a1b1879",
            "a3f206bbdca5419d8fa0eaaa857d6141",
            "4a76435cdf3f4043994030db53307ef2",
            "2a1aa0c194d34fa0acfc9db5c71ad7e6",
            "df14db4dc27648c6b539218b24fe3da4",
            "31c6b15d5661455da031599655e66470",
            "60915470b1804b5bbf5fe528509f909d",
            "68be7d204c6b446f8f22e235201cb64b",
            "faf53a2d87114f66bdd181ef52bfddc2",
            "242ca09ef3ca4384a8d1e647127bd99f",
            "8e171fb98f8244d7aed3f27657b80992"
          ]
        },
        "outputId": "7a518f96-404f-4f64-b2fe-f5874ba27802"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz\n",
            "Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to ./data/MNIST/raw/train-images-idx3-ubyte.gz\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/9912422 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "d55373061520403c96d61ffdad06993a"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Extracting ./data/MNIST/raw/train-images-idx3-ubyte.gz to ./data/MNIST/raw\n",
            "\n",
            "Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz\n",
            "Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to ./data/MNIST/raw/train-labels-idx1-ubyte.gz\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/28881 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "1e3357a13e954cb794d9af14d7430149"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Extracting ./data/MNIST/raw/train-labels-idx1-ubyte.gz to ./data/MNIST/raw\n",
            "\n",
            "Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz\n",
            "Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to ./data/MNIST/raw/t10k-images-idx3-ubyte.gz\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/1648877 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "8069d06f822e4c089d93fd3630916436"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Extracting ./data/MNIST/raw/t10k-images-idx3-ubyte.gz to ./data/MNIST/raw\n",
            "\n",
            "Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz\n",
            "Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to ./data/MNIST/raw/t10k-labels-idx1-ubyte.gz\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/4542 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "6c1402291ee6457395f9c57c7a1b1879"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Extracting ./data/MNIST/raw/t10k-labels-idx1-ubyte.gz to ./data/MNIST/raw\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Creat a DataLoader instance\n",
        "train_loader = DataLoader(train_dataset , batch_size = batch_size , shuffle = True)\n",
        "test_loader = DataLoader(test_dataset , batch_size = batch_size , shuffle = True)\n",
        "# testing \n",
        "ex = iter(train_loader)\n",
        "samples , labels = next(ex)\n",
        "print(samples.shape , labels.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fpdnkfmJ53n-",
        "outputId": "b9dea5c9-f470-4eae-c9b9-54d35f8a3884"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([100, 1, 28, 28]) torch.Size([100])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "samples[50].view(-1,28**2).shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5oSfipsb639D",
        "outputId": "4906aa86-694b-4b17-c66d-bab7a12da309"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "torch.Size([1, 784])"
            ]
          },
          "metadata": {},
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# plotting some random images\n",
        "for i in range(6):\n",
        "  plt.subplot(2,3,i+1)\n",
        "  plt.imshow(samples[i][0] , cmap='gray')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 267
        },
