CISC_873_DM_F22_A4.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "authorship_tag": "ABX9TyP0CLMLsxp5lIDE0KJIq4gJ",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU",
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "efb7bcbb8bf2490391f4a238b89d4b22": {
          "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_f712e239096c4bf79c16df9ba6fab628",
              "IPY_MODEL_7f03538152024fbca69ef08d507381d1",
              "IPY_MODEL_ef6087d76e1b41089f43efd1128e3484"
            ],
            "layout": "IPY_MODEL_ff3653f7e11d4268b3d6096a20cf71eb"
          }
        },
        "f712e239096c4bf79c16df9ba6fab628": {
          "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_fcc0bd48d43f4e4db0cad088f3513b10",
            "placeholder": "​",
            "style": "IPY_MODEL_1ac3936fee0a490cae81918ae7e277b7",
            "value": "100%"
          }
        },
        "7f03538152024fbca69ef08d507381d1": {
          "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_2ff3209b9906406a8a6c775b414b2b9e",
            "max": 5860,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_46413c7c59604f1c8bf0a099873a8003",
            "value": 5860
          }
        },
        "ef6087d76e1b41089f43efd1128e3484": {
          "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_4304108b4e584cc8967d93033e11a830",
            "placeholder": "​",
            "style": "IPY_MODEL_06d78c89d6474ba5afe611816960791e",
            "value": " 5860/5860 [01:01&lt;00:00, 117.35it/s]"
          }
        },
        "ff3653f7e11d4268b3d6096a20cf71eb": {
          "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
          }
        },
        "fcc0bd48d43f4e4db0cad088f3513b10": {
          "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
          }
        },
        "1ac3936fee0a490cae81918ae7e277b7": {
          "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": ""
          }
        },
        "2ff3209b9906406a8a6c775b414b2b9e": {
          "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
          }
        },
        "46413c7c59604f1c8bf0a099873a8003": {
          "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": ""
          }
        },
        "4304108b4e584cc8967d93033e11a830": {
          "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
          }
        },
        "06d78c89d6474ba5afe611816960791e": {
          "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": ""
          }
        },
        "9d00817cf3554d728253e397f7fdd62e": {
          "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_ddee07994bf642eab2d4a0467775b374",
              "IPY_MODEL_94cc02ea8d604cc9bccd1fa282650353",
              "IPY_MODEL_f5ec5d17b93e49599c92b8b22e038499"
            ],
            "layout": "IPY_MODEL_133a410ae8984f179de85f9c5dd2b6e9"
          }
        },
        "ddee07994bf642eab2d4a0467775b374": {
          "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_acc5d2e63056423fb528242b40b05d64",
            "placeholder": "​",
            "style": "IPY_MODEL_447c9ce6ccb3481abb5da85ce471739d",
            "value": "100%"
          }
        },
        "94cc02ea8d604cc9bccd1fa282650353": {
          "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_a8e4ba30b30a4a5c839e59460d015588",
            "max": 1466,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_2c78f8833c4543b18328df3b649299e6",
            "value": 1466
          }
        },
        "f5ec5d17b93e49599c92b8b22e038499": {
          "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_ac6166ea9509499395f0a548c74b1787",
            "placeholder": "​",
            "style": "IPY_MODEL_5d1b7b1e71d44480b051a660219e6fae",
            "value": " 1466/1466 [00:14&lt;00:00, 111.48it/s]"
          }
        },
        "133a410ae8984f179de85f9c5dd2b6e9": {
          "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
          }
        },
        "acc5d2e63056423fb528242b40b05d64": {
          "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
          }
        },
        "447c9ce6ccb3481abb5da85ce471739d": {
          "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": ""
          }
        },
        "a8e4ba30b30a4a5c839e59460d015588": {
          "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
          }
        },
        "2c78f8833c4543b18328df3b649299e6": {
          "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": ""
          }
        },
        "ac6166ea9509499395f0a548c74b1787": {
          "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
          }
        },
        "5d1b7b1e71d44480b051a660219e6fae": {
          "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": ""
          }
        },
        "786f637c189d457aaaa1f54b88ca1cff": {
          "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_35244a1845b6499e989864e0ebc6238d",
              "IPY_MODEL_6b6f92e7daed47c5b6dae08f54b54ea9",
              "IPY_MODEL_021d31b7f7dc4f798c317109a79c142b"
            ],
            "layout": "IPY_MODEL_1faad506bbf84271ae7669c652edfdc0"
          }
        },
        "35244a1845b6499e989864e0ebc6238d": {
          "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_c9e35f22cd9e4561a14daaed3abca124",
            "placeholder": "​",
            "style": "IPY_MODEL_954e6968261b4d1a9f9bbea55a0c242d",
            "value": "100%"
          }
        },
        "6b6f92e7daed47c5b6dae08f54b54ea9": {
          "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_86479a7791b34265a2f7dcc68d20d8bd",
            "max": 5860,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_642a5ac2f4b2427a89ef498643392651",
            "value": 5860
          }
        },
        "021d31b7f7dc4f798c317109a79c142b": {
          "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_f2e003882f804290acac27ef5439f4cc",
            "placeholder": "​",
            "style": "IPY_MODEL_7abd1b443f3f4fdd9a2b589244125b9c",
            "value": " 5860/5860 [01:03&lt;00:00, 102.96it/s]"
          }
        },
        "1faad506bbf84271ae7669c652edfdc0": {
          "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
          }
        },
        "c9e35f22cd9e4561a14daaed3abca124": {
          "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
          }
        },
        "954e6968261b4d1a9f9bbea55a0c242d": {
          "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": ""
          }
        },
        "86479a7791b34265a2f7dcc68d20d8bd": {
          "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
          }
        },
        "642a5ac2f4b2427a89ef498643392651": {
          "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": ""
          }
        },
        "f2e003882f804290acac27ef5439f4cc": {
          "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
          }
        },
        "7abd1b443f3f4fdd9a2b589244125b9c": {
          "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": ""
          }
        },
        "03c5635e6de2450dbd9eea85d2e0d9fb": {
          "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_9db137265a0c487a935c2969a44558d0",
              "IPY_MODEL_ef212e2db02c4500aac47ecf8b770981",
              "IPY_MODEL_feaddee22c7e4742a880d808c1504c52"
            ],
            "layout": "IPY_MODEL_78ccfb12800b46c090fb98f9f19f2d9f"
          }
        },
        "9db137265a0c487a935c2969a44558d0": {
          "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_e1d54376d8734b64b814806e9fef5add",
            "placeholder": "​",
            "style": "IPY_MODEL_e76d113ae5904a04b565035a1fadfc88",
            "value": "100%"
          }
        },