        "id": "uDWVlUtI60jk",
        "outputId": "2dc2bda4-6f79-480c-e7de-2b22b5d1a07d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 6 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXAAAAD6CAYAAAC4RRw1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAeCklEQVR4nO3dfXRUxfkH8O9jAN8QS5BiBIQqItBT5UWFn1oqIAhoAbVVrLVoacG3FmyrBhVEoS21glRFaKgUtFSxikJBQUTEii8FPFgDFIiegmggBaoiCojM748s48yYu9nsy907d7+fc3LyzM7u3sc8cbiZvTNXlFIgIiL/HJbvBIiIKD0cwImIPMUBnIjIUxzAiYg8xQGciMhTHMCJiDyV0QAuIn1FZIOIVIhIabaSovxiXeOLtY0XSfc6cBEpArARQG8AWwGsBHCFUmpd9tKjsLGu8cXaxk+9DF57FoAKpdS7ACAijwMYCCDwl0FEuGooIpRSEtDFunosSV2BOtaWdY2UHUqppu6DmUyhNAfwntHemnjMIiLDRGSViKzK4FgUHtY1vmqtLesaWZtrejCTM/CUKKXKAJQB/Bc9TljXeGJd/ZLJGfj7AFoa7RaJx8hvrGt8sbYxk8kAvhLAKSLyDRFpAGAwgPnZSYvyiHWNL9Y2ZtKeQlFKHRCRGwEsBlAEYIZSam3WMqO8YF3ji7WNn7QvI0zrYJxTi4xarlaoE9Y1OljX2FqtlDrDfZArMYmIPMUBnIjIUxzAiYg8xQGciMhTHMCJiDzFAZyIyFM5X0pP5IOSkhKrPXPmTKvdu3fvwNeKfHnl3po1a6y+Tp06ZZ4cJVWvnj2MmbUcPXq01efW49lnn9XxlClTrL6qqqpspZgzPAMnIvIUB3AiIk9xACci8hTnwKlgHXvssTqeNm2a1Xf++edb7WRbTph9Bw8ezFJ2lMwJJ5yg4z/+8Y9WX79+/VJ+n86dO+v4uuuus/rM9x0/frzVt2/fvpSPkUs8Ayci8hQHcCIiT3EKhQrWrFmzdHzRRRel/LoDBw5Y7f379+s4Kn9ax93cuXN1fOaZZ1p9yaa7Nm7caLVHjBih4+nTp1t9o0aN0vFHH31k9d17772pJ5tDPAMnIvIUB3AiIk9xACci8lRBzYEfdtiX/161adPG6jv99NN1fPjhh1t9rVu3DnzPAQMGWO2dO3fq+Oabb7b6ysvLU86Vsu+0006z2l26dEn5tZ988omOS0tLrb6pU6dmlhjVmbv1QZBx48ZZ7cmTJ1ttc27b/Rzkueee0/Edd9wR+Dp37jxMPAMnIvIUB3AiIk95MYWyePFiq/3MM8/ouGfPnlbfMcccE/g+RUVFga/LBXPKBgD69++v4y+++CLnxyebu9LOXM1XG3NXO06ZhMOcylyxYoXV16JFCx1//PHHVl/Xrl117F42mIw7xWlOjbhTKGPHjtXx7Nmzrb5PP/005WNmimfgRESe4gBOROQpDuBERJ6SZMtOs34wkbQO5uYY9o5vn3/+eWCfOa8OfHXe22TOz4c5T1YTpZTU/qzUpFvXMPTp00fHf/nLX6y+Jk2aBL7OnfM0l04n+33IN5/r6l6+e9999+l42LBhga+76667rLZ76WA2vP7661bbXL7/yCOPWH3XXHNN1o8PYLVS6gz3QZ6BExF5qtYBXERmiEiViJQbjxWLyBIR2ZT43ji3aVK2sa7xxdoWjlQuI5wJ4EEA5t8JpQCWKqUmiEhpon1r9tOrZq5uBIDGjb/83XMvE0r25+1bb72l4x07dlh95g5zZ599ttV3zjnnBL5njx49rPYLL7wQ+NyImYk81zUX3D/DL7jgAh0nmzJxbd261WpHedqkBjPhYW3d/++STZuY0yS5mDJxzZkzx2qbUygXX3yx1Tdx4kQd53r1da1n4EqplwHsch4eCODQXpyzAAzKcl6UY6xrfLG2hSPdOfBmSqnKRLwNQLMs5UP5xbrGF2sbQxmvxFRKqWSfVovIMADBfwtRJLGu8ZWstqyrX9IdwLeLSIlSqlJESgBUBT1RKVUGoAxI/7Kkbt26We2WLVvq+LXXXrP69u7dm84h0ta2bdtQj5djodY1Fxo2bGi1R44cmadMIiel2oZdV3NHSHOLDNfChQuttnuT4Xxyf+fcdi6lO4UyH8CQRDwEwLzspEN5xrrGF2sbQ6lcRvgYgNcAnCoiW0VkKIAJAHqLyCYA5yfa5BHWNb5Y28JR6xSKUuqKgK5eWc4lUEVFRdJ2PrmXOJrcm99GSRTqmgtNmzZN63VPP/201Z43z98TVJ9q2717dx03atTI6jOnQ93VlmHv5mmuCgXsSwXziSsxiYg8xQGciMhTHMCJiDzlxR15osy9Ua7p0Ucftdr53oGwEDz22GMpP3fPnj06XrRokdXn3uWFcuPCCy/UsbvL6IIFC3T85ptvhpZTTdy7N5k7pIa5o6uLZ+BERJ7iAE5E5ClOoeRQ+/btrXb9+vV17NnudrG0a9eX+z396U9/ymMmhaNXL/tKRnMHQneKMSqX6gHA5Zdfnu8UasQzcCIiT3EAJyLyFAdwIiJPcQ48h9asWWO1Oe9Nhe6WW26x2g0aNNDx+++/b/W5NxKOqsrKyqTtXOIZOBGRpziAExF5igM4EZGnOAeeoWbNgm8tOH369BAzoboy75xy7bXXWn0vvvii1d64cWMoOcXR4YcfruMmTZpYfSKi48mTJ4eWUyo6duyo47Fjx1p9Zt7ulhmbN2/OaV4mnoETEXmKAzgRkac4hVJHJ510ktX+wQ9+EPhcc6c1AKhX78sft7vM/u9//7uO27RpY/Ude+yxOl66dGnqyVJSjRs31vGUKVOsvpUrV1pt85I23ii5bjp16qRjc1oCsHfyc29cnG9mnY8++mirb8eOHTp+6KGHQsvJxTNwIiJPcQAnIvIUB3AiIk9xDjzhiCOO0LF7d+w+ffro2L079lFHHRX4nnfffbfVvuOOO3RsLiEGgKqqKh2bl7cBwIcffqjjsrIyq2/cuHGBx6f0nXnmmVa7c+fOOna3RLj55ptDySnu8nGppvn/709+8hOrb9CgQYGvu+2223S8devW7CeWIp6BExF5igM4EZGnOIWSYF4KNGTIkLTfx7wZrjllAgD//e9/A19XUVGh45KSEqsvapdXxcXu3bt1/Oqrr1p9F1xwgdUuKirScbdu3ay+tm3b6pgrNv0yZ84cHffr1y/wee6q6qjcwYln4EREnuIATkTkqVoHcBFpKSLLRGSdiKwVkRGJx4tFZImIbEp8b1zbe1F0sK7xxLoWFjGXstb4BJESACVKqTdF5BgAqwEMAnA1gF1KqQkiUgqgsVLq1lreK/nB8sic/7r00ksDn3fyySdb7e7du1vtFStWBPZFzAmIYV3decwFCxYEPte8C/q0adOsPnfJd8+ePQPfx7xj+ZNPPplSnjkU6bq6nzV07dpVx+7PePny5Rkf79xzz7Xa7mW33/nOd3S8b98+q+/+++/X8fjx460+8/OTkKxWSp3hPljrGbhSqlIp9WYi3g1gPYDmAAYCmJV42ixU/5KQJ1jXeGJdC0udrkIRkdYAOgF4A0AzpdShm79tA1DjxtgiMgzAsPRTpFxjXeOJdY2/lAdwEWkI