        "ef212e2db02c4500aac47ecf8b770981": {
          "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_cb8db989e41345ea9f3ff017bdbd6ca2",
            "max": 1466,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_05f06f0b7a6c4abba0537eac94652aaa",
            "value": 1466
          }
        },
        "feaddee22c7e4742a880d808c1504c52": {
          "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_f54122d9b8df4735ab01b24745338daf",
            "placeholder": "​",
            "style": "IPY_MODEL_258d76ec903d41ce85ed7cd695d78703",
            "value": " 1466/1466 [00:21&lt;00:00, 106.74it/s]"
          }
        },
        "78ccfb12800b46c090fb98f9f19f2d9f": {
          "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
          }
        },
        "e1d54376d8734b64b814806e9fef5add": {
          "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
          }
        },
        "e76d113ae5904a04b565035a1fadfc88": {
          "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": ""
          }
        },
        "cb8db989e41345ea9f3ff017bdbd6ca2": {
          "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
          }
        },
        "05f06f0b7a6c4abba0537eac94652aaa": {
          "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": ""
          }
        },
        "f54122d9b8df4735ab01b24745338daf": {
          "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
          }
        },
        "258d76ec903d41ce85ed7cd695d78703": {
          "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": ""
          }
        },
        "be3d61d8fe314e81ae223d0e75158178": {
          "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_c6a8b9ce9e5f47fcb5d204d11725b509",
              "IPY_MODEL_a10dc24e06c2488ea2d0fbee58246b7b",
              "IPY_MODEL_472ccd92d91f475595b2f424d859d951"
            ],
            "layout": "IPY_MODEL_1207fc6f1f59478fbe4014cc758d1bb1"
          }
        },
        "c6a8b9ce9e5f47fcb5d204d11725b509": {
          "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_a6d9fadd10d94e83a1d374f1f38eb42c",
            "placeholder": "​",
            "style": "IPY_MODEL_7255f4c4a0554272bb37f38a55cb0da2",
            "value": "100%"
          }
        },
        "a10dc24e06c2488ea2d0fbee58246b7b": {
          "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_1bc074406aa947758cfb3686286b1b6f",
            "max": 1466,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_6bbba4cf396f4d6581e32f8db511c6d4",
            "value": 1466
          }
        },
        "472ccd92d91f475595b2f424d859d951": {
          "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_e761f4aba4d7443a986cd62e3c93b4b3",
            "placeholder": "​",
            "style": "IPY_MODEL_507b5beafe374b7f92759edccec7581d",
            "value": " 1466/1466 [00:17&lt;00:00, 89.46it/s]"
          }
        },
        "1207fc6f1f59478fbe4014cc758d1bb1": {
          "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
          }
        },
        "a6d9fadd10d94e83a1d374f1f38eb42c": {
          "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
          }
        },
        "7255f4c4a0554272bb37f38a55cb0da2": {
          "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": ""
          }
        },
        "1bc074406aa947758cfb3686286b1b6f": {
          "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
          }
        },
        "6bbba4cf396f4d6581e32f8db511c6d4": {
          "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": ""
          }
        },
        "e761f4aba4d7443a986cd62e3c93b4b3": {
          "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
          }
        },
        "507b5beafe374b7f92759edccec7581d": {
          "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": ""
          }
        },
        "8232384db7c749789535d33d54590666": {
          "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_3f46695168144e3da72a9f524d5e1950",
              "IPY_MODEL_33f53533420c40f1b70e03992a69b459",
              "IPY_MODEL_649131dc159f47069eb061e4dc009445"
            ],
            "layout": "IPY_MODEL_f149f3c00b0d4999a254c0dc99757191"
          }
        },
        "3f46695168144e3da72a9f524d5e1950": {
          "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_fe56a1e7d8d64a3d9f9521729810d7e1",
            "placeholder": "​",
            "style": "IPY_MODEL_da88b3466a3b4ab4b6763e336886aae0",
            "value": "100%"
          }
        },
        "33f53533420c40f1b70e03992a69b459": {
          "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_b1082cb5983141da877f56673e442a5f",
            "max": 7360,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_c38068e182eb46639b8cd8f57bc25ccd",
            "value": 7360
          }
        },
        "649131dc159f47069eb061e4dc009445": {
          "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_7f02f1d13dc24d4280439847a917b8cc",
            "placeholder": "​",
            "style": "IPY_MODEL_8a1543e994d54748bfb148719aa6f229",
            "value": " 7360/7360 [01:20&lt;00:00, 77.50it/s]"
          }
        },
        "f149f3c00b0d4999a254c0dc99757191": {
          "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
          }
        },
        "fe56a1e7d8d64a3d9f9521729810d7e1": {
          "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
          }
        },
        "da88b3466a3b4ab4b6763e336886aae0": {
          "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": ""
          }
        },
        "b1082cb5983141da877f56673e442a5f": {
          "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
          }
        },
        "c38068e182eb46639b8cd8f57bc25ccd": {
          "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": ""
          }
        },
        "7f02f1d13dc24d4280439847a917b8cc": {
          "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
          }
        },
        "8a1543e994d54748bfb148719aa6f229": {
          "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",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/Mostafa3zazi/CISC-873-DM-Data-Mining/blob/main/CISC_873_DM_F22_A4.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#CISC-873-DM-F22-A4: Airbnb price category prediction"
      ],
      "metadata": {
        "id": "U4F8PnGrGDvQ"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Download data"
      ],
      "metadata": {
        "id": "ctN_lrFcdvbz"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FLUqiPMEdhX1",
        "outputId": "239c2ea4-1d80-4784-bbbf-922ebb653191"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2022-03-24 23:24:14--  https://github.com/CISC-873/Information-2021/releases/download/data/a4.zip\n",
            "Resolving github.com (github.com)... 140.82.112.4\n",
            "Connecting to github.com (github.com)|140.82.112.4|:443... connected.\n",
            "HTTP request sent, awaiting response... 302 Found\n",
            "Location: https://objects.githubusercontent.com/github-production-release-asset-2e65be/406495726/4d095bba-8b9b-4be4-8738-83f8ff5b0d18?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20220324%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20220324T232239Z&X-Amz-Expires=300&X-Amz-Signature=9439e109891f06f90e0ae3062eea9438946f6b115ebf54a18f05e5ec95969911&X-Amz-SignedHeaders=host&actor_id=0&key_id=0&repo_id=406495726&response-content-disposition=attachment%3B%20filename%3Da4.zip&response-content-type=application%2Foctet-stream [following]\n",
            "--2022-03-24 23:24:14--  https://objects.githubusercontent.com/github-production-release-asset-2e65be/406495726/4d095bba-8b9b-4be4-8738-83f8ff5b0d18?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20220324%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20220324T232239Z&X-Amz-Expires=300&X-Amz-Signature=9439e109891f06f90e0ae3062eea9438946f6b115ebf54a18f05e5ec95969911&X-Amz-SignedHeaders=host&actor_id=0&key_id=0&repo_id=406495726&response-content-disposition=attachment%3B%20filename%3Da4.zip&response-content-type=application%2Foctet-stream\n",
            "Resolving objects.githubusercontent.com (objects.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
            "Connecting to objects.githubusercontent.com (objects.githubusercontent.com)|185.199.108.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 639078419 (609M) [application/octet-stream]\n",
            "Saving to: ‘a4.zip.1’\n",
            "\n",
            "a4.zip.1            100%[===================>] 609.47M   217MB/s    in 2.8s    \n",
            "\n",
            "2022-03-24 23:24:17 (217 MB/s) - ‘a4.zip.1’ saved [639078419/639078419]\n",
            "\n",
            "replace __MACOSX/._img_test? [y]es, [n]o, [A]ll, [N]one, [r]ename: "
          ]
        }
      ],
      "source": [
        "! wget https://github.com/CISC-873/Information-2021/releases/download/data/a4.zip\n",
        "\n",
        "! unzip -q a4.zip"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Inspecting Data"
      ],
      "metadata": {
        "id": "MHRzMh7EdySl"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "import os\n",
        "\n",
        "import pandas as pd\n",
        "from tqdm.notebook import tqdm\n",
        "from PIL import Image\n",
        "import pandas as pd\n",
        "import os\n",
        "import numpy as np\n",
        "from ast import literal_eval\n",
        "\n",
        "# after unzipping the datset\n",
        "# we find that (through the file browser on the left) there is a csv file and a \n",
        "# folder of images\n",
        "df = pd.read_csv('train_xy.csv')\n",
        "df"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 424
        },
        "id": "YnL5IGa2eMXV",
        "outputId": "00a3589f-fd21-4181-fba8-e8c318f9bd6c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                                summary               image  \\\n",
              "0     Spacious, sunny and cozy modern apartment in t...     img_train/0.jpg   \n",
              "1     Located in one of the most vibrant and accessi...     img_train/1.jpg   \n",
              "2     Logement coquet et douillet à 10 minutes du ce...     img_train/2.jpg   \n",
              "3     Beautiful and spacious (1076 sc ft, / 100 mc) ...     img_train/3.jpg   \n",
              "4     Très grand appartement ''rustique'' et très ag...     img_train/4.jpg   \n",
              "...                                                 ...                 ...   \n",
              "7622  Un grand logement 4 et 1/2, tout inclut, bien ...  img_train/7626.jpg   \n",
              "7623  Magnificent condo directly on the river. You w...  img_train/7627.jpg   \n",
              "7624  This apartment is perfect for anyone visiting ...  img_train/7628.jpg   \n",
              "7625  It is a cozy ,clean ,and comfortable apartment...  img_train/7629.jpg   \n",
              "7626  Modern country style (newly-renovated); open c...  img_train/7630.jpg   \n",
              "\n",
              "           type  price  \n",
              "0     Apartment      1  \n",
              "1     Apartment      0  \n",
              "2     Apartment      1  \n",
              "3     Apartment      1  \n",
              "4     Apartment      0  \n",
              "...         ...    ...  \n",
              "7622  Apartment      0  \n",
              "7623  Apartment      2  \n",
              "7624  Apartment      1  \n",
              "7625  Apartment      0  \n",
              "7626      House      1  \n",
              "\n",
              "[7627 rows x 4 columns]"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-e33d9f35-5068-47de-82d0-b88a01101057\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>summary</th>\n",
              "      <th>image</th>\n",
              "      <th>type</th>\n",
              "      <th>price</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Spacious, sunny and cozy modern apartment in t...</td>\n",
              "      <td>img_train/0.jpg</td>\n",
              "      <td>Apartment</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Located in one of the most vibrant and accessi...</td>\n",
              "      <td>img_train/1.jpg</td>\n",
              "      <td>Apartment</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Logement coquet et douillet à 10 minutes du ce...</td>\n",
              "      <td>img_train/2.jpg</td>\n",
              "      <td>Apartment</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>Beautiful and spacious (1076 sc ft, / 100 mc) ...</td>\n",
              "      <td>img_train/3.jpg</td>\n",
              "      <td>Apartment</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>Très grand appartement ''rustique'' et très ag...</td>\n",
              "      <td>img_train/4.jpg</td>\n",
              "      <td>Apartment</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7622</th>\n",
              "      <td>Un grand logement 4 et 1/2, tout inclut, bien ...</td>\n",
              "      <td>img_train/7626.jpg</td>\n",
              "      <td>Apartment</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7623</th>\n",
              "      <td>Magnificent condo directly on the river. You w...</td>\n",
              "      <td>img_train/7627.jpg</td>\n",
              "      <td>Apartment</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7624</th>\n",
              "      <td>This apartment is perfect for anyone visiting ...</td>\n",
              "      <td>img_train/7628.jpg</td>\n",
              "      <td>Apartment</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7625</th>\n",
              "      <td>It is a cozy ,clean ,and comfortable apartment...</td>\n",
              "      <td>img_train/7629.jpg</td>\n",
              "      <td>Apartment</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7626</th>\n",
              "      <td>Modern country style (newly-renovated); open c...</td>\n",
              "      <td>img_train/7630.jpg</td>\n",
              "      <td>House</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>7627 rows × 4 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-e33d9f35-5068-47de-82d0-b88a01101057')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-e33d9f35-5068-47de-82d0-b88a01101057 button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-e33d9f35-5068-47de-82d0-b88a01101057');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 1
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df.info()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "A4av9qPlezpz",
        "outputId": "e0df5580-c907-4c98-ee62-c9eb66cd1ad0"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "<class 'pandas.core.frame.DataFrame'>\n",
            "RangeIndex: 7627 entries, 0 to 7626\n",
            "Data columns (total 4 columns):\n",
            " #   Column   Non-Null Count  Dtype \n",
            "---  ------   --------------  ----- \n",
            " 0   summary  7326 non-null   object\n",
            " 1   image    7627 non-null   object\n",
            " 2   type     7627 non-null   object\n",
            " 3   price    7627 non-null   int64 \n",
            "dtypes: int64(1), object(3)\n",
            "memory usage: 238.5+ KB\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df.price.value_counts()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MuV99ZGLe3yo",