4CkAI5VSH5sbmiulVNCfW0qpMgBlifeIzJ/arueee67G2OVu/B/xaZJaxa2u//znP1N+rrkKr127dlbfunXrrLb5p7i7ijaKolpXc9dNADjrrLN07E5vXH/99TouLy+3+k444QQdn3jiiVafuQvo8OHDrb7i4mKr/fzzz+v4d7/7ndW3bNmyr/4HRExKV6GISH1U/zLMVkrNTTy8PTE/fmievCro9RRNrGs8sa6FI5WrUATAwwDWK6UmGV3zARxa8TIEwLzsp0e5wrrGE+taWFKZQjkHwFUA3haRNYnHbgMwAcATIjIUwGYAl+UmRcoR1jWeWNcCUusArpR6BYAEdPfKbjrRd/XVV+c7hayIa10/++wzqz137lwdX3LJJYGv69+/v9V+6qmnrPbevXt1HOU58KjX1d1Ns2/fvjp2L/l75ZVXdLx27Vqrz9xuolWrVlafeWm0e3PqMWPGWG3zTks+4kpMIiJPcQAnIvIUdyOkWDFXVwLAE088oeNkUyiuZKtx58+fb7VffvnllN+30O3cudNqX3TRRTpu3ry51Td06FAdDx482Oozp1DcFZvm7p0zZ85Menzf8QyciMhTHMCJiDzFAZyIyFO17kaY1YNFaMl1utzLjtyb31555ZU6fvzxx0PJKR1KqaBLzeosDnWNC9Y1ttLbjZCIiKKJAzgRkad4GWEdbdmyxWq7UygHDx4MMx0iKmA8Ayci8hQHcCIiT3EAJyLyFOfA6+iee+6x2k2bNrXaFRUVYaZDRAWMZ+BERJ7iAE5E5CmuxCxQXLEXT6xrbHElJhFRnHAAJyLyFAdwIiJPhX0Z4Q5U3xH7uEQcBYWYS6van1InrGtyrGv2FGouNdY21A8x9UFFVtU0IZ8PzCV7opQ/c8meKOXPXGycQiEi8hQHcCIiT+VrAC/L03FrwlyyJ0r5M5fsiVL+zMWQlzlwIiLKHKdQiIg8xQGciMhToQ7gItJXRDaISIWIlIZ57MTxZ4hIlYiUG48Vi8gSEdmU+N44hDxaisgyEVknImtFZES+cskG1tXKJTa1ZV2tXCJZ19AGcBEpAjAFQD8AHQBcISIdwjp+wkwAfZ3HSgEsVUqdAmBpop1rBwD8UinVAUA3ADckfhb5yCUjrOtXxKK2rOtXRLOuSqlQvgD8H4DFRnsUgFFhHd84bmsA5UZ7A4CSRFwCYEMecpoHoHcUcmFdWVvW1Z+6hjmF0hzAe0Z7a+KxfGumlKpMxNsANAvz4CLSGkAnAG/kO5c0sa4BPK8t6xogSnXlh5gGVf3PaGjXVYpIQwBPARiplPo4n7nEWT5+lqxt7rGu4Q7g7wNoabRbJB7Lt+0iUgIAie9VYRxUROqj+hdhtlJqbj5zyRDr6ohJbVlXRxTrGuYAvhLAKSLyDRFpAGAwgPkhHj/IfABDEvEQVM9t5ZSICICHAaxXSk3KZy5ZwLoaYlRb1tUQ2bqGPPHfH8BGAO8AuD0PHzw8BqASwOeontMbCqAJqj893gTgBQDFIeRxLqr/1PoXgDWJr/75yIV1ZW1ZV3/ryqX0RESe4oeYRESe4gBOROSpjAbwfC+1pdxgXeOLtY2ZDCb1i1D94cZJABoAeAtAh1peo/gVjS/WNZ5f2fx/Nt//Lfyyvv5bU40yOQM/C0CFUupdpdR+AI8DGJjB+1E0sK7xxdr6a3NND2YygKe01FZEhonIKhFZlcGxKDysa3zVWlvW1S/1cn0ApVQZErceEhGV6+NROFjXeGJd/ZLJGXhUl9pSZljX+GJtYyaTATyqS20pM6xrfLG2MZP2FIpS6oCI3AhgMao/3Z6hlFqbtcwoL1jX+GJt4yfUpfScU4sOpZRk671Y1+hgXWNrtVLqDPdBrsQkIvIUB3AiIk9xACci8lTOrwMnIgpTSUmJji+77DKrb/LkyVbb/Azw29/+ttW3YsWKHGSXXTwDJyLyFAdwIiJPcQolQy1btrTat99+u46HDx+e1nvu3LnTah933HFpvQ9RXNWvX1/H7v+DEydO1PGAAQOsvoMHD+Y2sZDxDJyIyFMcwImIPMUBnIjIU5wDr6MePXpY7cWLF1vtevW+/JGmu01BcXGx1b7xxht1/OCDD6b1npSZ5s2/3Da7vLzc6vv1r3+t43vvvTe0nArZ6NGjdWx+7pSJLl26WG1eRkhERDnDAZyIyFOcQknBmDFjdHzrrbdafeaUiWvHjh1We/bs2Tp+8sknrb42bdro+M9//rPVd9hh/Hc23370ox/puFGjRlZfv379dMwplHC4l9oG2b17t9V+9NFHrfZ3v/tdHf/+97+3+j766CMdz5o1q64phoIjAxGRpziAExF5igM4EZGnOAeeYM5Bz5s3z+pr27atjouKipK+z8KFC3U8fvx4q++NN97QsTuPOnbsWB2bS4EBoKysLOkxKfd69+4d2Ld8+fIQMyEAeOCBB3T8n//8J/B5q1atstp79uyx2uYcuPt51pFHHplBhuHgGTgRkac4gBMReapgp1Dcjd7N6Q5zOsVVWVlptb///e9b7dWrV+t43759ge/z6aefWu0f/vCHOt6+fXvg66huOnbsqGP38sxevXrpeNeuXVafO8XVtWvXwGO88847maRIaTB3FXSnPJM59dRTU36uOXUaVTwDJyLyFAdwIiJPcQAnIvJUQc2Bm3Pb7q5+ye56c//99+t4woQJVt+2bdvSyuXAgQNWm/Pe2WHeqQUAxo0bp+PTTjvN6rv88st1PHXqVKvP3ZnuiCOO0LG5xBoAlixZkl6ylHPHH3+81f7HP/5htZs0aaLjt99+2+p7+OGHc5dYlvAMnIjIU7UO4CIyQ0SqRKTceKxYRJaIyKbE98a5TZOyjXWNL9a2cKQyhTITwIMAHjEeKwWwVCk1QURKE+1ba3htXg0cONBqm7uNuVMmixYt0vHKlSutvt/+9rc63rt3bzZTzKeZ8LSuybRv395qX3jhhTr+4osvrL4333wz8H3cS0lFRMebNm2y+qqqquqcZ47NRAxrm6p27drp+IknnrD6zCkTl7kaGgDWrl2b1bxyodYzcKXUywB2OQ8PBHBof8VZAAZlOS/KMdY1vljbwpHuh5jNlFKHVrRsA9As6IkiMgzAsDSPQ+FiXeMrpdqyrn7J+CoUpZQSkcCbPyqlygCUAUCy51G0sK7xlay2rKtf0h3At4tIiVKqUkRKAERmEvDkk0/WsXv3jYYNGwa+buTIkTreuHFj9hPzQ2TrmsxRRx2l4/vuu8/qM28s/frrr1t95u6QrksvvTTwfZ599tm08swzL2sbpEOHDjqeNm2a1depUycdm78btZk0aZLVNi87vfvuu+uaYijSvYxwPoAhiXgIgNQ3I6AoY13ji7WNoVQuI3wMwGsAThWRrSIyFMAEAL1FZBOA8xNt8gjrGl+sbeGodQpFKXVFQFevgMfz6he/+IWOk02ZmKsrAWDLli05yymKfKtrMnPnztVxjx49rD5zR8ibbrop8D3MS88AoE+fPlbbnEL597//nVaeYYlTbQ/p3Lmz1Tanv5Ld9NtdeenelNzcPdRdiXnllVfq2F2BvX///loyDgdXYhIReYoDOBGRpziAExF5Kna7EV511VWBfebSaXfZbLaWyJt3gGncOPXtJsw5tRUrVmQll7gaNWqU1TbvrOPu8jhmzBgduze4Nblz4K41a9bo+Omnn04pT8oe93OH3/zmNzouLi62+swl8HPmzLH6/ve