        "outputId": "22ae69fd-aa9b-4574-b776-b0c3604011b8"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0    4737\n",
              "1    2403\n",
              "2     487\n",
              "Name: price, dtype: int64"
            ]
          },
          "metadata": {},
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df['type'].value_counts()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gEflVqAWfAj0",
        "outputId": "23d0b9c8-fcac-43f5-9edd-beb2d9aa4155"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Apartment                 5765\n",
              "Condominium                691\n",
              "House                      406\n",
              "Loft                       324\n",
              "Townhouse                  167\n",
              "Serviced apartment          77\n",
              "Bed and breakfast           38\n",
              "Guest suite                 32\n",
              "Hostel                      26\n",
              "Bungalow                    25\n",
              "Guesthouse                  14\n",
              "Cottage                     12\n",
              "Aparthotel                  12\n",
              "Boutique hotel              10\n",
              "Other                        8\n",
              "Villa                        7\n",
              "Tiny house                   3\n",
              "Boat                         2\n",
              "Cabin                        2\n",
              "Camper/RV                    2\n",
              "Casa particular (Cuba)       1\n",
              "Hotel                        1\n",
              "Earth house                  1\n",
              "Castle                       1\n",
              "Name: type, dtype: int64"
            ]
          },
          "metadata": {},
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df = df.dropna()"
      ],
      "metadata": {
        "id": "Tu2AjKFsfFma"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "df.isna().sum()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "M-ps4dGKeNUk",
        "outputId": "26f87849-8876-4a35-ad72-b31d596f3a5e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "summary     0\n",
              "image       0\n",
              "type        0\n",
              "price       0\n",
              "type_cat    0\n",
              "dtype: int64"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df.price.value_counts()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8OaRFUdRf2HI",
        "outputId": "fdfc8a72-2e0a-406e-9d96-d301f344236b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0    4546\n",
              "1    2312\n",
              "2     468\n",
              "Name: price, dtype: int64"
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "#converting the type of \"type\" column to categorical and encode it to labels of numbers\n",
        "#assigning unique number to each type\n",
        "df['type_cat'] = df['type'].astype('category').cat.codes\n",
        "len_price = len(df.price.unique())\n",
        "len_type = len(df['type_cat'].unique())\n",
        "\n",
        "\n",
        "\n",
        "# split data (training/testing set)\n",
        "df_train, df_test = train_test_split(df, test_size=0.2, random_state=10)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ttcUJDg9f9kf",
        "outputId": "7f8da45f-aa3e-4fc4-ec05-56d4dab4a1e8"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:3: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  This is separate from the ipykernel package so we can avoid doing imports until\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Problem formulation\n",
        "The problem when people prepare to post a new listing on airbnb is, how much should one ask for? \n",
        "\n",
        "the objective is to predict the listing price based on the listing characteristics, in this way to optimize user experience and lower the bar to be a new host.\n",
        "\n",
        "input is: the data has 2 inputs, text summary and an image.\n",
        "\n",
        "output is:  a prediction of the price range of a new listing, along with the type of rental being advertised.\n",
        "\n",
        "this is a multi-class classification problem that consists of 2 tasks (the price and type) to solve this problem we will use techniques to preprocess text and images before the training process. for the models we need 2 types of model, one to train on images and one to train on text, the problem is that we have 2 different models and we want to use them to together to predict the type and price. Moreover, we need to decide which model we will use and the implementation.\n",
        "for image model we will use a CNN model and for text we will use RNN.\n",
        "Multi-modality learning will be used to make a model with can combine the two different types of data (image and text) and Multi-objective learning to predict both price and type using one combined model.\n",
        "\n",
        "An ideal solution is classifying the appropriate price range accurately for new Airbnb. the impact of this solution will optimize user experience and lower the bar to be a new host but the downside is that hosts may write fake summaries to fool the model and have higher price range (a solution to this problem can be add more weight on images than summary)"
      ],
      "metadata": {
        "id": "v0Lt3JOjE02b"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Data Preprocessing\n",
        "\n",
        "We have image and text data. \n",
        "\n",
        "- Image data:covert to gray scale then resize\n",
        "- Text data: tokenization and converting to integer IDs"
      ],
      "metadata": {
        "id": "FUoMACSf_MRK"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# preprocess image data\n",
        "import os\n",
        "\n",
        "def load_image(file):\n",
        "    try:\n",
        "      #read image, convert it to gray scale and resize it\n",
        "        image = Image.open(\n",
        "            file\n",
        "        ).convert('L').resize((64, 64))\n",
        "        arr = np.array(image)\n",
        "    except:\n",
        "        arr = np.zeros((64, 64))\n",
        "    return arr\n",
        "\n",
        "\n",
        "# loading images:\n",
        "x_train_image = np.array([load_image(i) for i in tqdm(df_train.image)])\n",
        "\n",
        "# loading summary: (force convert some of the non-string cell to string)\n",
        "x_train_text = df_train.summary.astype('str')\n",
        "\n",
        "# get price \n",
        "y_train_price = df_train.price\n",
        "\n",
        "# get type\n",
        "y_train_type = df_train['type_cat']"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 49,
          "referenced_widgets": [
            "efb7bcbb8bf2490391f4a238b89d4b22",
            "f712e239096c4bf79c16df9ba6fab628",
            "7f03538152024fbca69ef08d507381d1",
            "ef6087d76e1b41089f43efd1128e3484",
            "ff3653f7e11d4268b3d6096a20cf71eb",
            "fcc0bd48d43f4e4db0cad088f3513b10",
            "1ac3936fee0a490cae81918ae7e277b7",
            "2ff3209b9906406a8a6c775b414b2b9e",
            "46413c7c59604f1c8bf0a099873a8003",
            "4304108b4e584cc8967d93033e11a830",
            "06d78c89d6474ba5afe611816960791e"
          ]
        },
        "id": "3upEJzNJgwJG",
        "outputId": "30511c50-8672-44c8-aec3-c53ca0854f38"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/5860 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "efb7bcbb8bf2490391f4a238b89d4b22"
            }
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# check image loading\n",
        "import matplotlib.pyplot as plt\n",
        "plt.imshow(x_train_image[0, :, :])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 286
        },
        "id": "zr1BINQLhrsI",
        "outputId": "a106ef03-0754-4954-bcf0-1b201781de81"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f9294ef0fd0>"
            ]
          },
          "metadata": {},
          "execution_count": 11
        },
        {
          "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": [
        "# preprocess text data\n",
        "\n",
        "from tensorflow.keras.preprocessing.text import Tokenizer\n",
        "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
        "from sklearn.model_selection import train_test_split\n",
        "from pprint import pprint\n",
        "\n",
        "vocab_size = 40000\n",
        "max_len = 100\n",
        "\n",
        "\n",
        "# build vocabulary from training set\n",
        "tokenizer = Tokenizer(num_words=vocab_size)\n",
        "tokenizer.fit_on_texts(x_train_text)\n",
        "\n",
        "\n",
        "def _preprocess(list_of_text):\n",
        "  #add padding of zeroes to text so that each recored has the same lenght (the padding is post padding)\n",
        "    return pad_sequences(\n",
        "        tokenizer.texts_to_sequences(list_of_text),\n",
        "        maxlen=max_len,\n",
        "        padding='post',\n",
        "    )\n",
        "    \n",
        "\n",
        "# padding is done inside: \n",
        "x_train_text_id = _preprocess(x_train_text)\n",
        "\n",
        "print(x_train_text_id.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "IWokcNBfqwqL",
        "outputId": "8cbcf89d-658f-4cae-bc4b-7a0872006672"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(5860, 100)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "pprint(tokenizer.sequences_to_texts(x_train_text_id[:5]))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "om24VtVRs22k",
        "outputId": "ca5e6a99-c76d-4d6d-af9c-73911143e082"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "['featured in the new york times the wall street journal and fast company '\n",
            " 'we’ve hosted over 350k guests in 2 500 spaces with an average rating of 4 6 '\n",
            " 'stars each space is designed and run by us so you’ll always know what to '\n",
            " 'expect from your stay at sonder we’ve combined the best parts of home and '\n",
            " 'hotel 24 7 support via text email or phone coffee fresh towels and bathroom '\n",
            " 'essentials provided hotel standard cleaning before you arrive self check in',\n",
            " 'our home is on the first floor of a charming duplex it has two large '\n",
            " 'bedrooms a living room a washroom with a shower bathtub an open kitchen and '\n",
            " 'dining room a beautiful spacious deck with a bbq and a backyard our home is '\n",
            " \"perfect for young families it is close to numerous parks and has children's \"\n",
            " 'books toys and board games bath toys a sandbox in the backyard and a playpen '\n",
            " 'travel crib our extensive board game collection is also perfect for game '\n",
            " 'lovers of all ages',\n",
            " 'my place is good for couples solo adventurers business travellers and '\n",
            " 'families with kids the luxury condo is located in the heart of montreal in '\n",
            " \"the complexe les étoiles at the embassy suites hilton's building a free \"\n",
            " 'fully equipped gym included there is an indoor parking with 24h access in '\n",
            " 'and out for a competitive extra fee the booking from the greater montreal '\n",
            " 'area will not be accepted',\n",
            " 'beautiful entire apt for rent in the heart of the plateau 10 mins walk to '\n",
            " 'metro station mont royal 15mins to downtown 1 min to the park la fontaine 3 '\n",
            " 'mins walk to the ave montroyal',\n",
            " 'luxury unit in downtown montreal 1 queen bed 1 sofa bed steps away from the '\n",
            " 'famous st catherine street place des arts subway station excellent location '\n",
            " 'for tourists business travellers location is probably the best thing about '\n",
            " 'this loft less than 10min walk to old montreal metered parking on the street '\n",
            " 'indoor and outdoor parking lots available only steps away']\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "tokenizer.num_words"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "r7JXqA31s8B6",
        "outputId": "4b07cd93-aa61-4ac6-fd2e-ac0cdf42ce88"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "40000"
            ]
          },
          "metadata": {},
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Trials\n",
        "define some usefull functions and prepare our input data"
      ],
      "metadata": {
        "id": "yBUPgBcze6h-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from matplotlib import pyplot as plt\n",
        "def train_val_acc_plt (history):\n",
        "  plt.plot(history.history['price_sparse_categorical_accuracy'])\n",
        "  plt.plot(history.history['val_price_sparse_categorical_accuracy'])\n",
        "  plt.title('model accuracy')\n",
        "  plt.ylabel('accuracy')\n",
        "  plt.xlabel('epoch')\n",
        "  plt.legend(['train', 'val'], loc='upper left')\n",
        "  plt.show()"
      ],
      "metadata": {
        "id": "pUHsErE_hFLo"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def train_val_loss_plt (history):\n",
        "  plt.plot(history.history['price_loss'])\n",
        "  plt.plot(history.history['val_price_loss'])\n",
        "  plt.title('model loss')\n",
        "  plt.ylabel('loss')\n",
        "  plt.xlabel('epoch')\n",
        "  plt.legend(['train', 'loss'], loc='upper left')\n",
        "  plt.show()"
      ],
      "metadata": {
        "id": "xphy2vbHhUnC"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# loading images:\n",
        "x_test_image = np.array([load_image(i) for i in tqdm(df_test.image)])\n",
        "# loading summary: (force convert some of the non-string cell to string)\n",
        "x_test_text = _preprocess(df_test.summary.astype('str'))"
      ],
      "metadata": {
        "id": "yz0rOjh-h_do",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 49,
          "referenced_widgets": [
            "9d00817cf3554d728253e397f7fdd62e",
            "ddee07994bf642eab2d4a0467775b374",
            "94cc02ea8d604cc9bccd1fa282650353",
            "f5ec5d17b93e49599c92b8b22e038499",
            "133a410ae8984f179de85f9c5dd2b6e9",
            "acc5d2e63056423fb528242b40b05d64",
            "447c9ce6ccb3481abb5da85ce471739d",
            "a8e4ba30b30a4a5c839e59460d015588",
            "2c78f8833c4543b18328df3b649299e6",
            "ac6166ea9509499395f0a548c74b1787",
            "5d1b7b1e71d44480b051a660219e6fae"
          ]
        },
        "outputId": "3477502a-12a7-465d-b28e-2ad57c5dccd3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/1466 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "9d00817cf3554d728253e397f7fdd62e"
            }
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Trial 1\n",
        "working with text input only (Embedding)\n",
        "\n",
        "using simple model with text as input and multi-task learning to predict (price and type)\n"
      ],
      "metadata": {
        "id": "i0zq4Pgve9KB"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "286d1840-7342-4b0c-ecf9-91d59ef2c013",