//wUew9wyA7Dv4DR69Girz23nC8/AiYg8xQGciMhTsZtCSaZ58+Y6PvLII62+Dz/8MPB15rTIt771Lavvmmuusdrm6i33T7tkzCkUd4rA/DPwgw8+SPk94+TEE0/UcWlpqdVnXkb217/+1eozd6BMZujQoUn7ly9fruOoXEJWSNybgN95551ZP8bOnTsD+6J6g2OegRMReYoDOBGRpziAExF5qqDmwBcuXKjjyspKq++KK75cfTxkyBCrr1u3bjpu1KhR0mOYOxm+++67KefWunVrHU+cONHqM+8k424BUCh+/OMf6/iYY46x+sxa/uxnP7P6vva1rwW+57nnnqtj8849gH0HHsC+zNS93Ox73/uejs3dMN33MZfjA8Bdd91VY0z5kWyee8OGDSFmkjqegRMReYoDOBGRpziAExF5qqDmwM3rst0l9w899JCOjz76aKuvoqJCx8uWLbP6zOW2gH3Xl7rcrdzcBnXp0qVW3/HHH5/y+xQCdy7Z/Pns2rUr6XNTfU+Xuw1pOu+Tai6UHy1atAjsO/XUU0PMJHU8Ayci8hQHcCIiTxXUFMqgQYNqjAFg9+7dOnYvRTOXZyfbzSwT7p/+phtuuEHHt912W06OH3UvvfSSjutyqWC63Hq8+OKLOjZ3u6uLV155xWqn+z5R595I2F0GH1Xf/OY3851CnfEMnIjIUxzAiYg8xQGciMhT3s+Bu/Of7hLoIJ988onVNpfPP/PMM5knVoMGDRroeMSIEVbf8OHDdXzw4EGrb/PmzTnJxyfmHPjpp59u9RUVFem4fv36Vp+5RL5Lly5Wn3nXcdcf/vAHqz116lQd79ixo/aEY+6BBx6w2ubv6IIFC6w+9246UWJuk3HeeecFPu/5558PIZu64xk4EZGnOIATEXnK+ykUcxdB4KuXMAXZvn271TZXUGZLkyZNrPbPf/5zHbs3RTV3HFy3bp3VZ97lh4CtW7em/FxzKuT2228PfJ55w2sAGDduXN0TKyDurovXX3+9jlevXm31vffeezres2dPbhOrhbub6PTp03VsTnEC9srZqF7yyTNwIiJP1TqAi0hLEVkmIutEZK2IjEg8XiwiS0RkU+J749ynS9nCusYT61pYUjkDPwDgl0qpDgC6AbhBRDoAKAWwVCl1CoCliTb5g3WNJ9a1gNQ6B66UqgRQmYh3i8h6AM0BDARwXuJpswC8BCC1LduyyLy8C7DvsuLOj5vcObxJkybp2J0PN5fSu8uC27VrZ7V/+tOf6rhv375WX6tWrXTsXir44IMP6vimm24KzDtbol7XbDHvWN+9e3erz7zktC53T4qysOp68cUXW+3169fr+IUXXrD6ysvLdexefpjsLlnZ4O4wOGPGDKvdoUMHHbu7RZo7lK5cuTLruWVDnT7EFJHWADoBeANAs8QvCwBsA9As4DXDAAxLP0XKNdY1nljX+Ev5Q0wRaQjgKQAjlVIfm32q+p+uGjc7VkqVKaXOUEqdkVGmlBOsazyxroVBUtlkXkTqA1gAYLFSalLisQ0AzlNKVYpICYCXlFJJdz0XkZzvaN+xY0cdL1q0yOpr2rSpmUta7+9Or5gruepiypQpVtvdYS/XlFLiU13TZd4s+p133gl83uDBg6323/72t1yllFP5qqt5CZ55A2oAuOWWW3RsTiMCwP79+3X86quvWn3m1Itr/vz5Oh4wYEDg83r27Gm127dvb7XNceDOO++0+u65554a88yT1TX9o5rKVSgC4GEA6w/9MiTMB3Bo/fkQAPOykSWFg3WNJ9a1sKQyB34OgKsAvC0iaxKP3QZgAoAnRGQogM0ALstNipQjrGs8sa4FJJWrUF4BEDTf0Cu76VBYWNd4Yl0LS0pz4Fk7WJ7nSq+77jodjxkzxuozl73Xq5edHQYOHDhgtc3LEc35NeCry+dzTSmV3ocANch3XZNJNge+atUqHZ999tlWn7m1gU+iWFdzx1D30t5LLrlEx+58dS6sWLHCav/qV7/SsbtFwwcffJDzfOogvTlwIiKKJg7gRESeKqgplGTMFVnuKrNrr71Wx82bN7f6tmzZYrXN3c3cS9E2btyYcZ7ZEsU/tXMh2RSK+Sezu6L2s88+y2leuVIodS1AnEIhIooTDuBERJ7iAE5E5CnOgReoQpkr/frXv67jZcuWWX3mzYndG9qG+f9FNhVKXQsQ58CJiOKEAzgRkac4hVKg+Kd2PLGuscUpFCKiOOEATkTkKQ7gRESe4gBOROQpDuBERJ7iAE5E5CkO4EREnuIATkTkKQ7gRESe4gBOROSp7Ny9N3U7AGwGcFwijoJCzKVVlt+PdU2Odc2eQs2lxtqGuheKPqjIqprW9ecDc8meKOXPXLInSvkzFxunUIiIPMUBnIjIU/kawMvydNyaMJfsiVL+zCV7opQ/czHkZQ6ciIgyxykUIiJPcQAnIvJUqAO4iPQVkQ0iUiEipWEeO3H8GSJSJSLlxmPFIrJERDYlvjcOIY+WIrJMRNaJyFoRGZGvXLKBdbVyiU1tWVcrl0jWNbQBXESKAEwB0A9ABwBXiEiHsI6fMBNAX+exUgBLlVKnAFiaaOfaAQC/VEp1ANANwA2Jn0U+cskI6/oVsagt6/oV0ayrUiqULwD/B2Cx0R4FYFRYxzeO2xpAudHeAKAkEZcA2JCHnOYB6B2FXFhX1pZ19aeuYU6hNAfwntHemngs35oppSoT8TYAzcI8uIi0BtAJwBv5ziVNrGsAz2vLugaIUl35IaZBVf8zGtp1lSLSEMBTAEYqpT7OZy5xlo+fJWube6xruAP4+wBaGu0WicfybbuIlABA4ntVGAcVkfqo/kWYrZSam89cMsS6OmJSW9bVEcW6hjmArwRwioh8Q0QaABgMYH6Ixw8yH8CQRDwE1XNbOSUiAuBhAOuVUpPymUsWsK6GGNWWdTVEtq4hT/z3B7ARwDsAbs/DBw+PAagE8Dmq5/SGAmiC6k+PNwF4AUBxCHmci+o/tf4FYE3iq38+cmFdWVvW1d+6cik9EZGn+CEmEZGnOIATEXmKAzgRkac4gBMReYoDOBGRpziAExF5igM4EZGn/h/tkL67OSL/EQAAAABJRU5ErkJggg==\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "torch.tensor([1,2,3]).sum()#.item()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "akisziO3PAvi",
        "outputId": "0d018d3f-abcd-42f8-daf2-8cf2811b43a3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor(6)"
            ]
          },
          "metadata": {},
          "execution_count": 65
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# define class\n",
        "class FeedForward(nn.Module):\n",
        "  def __init__(self,input_size ,hidden_size , n_classes):\n",
        "    super(FeedForward , self).__init__()\n",