        "id": "HkbF-kXdfD-D"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_2 (InputLayer)           [(None, 100)]        0           []                               \n",
            "                                                                                                  \n",
            " embedding_1 (Embedding)        (None, 100, 100)     4000000     ['input_2[0][0]']                \n",
            "                                                                                                  \n",
            " tf.math.reduce_mean_1 (TFOpLam  (None, 100)         0           ['embedding_1[0][0]']            \n",
            " bda)                                                                                             \n",
            "                                                                                                  \n",
            " price (Dense)                  (None, 3)            303         ['tf.math.reduce_mean_1[0][0]']  \n",
            "                                                                                                  \n",
            " type (Dense)                   (None, 24)           2424        ['tf.math.reduce_mean_1[0][0]']  \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 4,002,727\n",
            "Trainable params: 4,002,727\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "from __future__ import absolute_import, division, print_function, unicode_literals\n",
        "import collections\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "from tensorflow.keras.layers import Conv2D, Flatten, Dense, MaxPool2D\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "\n",
        "\n",
        "# input for text text.\n",
        "in_text = keras.Input(batch_shape=(None, max_len))\n",
        "\n",
        "# text part\n",
        "# simple average of embedding. you can change it to anything else as needed\n",
        "embedded = keras.layers.Embedding(tokenizer.num_words, 100)(in_text)\n",
        "averaged = tf.reduce_mean(embedded, axis=1) #reduces input_tensor along the dimensions given in axis by computing the mean of elements across the dimensions in axis\n",
        "\n",
        "\n",
        "\n",
        "# multi-task learning (each is a multi-class classification)\n",
        "# one dense layer for each task\n",
        "p_price = Dense(len_price, activation='softmax', name='price')(averaged)\n",
        "p_type = Dense(len_type, activation='softmax', name='type')(averaged)\n",
        "\n",
        "\n",
        "# define model input/output using keys.\n",
        "model = keras.Model(\n",
        "    inputs={\n",
        "        'summary': in_text,\n",
        "    },\n",
        "    outputs={\n",
        "        'price': p_price,\n",
        "        'type': p_type,\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "# compile model with optimizer, loss values for each task, loss \n",
        "# weights for each task.\n",
        "#usign adam optimizer and sparse categorical crossentropy loss.\n",
        "model.compile(\n",
        "    optimizer=Adam(),\n",
        "    loss={\n",
        "        'price': 'sparse_categorical_crossentropy',\n",
        "        'type': 'sparse_categorical_crossentropy',\n",
        "    },\n",
        "    loss_weights={\n",
        "        'price': 0.5,\n",
        "        'type': 0.5,       \n",
        "    },\n",
        "    metrics={\n",
        "        'price': ['SparseCategoricalAccuracy'],\n",
        "        'type': ['SparseCategoricalAccuracy'],\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(\n",
        "    x={\n",
        "        'summary': x_train_text_id, #input to model\n",
        "    },\n",
        "    y={\n",
        "        'price': y_train_price, #model output\n",
        "        'type': y_train_type,\n",
        "    },\n",
        "    epochs=20,\n",
        "    batch_size=16,\n",
        "    validation_split=0.2,\n",
        "    #using early stopping with patience of 3 and monitor val_price_loss \n",
        "    callbacks=[\n",
        "        tf.keras.callbacks.EarlyStopping(monitor='val_price_loss', patience=3,\n",
        "                                         restore_best_weights=True)\n",
        "    ],\n",
        "    verbose=1\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "L2s46Ubsf-fe",
        "outputId": "5c415187-da91-4f9e-96e1-e888dc32cb13"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "293/293 [==============================] - 8s 19ms/step - loss: 1.2397 - price_loss: 0.8740 - type_loss: 1.6054 - price_sparse_categorical_accuracy: 0.6171 - type_sparse_categorical_accuracy: 0.7459 - val_loss: 0.9020 - val_price_loss: 0.8272 - val_type_loss: 0.9768 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 2/20\n",
            "293/293 [==============================] - 4s 15ms/step - loss: 0.8895 - price_loss: 0.7999 - type_loss: 0.9790 - price_sparse_categorical_accuracy: 0.6267 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8790 - val_price_loss: 0.8081 - val_type_loss: 0.9498 - val_price_sparse_categorical_accuracy: 0.6229 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 3/20\n",
            "293/293 [==============================] - 3s 10ms/step - loss: 0.8551 - price_loss: 0.7676 - type_loss: 0.9425 - price_sparse_categorical_accuracy: 0.6446 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8594 - val_price_loss: 0.7870 - val_type_loss: 0.9318 - val_price_sparse_categorical_accuracy: 0.6374 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 4/20\n",
            "293/293 [==============================] - 3s 10ms/step - loss: 0.8230 - price_loss: 0.7313 - type_loss: 0.9147 - price_sparse_categorical_accuracy: 0.6670 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8447 - val_price_loss: 0.7714 - val_type_loss: 0.9181 - val_price_sparse_categorical_accuracy: 0.6399 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 5/20\n",
            "293/293 [==============================] - 3s 10ms/step - loss: 0.7886 - price_loss: 0.6928 - type_loss: 0.8845 - price_sparse_categorical_accuracy: 0.6881 - type_sparse_categorical_accuracy: 0.7547 - val_loss: 0.8272 - val_price_loss: 0.7501 - val_type_loss: 0.9042 - val_price_sparse_categorical_accuracy: 0.6647 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 6/20\n",
            "293/293 [==============================] - 3s 10ms/step - loss: 0.7546 - price_loss: 0.6557 - type_loss: 0.8534 - price_sparse_categorical_accuracy: 0.7182 - type_sparse_categorical_accuracy: 0.7549 - val_loss: 0.8182 - val_price_loss: 0.7489 - val_type_loss: 0.8876 - val_price_sparse_categorical_accuracy: 0.6621 - val_type_sparse_categorical_accuracy: 0.7679\n",
            "Epoch 7/20\n",
            "293/293 [==============================] - 3s 10ms/step - loss: 0.7214 - price_loss: 0.6225 - type_loss: 0.8203 - price_sparse_categorical_accuracy: 0.7368 - type_sparse_categorical_accuracy: 0.7581 - val_loss: 0.8038 - val_price_loss: 0.7333 - val_type_loss: 0.8743 - val_price_sparse_categorical_accuracy: 0.6741 - val_type_sparse_categorical_accuracy: 0.7696\n",
            "Epoch 8/20\n",
            "293/293 [==============================] - 3s 10ms/step - loss: 0.6885 - price_loss: 0.5915 - type_loss: 0.7855 - price_sparse_categorical_accuracy: 0.7573 - type_sparse_categorical_accuracy: 0.7622 - val_loss: 0.7949 - val_price_loss: 0.7319 - val_type_loss: 0.8578 - val_price_sparse_categorical_accuracy: 0.6698 - val_type_sparse_categorical_accuracy: 0.7722\n",
            "Epoch 9/20\n",
            "293/293 [==============================] - 3s 10ms/step - loss: 0.6556 - price_loss: 0.5632 - type_loss: 0.7479 - price_sparse_categorical_accuracy: 0.7735 - type_sparse_categorical_accuracy: 0.7703 - val_loss: 0.7928 - val_price_loss: 0.7443 - val_type_loss: 0.8414 - val_price_sparse_categorical_accuracy: 0.6638 - val_type_sparse_categorical_accuracy: 0.7713\n",
            "Epoch 10/20\n",
            "293/293 [==============================] - 3s 10ms/step - loss: 0.6223 - price_loss: 0.5366 - type_loss: 0.7080 - price_sparse_categorical_accuracy: 0.7833 - type_sparse_categorical_accuracy: 0.7805 - val_loss: 0.7810 - val_price_loss: 0.7331 - val_type_loss: 0.8289 - val_price_sparse_categorical_accuracy: 0.6698 - val_type_sparse_categorical_accuracy: 0.7739\n",
            "Epoch 11/20\n",
            "293/293 [==============================] - 4s 13ms/step - loss: 0.5896 - price_loss: 0.5100 - type_loss: 0.6692 - price_sparse_categorical_accuracy: 0.7974 - type_sparse_categorical_accuracy: 0.7918 - val_loss: 0.7736 - val_price_loss: 0.7321 - val_type_loss: 0.8151 - val_price_sparse_categorical_accuracy: 0.6792 - val_type_sparse_categorical_accuracy: 0.7833\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_val_acc_plt(history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "hkrSedT1hgl3",
        "outputId": "079e4151-7a11-4117-89c5-dadcb54f3067"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "train_val_loss_plt (history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "j2pMDLv_hiFw",
        "outputId": "470cdfcb-9ebc-4847-f806-38a0e885034a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3deXxU1f3/8dcnGyEQ9rAlQMK+ySKLKDsiIC5YtSouFVu0Wjeo2mLbb7W2/WlrW5eKC+61CCKgYkVREQirECDsi0nYEpaEQFgSsk0+vz/uRUYMkEAmk8x8no/HPJi5y8xnUPLOOefec0RVMcYYY04X4u8CjDHGVE0WEMYYY0plAWGMMaZUFhDGGGNKZQFhjDGmVBYQxhhjSmUBYUwFEJF3ROQvZTx2p4gMv9D3McbXLCCMMcaUygLCGGNMqSwgTNBwu3YeE5H1IpIrIm+KSBMR+VxEjonI1yJS3+v4a0Vkk4jkiMhCEenkta+niKxxz/sAiDzts64WkWT33GUi0u08a75bRFJE5JCIzBGR5u52EZHnRCRTRI6KyAYR6eruGy0im93aMkTk0fP6CzNBzwLCBJsbgCuA9sA1wOfA74AYnH8PDwGISHtgGjDB3TcX+FREIkQkAvgYeA9oAHzovi/uuT2Bt4BfAg2B14A5IlKjPIWKyDDgaeAmoBmwC5ju7h4BDHK/R133mGx335vAL1U1GugKfFOezzXmJAsIE2z+raoHVDUDWAx8q6prVTUf+Ajo6R53M/CZqn6lqkXAP4CawGVAPyAceF5Vi1R1JrDK6zPuAV5T1W9V1aOq7wIF7nnlcRvwlqquUdUC4HHgUhGJB4qAaKAjIKq6RVX3uecVAZ1FpI6qHlbVNeX8XGMACwgTfA54PT9Ryuva7vPmOL+xA6CqJcAeINbdl6E/nOlyl9fzVsAjbvdSjojkAC3c88rj9BqO47QSYlX1G+AlYDKQKSJTRKSOe+gNwGhgl4gsEpFLy/m5xgAWEMacyV6cH/SA0+eP80M+A9gHxLrbTmrp9XwP8FdVref1iFLVaRdYQy2cLqsMAFV9UVV7AZ1xupoec7evUtUxQGOcrrAZ5fxcYwALCGPOZAZwlYhcLiLhwCM43UTLgOVAMfCQiISLyPVAX69zXwfuFZFL3MHkWiJylYhEl7OGacBdItLDHb/4fzhdYjtFpI/7/uFALpAPlLhjJLeJSF23a+woUHIBfw8miFlAGFMKVd0G3A78GziIM6B9jaoWqmohcD0wDjiEM14x2+vcJOBunC6gw0CKe2x5a/ga+D9gFk6rpQ1wi7u7Dk4QHcbphsoGnnX33QHsFJGjwL04YxnGlJvYgkHGGGNKYy0IY4wxpbKAMMYYUyoLCGOMMaWygDDGGFOqMH8XUFEaNWqk8fHx/i7DGGOqldWrVx9U1ZjS9gVMQMTHx5OUlOTvMowxploRkV1n2mddTMYYY0plAWGMMaZUFhDGGGNKFTBjEKUpKioiPT2d/Px8f5fic5GRkcTFxREeHu7vUowxASKgAyI9PZ3o6Gji4+P54cSbgUVVyc7OJj09nYSEBH+XY4wJEAHdxZSfn0/Dhg0DOhwARISGDRsGRUvJGFN5AjoggIAPh5OC5XsaYypPwAfEuZSosu/ICQqLPf4uxRhjqpSgD4hiTwmHjhey+9AJSnww9XlOTg4vv/xyuc8bPXo0OTk5FV6PMcaUVdAHRERYKHH1a5JXWMz+IxXfh3+mgCguLj7reXPnzqVevXoVXo8xxpRVQF/FVFZ1oyJoVOjh4PECatUIpW7NiAp770mTJpGamkqPHj0IDw8nMjKS+vXrs3XrVrZv3851113Hnj17yM/P5+GHH+aee+4BTk0dcvz4ca688koGDBjAsmXLiI2N5ZNPPqFmzZoVVqMxxpQmaALiT59uYvPeo2c95kSRB1WlZnhomQZ9OzevwxPXdDnrMc888wwbN24kOTmZhQsXctVVV7Fx48bvL0d96623aNCgASdOnKBPnz7ccMMNNGzY8Afv8d133zFt2jRef/11brrpJmbNmsXtt99+zvqMMeZCBH0Xk7fIMOevI7/Yd2u89+3b9wf3Krz44ot0796dfv36sWfPHr777rsfnZOQkECPHj0A6NWrFzt37vRZfcYYc1LQtCDO9Zv+SUdPFLEzO5eGtSKIrR9V4XXUqlXr++cLFy7k66+/Zvny5URFRTFkyJBS72WoUaPG989DQ0M5ceJEhddljDGnsxbEaerUDCcmugbZuYXk5BVe8PtFR0dz7NixUvcdOXKE+vXrExUVxdatW1mxYsUFf54xxlSUoGlBlEeTOpHkFXhIP3yCyPBQIsNDz/u9GjZsSP/+/enatSs1a9akSZMm3+8bNWoUr776Kp06daJDhw7069evIso3xpgKIeqDa//9oXfv3nr6gkFbtmyhU6dO5/V+hcUlpGQeJyxUaBtTm5CQqn+n8oV8X2NMcBKR1arau7R91sV0BhFhIbRoUJP8Ig97c6zP3xgTfHwaECIySkS2iUiKiEwqZX9LEVkgImtFZL2IjHa3x4vICRFJdh+v+rLOM4mODKdxdCSH8go5lHvh4xHGGFOd+GwMQkRCgcnAFUA6sEpE5qjqZq/D/gDMUNVXRKQzMBeId/elqmoPX9VXVk3q1CCvsJi9OSeIiriw8QhjjKlOfNmC6AukqGqaqhYC04Expx2jQB33eV1grw/rOS8iQosGUYSIsCs7D09JYIzZGGPMufgyIGKBPV6v091t3p4EbheRdJzWw4Ne+xLcrqdFIjKwtA8QkXtEJElEkrKysiqw9B8KDw2hZYOaFBZ7yMg5QaAM7BtjzNn4e5B6LPCOqsYBo4H3RCQE2Ae0VNWewK+B90Wkzuknq+oUVe2tqr1jYmJ8WmjtyHCa1Ikkx8YjjDFBwpcBkQG08Hod527z9gtgBoCqLgcigUaqWqCq2e721UAq0N6HtZZJTHQNoiPD2XsknxOFZ5+N9aTatWv7uCpjjPENXwbEKqCdiCSISARwCzDntGN2A5cDiEgnnIDIEpEYd5AbEWkNtAPSfFhrmYgILerXJCxE2HUoD0+J7+ZsMsYYf/NZQKhqMfAAMA/YgnO10iYReUpErnUPewS4W0TWAdOAcep08A8C1otIMjATuFdVD/mq1vIICw2hZYMoioqV9MNlH49QVR577DG6du3KRRddxAcffADAvn37GDRoED169KBr164sXrwYj8fDuHHjvj/2ueee8+VXMsaYUvl0qg1VnYsz+Oy97Y9ezzcD/Us5bxYwq0KL+XwS7N9QIW9VC+jgKeFI3Y5kX/kMjWrXOOc5s2fPJjk5mXXr1nHw4EH69OnDoEGDeP/99xk5ciS///3v8Xg85OXlkZycTEZGBhs3bgSwleWMMX7h70Hqais8VKgRFsK+I/nklWE8YsmSJYwdO5bQ0FCaNGnC4MGDWbVqFX369OHtt9/mySefZMOGDURHR9O6dWvS0tJ48MEH+eKLL6hT50fj88YY43PBM1nflc9U6NsJEOUpITzzOLuz82jbuDZhoeXP20GDBpGYmMhnn33GuHHj+PWvf83PfvYz1q1bx7x583j11VeZMWMGb731VoXWb4wx52ItiAsQFhpCy4ZRFJWcezxi4MCBfPDBB3g8HrKyskhMTKRv377s2rWLJk2acPfddzN+/HjWrFnDwYMHKSkp4YYbbuAvf/kLa9asqcRvZYwxjuBpQfhIVEQYzepGsjfnBAePFxATHVnqcT/5yU9Yvnw53bt3R0T4+9//TtOmTXn33Xd59tlnCQ8Pp3bt2vznP/8hIyODu+66ixL3Kqmnn366Mr+SMcYANt13hVBVdh/K4+iJYlrH1KJWDf/krk33bYwpL5vu+1yK8uACglJEiKtfk4gwYfehPIo9dn+EMab6s4DwFEHWNjiwCY5kQGHueYVFaIhzf0RxidOaCJSWmTEmeAV8QJzzB7WEQL1WEB4FuVlwcDtkboaje8vdsqgZEUbzepEcLygm81jBBVZePhZIxpiKFtCD1JGRkWRnZ9OwYUNEzrBkaEgoRDVwHiXFkH8EThyG4wecR2gNqFnfeYSXPgDtrUFUBHkFHjKP5lMrIpTakeEV/K1+TFXJzs4mMvLc9RljTFkFdEDExcWRnp7OeU0FXhIORSeg6CgUu7OWh0Y4LY2IKAg5819diSrZxwrISofG0TUIrYT1rCMjI4mLi/P55xhjgkdAB0R4eDgJCQkX/kZH98HmT2D9LEhf6WxrfjF0vR66/ATq/vgH8/YDx7j2pSX0aFGP//7ikvO6ic4YY/wpoC9z9Ymc3bDpI9g4G/YlO9ta9IOuN0DnMRDd5PtDZ65O59EP1/HgsLY8MqKD72szxphyOttlrhYQFyI7FTbNdsIic7Mz4B0/wAmLTtdCVAN+O3M9HyTt4Z27+jCkQ+PKrc8YY87BAqIyZG5xgmLjLDiU6oxRtB5KYafruCWxITuOhfLZQwNpXq+m/2o0xpjTWEBUJlXYv94Jio0fwZHdaEgE33i6sa7e5Tx474OE14z2d5XGGANYQPiPKqQnwcZZ5K+bSWR+FoUhkUR0utLphmp7RZkunTXGGF+xgKgKSjy8+f77RGz9mJtrrSai4DBEREOnq6HHrdBqAITYlU7GmMplAVFFFBR7uPGV5aRnH2XedULjnf9zLp8tPAb1WkL3W6H7LdCgAi7NNcaYMvDbZH0iMkpEtolIiohMKmV/SxFZICJrRWS9iIz22ve4e942ERnpyzorS42wUCbfejHFhHL3kmgKrn4RHt0O178BDdrAor/Biz3g7atg7VQoOO7vko0xQcxnLQgRCQW2A1cA6cAqYKy7DvXJY6YAa1X1FRHpDMxV1Xj3+TSgL9Ac+Bpor6qeM31edWhBnDRv035++d5qxl0Wz5PXdjm1I2cPrJ8Oye/DoTQIrwVdrnO6oFpeZl1QxpgK568WRF8gRVXTVLUQmA6MOe0YBU4uuFwX2Os+HwNMV9UCVd0BpLjvFxBGdmnKLwYk8M6ynXy6bu+pHfVawKDH4ME18PN5zp3am+fAO1c5LYuFz8DhXf4r3BgTVHwZELHAHq/X6e42b08Ct4tIOjAXeLAc5yIi94hIkogkndd8S37021EdubhlPR6avpbHPlxH5rH8UztFoGU/GPMSPLoNfjIF6sfDwqfhhW7wztWQPM2ZmtwYY3zE330WY4F3VDUOGA28JyJlrklVp6hqb1XtHRMT47MifSEiLIR3f96Xewa25uPkDIb9YxGvLkqloPi0XrSIWtD9ZrhzDkzYAEN/D0f2wMf3wj/aw8f3w86lF7TgkTHGlMaXAZEBtPB6Hedu8/YLYAaAqi4HIoFGZTy32ouODOfx0Z34cuJg+rVuwDOfb2XEc4l8uWl/6es71GsJg38DDyXDXZ874xObP4Z3RjtdUIv+7swVZYwxFcCXg9RhOIPUl+P8cF8F3Kqqm7yO+Rz4QFXfEZFOwHycrqTOwPucGqSeD7QLlEHqM0ncnsVT/9tMSuZxBrRtxP9d3ZkOTc9x13VhrjNOkTwVdi52tiUMgh63Q6drnKnJjTHmDPx2H4R72erzQCjwlqr+VUSeApJUdY57tdLrQG2cAevfqOqX7rm/B34OFAMTVPXzs31WIAQEQJGnhKkrdvGvr7aTW+jhtktaMnF4e+rXijj3yYd3wbrpTljk7HJuxOtyHfS4zRnTONOiScaYoGU3ylVDh3MLee7r7fx3xS6iI8OZOLwdt/VrRXhZ1pUoKYHdy5zLZTd9DEW50KC1c7lst1ucq6WMMQYLiGpt6/6jPPXpZpalZtOucW3+eE1nBrYrx4B8wXHYMscJi52LAYHWg527ttuPcJZSNcYELQuIak5V+XLzAf762RZ2H8pjeKcm/OGqTsQ3qlW+Nzq0w+mCWve+M5gtIRDXB9pcDm2HQ/MezhrdxpigYQERIAqKPby1ZCcvffMdhZ4Sft4/gQeGtSU6Mrx8b1RS4iydmvK189ibDKjTmmgzzA2MyyG6qU++hzGm6rCACDCZR/N5dt42PlydTqPaETw2sgM39mpBaMh5DkLnHoTUBZA6H1LmQ26ms71JVyco2lzuDHKH1ai4L2GMqRIsIALUuj05/OnTTazZnUPX2Do8cU0X+sQ3uLA3LSmBAxudlkXqN7B7BZQUOfNCJQx0uqLaDIOGbSrmSxhj/MoCIoCpKnPW7eWZz7ey70g+13RvzqQrOxJbUUubFhyDHYtPdUfluHNB1U9wWhdth