        "    self.l1 = nn.Linear(input_size ,hidden_size)\n",
        "    self.relu = nn.ReLU()\n",
        "    self.l2 = nn.Linear(hidden_size , n_classes)\n",
        "  def forward(self, x):\n",
        "    out = self.l1(x)\n",
        "    out = self.relu(out)\n",
        "    out = self.l2(out)\n",
        "    return out\n",
        "\n",
        "model = FeedForward(input_size , hidden_size , n_classes).to(device)\n",
        "\n",
        "# loss , optimizer\n",
        "criterion = nn.CrossEntropyLoss() #loss\n",
        "optimizer = torch.optim.Adam( model.parameters() ,lr = learning_rate)\n",
        "# training\n",
        "n_total_steps = len(train_loader)\n",
        "for epoch in range(num_epochs):\n",
        "  for i , (inputs , targets) in enumerate(train_loader):\n",
        "    # reshape inputs(100,1,28,28) => (100 , 28*28)\n",
        "    inputs = inputs.view(-1,28*28).to(device)\n",
        "    targets = targets.to(device)\n",
        "\n",
        "    # forward\n",
        "    pred = model(inputs)\n",
        "    l1 = criterion(pred,targets)\n",
        "    # backprog\n",
        "    optimizer.zero_grad()\n",
        "    l1.backward()\n",
        "    optimizer.step()\n",
        "\n",
        "    # printing\n",
        "    if (i+1)%100 == 0:\n",
        "      print(f\"epcoh {epoch+1}/{num_epochs} , step : {i+1}/{n_total_steps} , loss : {l1.item():.2f}\")\n",
        "\n",
        "\n",
        "with torch.no_grad():\n",
        "  n_correct = 0\n",
        "  n_samples = 0\n",
        "  for i , (inputs , targets) in enumerate(test_loader):\n",
        "    inputs = inputs.view(-1,28*28).to(device)\n",
        "    targets = targets.to(device)\n",
        "    # pred\n",
        "    out = model(inputs)\n",
        "    _,pred = torch.max(out , dim = 1)\n",
        "    n_correct += (pred == targets).sum().item()\n",
        "    n_samples += targets.shape[0]\n",
        "  \n",
        "  acc = 100 * n_correct / n_samples\n",
        "  print(f'acc : {acc:.2f}')\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-H9QbvMb9xEc",
        "outputId": "97ad39a9-e341-4273-ec05-3aba8d487d53"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "epcoh 1/2 , step : 100/600 , loss : 0.48\n",
            "epcoh 1/2 , step : 200/600 , loss : 0.45\n",
            "epcoh 1/2 , step : 300/600 , loss : 0.20\n",
            "epcoh 1/2 , step : 400/600 , loss : 0.20\n",
            "epcoh 1/2 , step : 500/600 , loss : 0.16\n",
            "epcoh 1/2 , step : 600/600 , loss : 0.24\n",
            "epcoh 2/2 , step : 100/600 , loss : 0.17\n",
            "epcoh 2/2 , step : 200/600 , loss : 0.12\n",
            "epcoh 2/2 , step : 300/600 , loss : 0.25\n",
            "epcoh 2/2 , step : 400/600 , loss : 0.23\n",
            "epcoh 2/2 , step : 500/600 , loss : 0.11\n",
            "epcoh 2/2 , step : 600/600 , loss : 0.11\n",
            "acc : 95.38\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Implementing a Convolutional neuarl network**"
      ],
      "metadata": {
        "id": "cIG5YHUrnnLl"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#Exercise "
      ],
      "metadata": {
        "id": "Lr6UHokRnGVz"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Transfer Learning**"
      ],
      "metadata": {
        "id": "Jkk8kEzA4SOu"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "from torch.optim import lr_scheduler\n",
        "import numpy as np\n",
        "import torchvision\n",
        "from torchvision import datasets , models , transforms\n",
        "from torch.utils.data import DataLoader\n",
        "import matplotlib.pyplot as plt\n",
        "import time \n",
        "import os\n",
        "import copy"
      ],
      "metadata": {
        "id": "43CVs7ZN4VaZ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "#device config\n",
        "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")"
      ],
      "metadata": {
        "id": "eHrG1ID_Uvxl"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# work with transforms\n",
        "mean = np.array([0.485 ,0.456 ,0.406])\n",
        "std = np.array([0.229 ,0.224 ,0.225])\n",
        "\n",
        "data_transforms = {\n",
        "    \"train\" :\n",
        "        transforms.Compose([\n",
        "        transforms.RandomResizedCrop(220),\n",
        "        transforms.RandomHorizontalFlip(),\n",
        "        transforms.ToTensor(),\n",
        "        transforms.Normalize(mean , std)                            \n",
        "    ]),\n",
        "    \"val\" : \n",
        "      transforms.Compose([\n",
        "      transforms.Resize(256),\n",
        "      transforms.CenterCrop(250),\n",
        "      transforms.ToTensor(),\n",
        "      transforms.Normalize(mean , std)                    \n",
        "      ]),\n",
        "}\n"
      ],
      "metadata": {
        "id": "rEwGPOTtWADG"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Dealing with dara\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UthmbcMPX_Iz",
        "outputId": "f9594470-61d3-415c-f0b9-7e9e857d8bde"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Datsets , DataLoader\n",
        "data_dir ='/content/drive/MyDrive/Notebooks/Pytorch_Tutorial/hymenoptera_data.zip (Unzipped Files)/hymenoptera_data/'\n",
        "sets = ['train' , 'val']\n",
        "image_dataset = {x : datasets.ImageFolder(os.path.join(data_dir , x) , data_transforms[x]) for x in sets}\n",
        "data_loaders = {x : DataLoader(image_dataset[x] , batch_size = 4 , shuffle = True , num_workers = 0) for x in sets}\n",
        "dataset_sizes = {x : len(image_dataset[x]) for x in sets} \n",
        "print(dataset_sizes)\n",
        "class_names = image_dataset['train'].classes\n",
        "print(class_names)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UT-AflO5rOU4",
        "outputId": "f4807499-2cc5-4ca1-ee31-8ea20eb260b6"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "{'train': 244, 'val': 153}\n",
            "['ants', 'bees']\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "A = np.random.rand(2,2,3)\n",
        "A"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5GhgbV_ORguQ",
        "outputId": "0b5f8327-1db8-4596-faeb-fb48b7883f8d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[[0.40414624, 0.97916179, 0.97843442],\n",
              "        [0.58613184, 0.1653862 , 0.9134577 ]],\n",
              "\n",
              "       [[0.85400859, 0.40918424, 0.27281436],\n",
              "        [0.5658334 , 0.75260135, 0.44823035]]])"
            ]
          },
          "metadata": {},
          "execution_count": 26
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "A.transpose(1,2,0) # (2,3,2)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pS6Q5ZbDRzqv",
        "outputId": "42750ad5-bfea-4409-aee0-a879a0b5b02b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[[0.40414624, 0.85400859],\n",
              "        [0.97916179, 0.40918424],\n",
              "        [0.97843442, 0.27281436]],\n",
              "\n",
              "       [[0.58613184, 0.5658334 ],\n",
              "        [0.1653862 , 0.75260135],\n",
              "        [0.9134577 , 0.44823035]]])"
            ]
          },
          "metadata": {},
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def imshow(inp, title):\n",
        "    \"\"\"Imshow for Tensor.\"\"\"\n",
        "    print(inp.shape)\n",
        "    inp = inp.numpy().transpose((1, 2, 0)) # in order to plot images, the channel dimention should be the last dim.\n",
        "    print(inp.shape)\n",
        "    inp = std * inp + mean\n",
        "    inp = np.clip(inp, 0, 1)\n",
        "    plt.imshow(inp)\n",
        "    plt.title(title)\n",
        "    plt.show()\n",
        "\n",
        "# Get a batch of training data\n",
        "inputs, classes = next(iter(data_loaders['train']))\n",
        "\n",
        "# Make a grid from batch\n",
        "out = torchvision.utils.make_grid(inputs) # concatenate images in a row (reduce from 4 dim into 3 dim)  (4,1,28,28) => (1,28,28*4)\n",
        "\n",
        "imshow(out , title=[class_names[x] for x in classes])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 183
        },
        "id": "BkHFRwqz36y-",
        "outputId": "e76dff82-a7f6-48cb-9fd3-4b8ebff4fae7"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([3, 224, 890])\n",
            "(224, 890, 3)\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Create the model\n",
        "def train_model(model, criterion, optimizer, scheduler, num_epochs=25):\n",
        "    since = time.time()\n",
        "\n",
        "    best_model_wts = copy.deepcopy(model.state_dict())\n",
        "    best_acc = 0.0\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        print('Epoch {}/{}'.format(epoch, num_epochs - 1))\n",
        "        print('-' * 10)\n",
        "\n",
        "        # Each epoch has a training and validation phase\n",
        "        for phase in ['train', 'val']:\n",
        "            if phase == 'train':\n",
        "                model.train()  # Set model to training mode\n",
        "            else:\n",
        "                model.eval()   # Set model to evaluate mode\n",
        "\n",
        "            running_loss = 0.0\n",
        "            running_corrects = 0\n",
        "\n",
        "            # Iterate over data.\n",
        "            for inputs, labels in data_loaders[phase]:\n",
        "                inputs = inputs.to(device)\n",
        "                labels = labels.to(device)\n",
        "\n",
        "                # forward\n",
        "                # track history if only in train\n",
        "                with torch.set_grad_enabled(phase == 'train'):\n",
        "                    outputs = model(inputs)\n",
        "                    _, preds = torch.max(outputs, 1)\n",
        "                    loss = criterion(outputs, labels)\n",
        "\n",
        "                    # backward + optimize only if in training phase\n",
        "                    if phase == 'train':\n",
        "                        optimizer.zero_grad()\n",
        "                        loss.backward()\n",
        "                        optimizer.step()\n",
        "\n",
        "                # statistics\n",
        "                running_loss += loss.item() * inputs.size(0)\n",
        "                running_corrects += torch.sum(preds == labels.data)\n",
        "\n",
        "            if phase == 'train':\n",
        "                scheduler.step()\n",
        "\n",
        "            epoch_loss = running_loss / dataset_sizes[phase]\n",
        "            epoch_acc = running_corrects.type(torch.float64) / dataset_sizes[phase]\n",
        "\n",
        "            print('{} Loss: {:.4f} Acc: {:.4f}'.format(\n",
        "                phase, epoch_loss, epoch_acc))\n",
        "\n",
        "            # deep copy the model\n",
        "            if phase == 'val' and epoch_acc > best_acc:\n",
        "                best_acc = epoch_acc\n",
        "                best_model_wts = copy.deepcopy(model.state_dict())\n",
        "\n",
        "        print()\n",
        "\n",
        "    time_elapsed = time.time() - since\n",
        "    print('Training complete in {:.0f}m {:.0f}s'.format(\n",
        "        time_elapsed // 60, time_elapsed % 60))\n",
        "    print('Best val Acc: {:4f}'.format(best_acc))\n",
        "\n",
        "    # load best model weights\n",
        "    model.load_state_dict(best_model_wts)\n",
        "    return model"
      ],
      "metadata": {
        "id": "dsBZFYO4wmE3"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Finetuning vs Freezing"
      ],
      "metadata": {
        "id": "M1pYwzgEjJcA"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#### Finetuning the convnet ####\n",
        "model = models.resnet18(pretrained = True)\n",
        "num_features = model.fc.in_features\n",
        "# assign a new output size\n",
        "model.fc = nn.Linear(num_features , 2)\n",
        "model.to(device)\n",
        "# loss , optimizer\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "optimizer = optim.SGD(model.parameters() , lr = 0.05)\n",
        "\n",
        "#scheduler\n",
        "step_lr_schedu = lr_scheduler.StepLR(optimizer , step_size = 10 , gamma = 0.1)\n",
        "\n",
        "model = train_model(model , criterion , optimizer , step_lr_schedu , num_epochs = 2)\n",
        "### Freezing all the layers except the last layer\n",
        "model = models.resnet18(pretrained = True)\n",
        "for param in model.parameters():\n",
        "  param.requires_grad= False\n",
        "num_features = model.fc.in_features\n",
        "# assign a new output size\n",
        "model.fc = nn.Linear(num_features , 2)\n",
        "model.to(device)\n",
        "# loss , optimizer\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "opimizer = optim.SGD(model.parameters() , lr = 0.05)\n",
        "\n",
        "#scheduler\n",
        "step_lr_schedu = lr_scheduler.StepLR(optimizer , step_size = 10 , gamma = 0.1)\n",
        "\n",
        "model = train_model(model , criterion , optimizer , step_lr_schedu , num_epochs = 2)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "v-7Wa6LH6jHr",
        "outputId": "9431bf7b-843a-49d0-ac03-8874b69505f9"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 0/1\n",
            "----------\n",
            "train Loss: 3.7550 Acc: 0.5492\n",
            "val Loss: 9.7473 Acc: 0.4771\n",
            "\n",
            "Epoch 1/1\n",
            "----------\n",
            "train Loss: 3.2041 Acc: 0.4959\n",
            "val Loss: 2.0744 Acc: 0.4314\n",
            "\n",
            "Training complete in 0m 10s\n",
            "Best val Acc: 0.477124\n",
            "Epoch 0/1\n",
            "----------\n",
            "train Loss: 0.8098 Acc: 0.5000\n",
            "val Loss: 0.8978 Acc: 0.4314\n",
            "\n",
            "Epoch 1/1\n",
            "----------\n",
            "train Loss: 0.8001 Acc: 0.5164\n",
            "val Loss: 0.9247 Acc: 0.4510\n",
            "\n",
            "Training complete in 0m 9s\n",
            "Best val Acc: 0.450980\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Saving Model**"
      ],
      "metadata": {
        "id": "hjLhM6YjjU2n"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Complete Model\n",
        "  * torch.save(model , PATH)\n",
        "  * torch.load(PATH)\n",
        "  * model.eval()\n",
        "  \n",
        "## STATE DICT\n",
        "  * torch.save(model.state_dict()))\n",
        "  * model = Model(*args , **kwargs)\n",
        "  * model.load_state_dict(torch.load(PATH))\n",
        "  * model.eval()"
      ],
      "metadata": {
        "id": "u7J0BqW2jvm3"
      }
    }
  ]
}