0P8QKhRu2I+zxhTqSwggkBeYTGvLkrjtUWpiMAvB7Xh3sFtqBlRgYPOqs4ssynznbDYuRiK8iAk3OmCOhkYTbraPRfGVBMWEEEk/XAeT3++lc/W76NZ3UgeH92Ja7o1Q3zxA7u4AHYvdwNjPmS6N8nXbnJqoLv1UKjVsOI/2xhTISwggtDKHYf406eb2LT3KL1b1eeJa7pwUVxd337o0b3OuEXKfOfP/BxAoHlPp2XR9nKI7Q2hYb6twxhTZhYQQcpTonyYtId/fLmN7NxCbrw4jsdGdaBxdKTvP7zEA3vXumMX8yEjCbQEatSFDqOcu7rjB9kiSMb4mQVEkDuaX8RL36Tw9tId1AgL5YFhbbmrfzw1wirxprgThyFtIXz3NWz5FAqOQN2WTlD0GOusd2GMqXQWEAaAHQdz+etnm/l6SyatGkbxl+u6lm/ajopSdAK2fgZr/+uEBupcCdXzduh0rc1Aa0wlsoAwP7BoexZ/mrOJtIO53Ngrjj9c1Yl6UWWYLdYXcvacmoH28A5nBtquP3GmK2/R166GMsbHLCDMj+QXeXhx/ne8lphG/agI/jymC1de1Mx/BanCrmVOUJycgbZhO6cLqvtYqOPH2owJYBYQ5ow2Zhzht7PWs2nvUUZ2acKfx3SlcZ1KGMQ+m4JjsPkTpwtq93JnUsE2l0PP26DDaJvyw5gKZAFhzqrYU8Lri3fw/NfbiQgL4f+u6sxPe8f55t6J8spOdVoVydPg2F5nQsGLfuqMVzTr7u/qjKn2LCBMmaRlHWfS7A2s3HGI/m0b8vRPutGyYRUZMC7xQNoCWDvVGeD2FECTi5xWxUU32c14xpwnCwhTZiUlyvsrd/PM51spLinh0REduKt/wvnPFOsLJw7DhplOy2LvWmeqjw6jnIHttsPtRjxjysGfa1KPAl7AWZP6DVV95rT9zwFD3ZdRQGNVrefu8wAb3H27VfXas32WBUTF2nfkBL//aCPfbM2ke4t6/P2GbnRoGu3vsn7swCZntbx10yHvoDPNR7ebnS6omA7+rs6YKs8vASEiocB24AogHVgFjFXVzWc4/kGgp6r+3H19XFXLPEWoBUTFOzlT7J8+3cyx/CLuG9KW+4e2qdwb7MrKUwTb5zmtiu3zQD3OtB49b4OuN0Ckj6cZMaaa8ldAXAo8qaoj3dePA6jq02c4fhnwhKp+5b62gKgiDuUW8tSnm/g4eS/tGtfmbzd24+KWVXgt6+OZsP4DZ7wiawuERUKna6DrjRDVAELDnW6p0AinOyo0wnmEnHzu7rdpQEwQ8FdA3AiMUtXx7us7gEtU9YFSjm0FrADiVNXjbisGkoFi4BlV/fhsn2cB4XsLtmbyu482sP9oPuMui+fRER2oVaMK9/erwt41TlBsnAn5R8p3voSeCozSQiUk/NS+0wPmB8eHQ6P2zmSFjdrbzX+mSjlbQFSVf923ADNPhoOrlapmiEhr4BsR2aCqqd4nicg9wD0ALVu2rLxqg9TQjo35cuIg/v7FNt5eupOvNh/g6esv8s90HWUhArG9nMfI/+dMGFiU76yQ5yl0uqU8RV6vi50/S4pO7fMUQklx2Y4vzAXP4R8fX5zvjI/MA+q2gDZDncH0hMFQs56//5aMOaMq0cUkImuB+1V12Rne6x3gf6o680yfZy2IyrVyxyEmzVpfNabrqA4O73KmQE+dD2mLoOCo00KJ631q7YzmPSGkCo7vVCclJU5Ah9n/i2Xlry6mMJxB6suBDJxB6ltVddNpx3UEvgAS1C1GROoDeapaICKNgOXAmDMNcIMFhD9Uuek6qgtPEaQnOWGRMt+5VBd1bgJsPeRUYNRp7udCq4EThyF9NaSvch4ZSVBwHBp3gmY9oHkP58+mXSG8gpbhDTD+vMx1NPA8zmWub6nqX0XkKSBJVee4xzwJRKrqJK/zLgNeA0qAEOB5VX3zbJ9lAeE/VXK6juokN9u5CfDkYkvH9zvbG3eGNsOcsGh5GYQH+d9piQeytsKelU7Apq+Eg9udfRLi/H3F9XGCdv962JvsdO2B01qz0CiV3ShnfK5KT9dRnahC5uZTCy3tXu6MZ4TVhPj+p1oXwTDYnZvttAj2rHRbB2ug8JizL6qhEwYnH7EXQ43T7tNRhaMZTlDsS3ZaaqeHRkzHU4HRvIeznnqQTTdvAWEqTZWerqM6KsyFnUtPdUdlf+dsD7TBbk+xs6a5d+vgUJqzT0Kd3/bj+kBcX2fcpkHr8wvIH4WG+2du1qnPiulwWkvjooAODQsIU6mqxXQd1VWgDHYfz3RaBScDYe8aKMpz9tVq7KwFcrJ10Lynb39AqzrrqXsHxt61XqER4rQ0AjQ0LCCMX1Sb6Tqqq7IMdre81PlBFlrDubLn5E2Bldk9VVwIBzY4tZ7sLsrZ5ewLCYdm3X7YXVSvpf+7z0oNjWTIzXT2Swg06vDD7qmmF0FELf/WfR4sIIzfVKvpOqq7Mw12lyYk3FlXIzTcCY/QiB8GSFiNU89/sM89x/vcH7yPu09CIHOLEwr7kp17QQCim0OLPqe6i5p1rz6D76pwbN+Pu6eOHzh1TEiY+wh3WnGh4T98HRLmbgt1t3m/DvPaFnaG116P7987DOrEOmu7nwcLCON33tN1dGgSzb9u7k6X5jY/ks+cHOzet9754ewpcqZI9xQ6v9F7Ctyb+NxtJx/FBV7Hnrb/B/tOvk+hM+9VaUJrOL9Ze7cO6sZW7t9DZTi6z2m9ZW5y1lsvKXbGVEqKnZsof/D65DaPe9Ol1/YzvvaU8j7uay1xaojrA+O/Pq/yLSBMlfHN1gP8dtYGcvIKmTC8Pb8c1JqwUJvzqFor8ZQeIHVibfU/Xzt5Y6CWnHdL7GwBYf8yTaUa1rEJX04YxIjOTXl23jZuem05Ow/m+rsscyFCQp37CWrWg9oxUDfOucrIwsH3QkKc7j8fddNZQJhKV79WBC/d2pMXbulBSuZxrnxhMf9dsYtAac0aEygsIIxfiAhjesQyb+IgesfX5w8fb2Tc26s4cDTf36UZY1wWEMavmtWtyX9+3pc/j+nCtzuyGfFcIp+u2+vvsowxWECYKkBEuOPSeOY+NJCERrV4cNpaHpy2lpy8Qn+XZkxQs4AwVUbrmNrMvPdSHh3Rns837GPEc4ks3Jbp77KMCVoWEKZKCQsN4YFh7fj4/v7UrRnOuLdX8YePN5BXWOzv0owJOhYQpkrqGluXTx8cwN0DE5j67W5Gv7CY1bsO+7ssY4KKBYSpsiLDQ/n9VZ2Zdnc/ijzKT19dxrPztlJYXOLv0owJChYQpsrr17ohX0wYyE97tWDyglTGTF7Ktv3H/F2WMQHPAsJUC9GR4fztxm68/rPeZB3L55p/L2FKYiqeEru5zhhfsYAw1coVnZswb8IghnaM4f/N3crYKSvYcyjP32UZE5AsIEy107B2DV69vRf//Gl3tuw7yqjnE5m+crdN1WFMBfNpQIjIKBHZJiIpIjKplP3PiUiy+9guIjle++4Uke/cx52+rNNUPyLCDb3i+GLiILq3qMek2RsY/24Smcdsqg5jKkqZAkJEHhaROuJ4U0TWiMiIc5wTCkwGrgQ6A2NFpLP3Mao6UVV7qGoP4N/AbPfcBsATwCVAX+AJEalf3i9nAl9svZr89xeX8MQ1nVmScpCRzyXy+YZ9/i7LmIBQ1hbEz1X1KDACqA/cATxzjnP6AimqmqaqhcB0YMxZjh8LTHOfjwS+UtVDqnoY+AoYVcZaTZAJCRHu6p/AZw8NoEWDKO6buoZff5DMkRNF/i7NmGqtrAFxcoHY0cB7qrrJa9uZxAJ7vF6nu9t+/OYirYAE4JvynCsi94hIkogkZWVlnfNLmMDWtnE0s+67jAnD2/HJur2Mej6RJd8d9HdZxlRbZQ2I1SLyJU5AzBORaKAi71a6BZipeqa1C0unqlNUtbeq9o6JianAckx1FR4awoTh7Zl932XUjAjl9je/5ck5mzhRWK7/tYwxlD0gfgFMAvqoah4QDtx1jnMygBZer+PcbaW5hVPdS+U915gf6d6iHnMfGshd/eN5Z9lOrvr3YtbtyTn3icaY75U1IC4FtqlqjojcDvwBOHKOc1YB7UQkQUQicEJgzukHiUhHnHGN5V6b5wEjRKS+Ozg9wt1mTJlFhofyxDVdeH/8JeQXerj+lWX8/YutFBRba8KYsihrQLwC5IlId+ARIBX4z9lOUNVi4AGcH+xbgBmquklEnhKRa70OvQWYrl4XsavqIeDPOCGzCnjK3WZMuV3WthGfTxjEDRfH8vLCVK5+cQnJ1pow5pykLDcXicgaVb1YRP4IZKjqmye3+b7Esundu7cmJSX5uwxTxS3clsnjszdw4Gg+dw9szcQr2hMZHurvsozxGxFZraq9S9tX1hbEMRF5HOfy1s9EJARnHMKYamVIh8Z8OXEQN/dpyWuJae404tY4NaY0ZQ2Im4ECnPsh9uMMGj/rs6qM8aHoyHCevv4ipo6/hILiEm58dTlPfbrZrnQy5jRlCgg3FKYCdUXkaiBfVc86BmFMVde/bSPmTRzE7Ze04q2lOxj1QiLfpmX7uyxjqoyyTrVxE7AS+ClwE/CtiNzoy8KMqQy1a4Tx5+u6Mu3ufqjCzVNW8MQnG8ktsCVOjSnrIPU64ApVzXRfxwBfq2p3H9dXZjZIbS5UXmExz87bxjvLdhJbryZ/v6Ebl7Vt5O+yjPGpihikDjkZDq7scpxrTLUQFRHGE9d0YcYvLyU8NIRb3/iW3320gWP5NqeTCU5l/SH/hYjME5FxIjIO+AyY67uyjPGfPvENmPvQQO4emMD0lbsZ9fxiErfbXF8m+JSpiwlARG4A+rsvF6vqR8/E6rYAABY1SURBVD6r6jxYF5PxhTW7D/PYh+tIzcrl5t4t+P3VnagTaVd4m8Bxti6mMgdEVWcBYXwlv8jDC/O/47VFqTSOjuTp6y9iaMfG/i7LmApx3mMQInJMRI6W8jgmIkd9U64xVUtkeCi/HdWRj37Vnzo1w7jrnVX8ekYyR/JsbMIEtrMGhKpGq2qdUh7Rqlqnsoo0piro3qIenz44gAeHteWT5L0Mf24RX27a7++yjPEZuxLJmHKoERbKIyM68Mn9/WlUuwb3vLeah6at5VBuob9LM6bCWUAYcx66xtblk/v7M3F4e+Zu2MeI5xbZWtgm4FhAGHOeIsJCeHh4Oz59cABN60Zy39Q13D91DQePF/i7NGMqhAWEMReoU7M6fPSr/jw2sgNfbT7AiOcSmbNuL4FyhaAJXhYQxlSA8NAQ7h/alv89NIAWDaJ4aNpafvneajKP5fu7NGPOmwWEMRWofZNoZt17KZOu7MjC7Vlc8a9EZq9Jt9aEqZYsIIypYGGhIdw7uA1zHxpIm5ha/HrGOu58exVpWcf9XZox5eLTgBCRUSKyTURSRGTSGY65SUQ2i8gmEXnfa7tHRJLdxxxf1mmML7RtXJsP772MP17dmbW7DjPy+USe+XyrTSVuqg2fTbUhIqHAduAKIB1YBYxV1c1ex7QDZgDDVPWwiDT2mlL8uKrWLuvn2VQbpirLPJbP3z7fxqw16TStE8nvrurENd2aISL+Ls0EuYqY7vt89AVSVDVNVQuB6cCY0465G5isqocBTptS3JiA0Tg6kn/e1J1Z911Go+gIHpq2llumrGDrfpuxxlRdvgyIWGCP1+t0d5u39kB7EVkqIitEZJTXvkgRSXK3X1faB4jIPe4xSVlZNh2zqfp6tarPJ/cP4K8/6cq2A8e46sUlPDlnE0dO2LxOpurx9yB1GNAOGAKMBV4XkXruvlZus+dW4HkRaXP6yao6RVV7q2rvmJiYyqrZmAsSGiLcdkkrFjwyhLF9W/Du8p0M+8dCZiTtoaTErnYyVYcvAyIDaOH1Os7d5i0dmKOqRaq6A2fMoh2Aqma4f6YBC4GePqzVmEpXv1YEf7nuIj59YACtGkbxm5nruf6VZaxPz/F3acYAvg2IVUA7EUkQkQjgFuD0q5E+xmk9ICKNcLqc0kSkvojU8NreH9iMMQGoa2xdZt57Gf/8aXfSD59gzOSlPD57vU0AaPzOZwGhqsXAA8A8YAswQ1U3ichTInKte9g8IFtENgMLgMdUNRvoBCSJyDp3+zPeVz8ZE2hCQoQbesXxzaOD+UX/BGYkpTP0Hwt5b/lOPNbtZPzEVpQzpgrafuAYT87ZxLLUbDo3q8OfxnShT3wDf5dlApC/LnM1xpyn9k2imTr+El6+7WJy8gr56avLmfhBMplHbW4nU3ksIIypokSE0Rc14+tHBvPA0LZ8tn4fQ/+xkNcT0yjylPi7PBMELCCMqeKiIsJ4dGQHvpw4iEtaN+Svc7dw5QuLWfLdQX+XZgKcBYQx1UR8o1q8Na4Pb97ZmyJPCbe/+S33/Xc16Yfz/F2aCVBh/i7AGFM+l3dqQv+2jXhjcRovLUhhwbZM7h/SlrsHtSYyPNTf5ZkAYi0IY6qhyPBQHhjWjvmPDOHyjk3451fbGfFcIl9vPuDv0kwAsYAwphqLrVeTybddzNTxlxARFsL4/yRx19sr2XEw19+lmQBgAWFMAOjfthGfPzyQP1zViVU7DzPyuUSenbeVvEJbe8KcPwsIYwJEeGgI4we25ptHBnN192ZMXpDKsH8sYtbqdJsE0JwXCwhjAkzjOpH866YezLz3UprUqcEjH67j6n8vYWmKXRZryscCwpgA1Tu+AR/9qj8v3NKDIyeKuO2Nbxn39kq27T/m79JMNWEBYUwACwkRxvSIZf4jg/nd6I6s3nWYK19I5PHZ623aDnNONlmfMUHkcG4h//4mhfdW7CQ8NIR7BrXmnkGtiYqwW6KClU3WZ4wBnEWK/nhNZ76aOJghHWJ4/uvvGPLsQqav3G3TipsfsYAwJgjFN6rFy7f1YtZ9lxJXvyaTZm9g9AuLWbAtk0DpVTAXzgLCmCDWq1UDZt13GS/fdjH5xR7uensVd7y5kk17j/i7NFMFWEAYE+ROTiv+1cTB/PHqzmzce4Sr/72ERz9cx74jJ/xdnvEjG6Q2xvzAkbwiJi9M4Z2lOwkJgfEDWvPLwa2Jjgz3d2nGB/w2SC0io0Rkm4ikiMikMxxzk4hsFpFNIvK+1/Y7ReQ793GnL+s0xpxSNyqc343uxPxHBjOic1NeWpDCkGcX8t6KXRTbQkVBxWctCBEJBbYDVwDpwCpgrKpu9jqmHTADGKaqh0WksapmikgDIAnoDSiwGuilqofP9HnWgjDGN9btyeGvc7ewcsch2sTUYtKVnRjeqTEi4u/STAXwVwuiL5CiqmmqWghMB8acdszdwOSTP/hVNdPdPhL4SlUPufu+Akb5sFZjzBl0b1GPD+7px5Q7eqHA3f9J4pYpK1ifnuPv0oyP+TIgYoE9Xq/T3W3e2gPtRWSpiKwQkVHlOBcRuUdEkkQkKSsrqwJLN8Z4ExFGdGnKvAmD+POYLnyXeZxrX1rKw9PX2op2AczfVzGFAe2AIcBY4HURqVfWk1V1iqr2VtXeMTExPirRGHNSeGgId1waz8LHhvCrIW34YuN+hv1zEU9/voUjJ4r8XZ6pYL4MiAyghdfrOHebt3RgjqoWqeoOnDGLdmU81xjjJ3Uiw/nNqI4seHQIV3drxpTENIY8u4C3l+6gsNgGsgOFLwNiFdBORBJEJAK4BZhz2jEf47QeEJFGOF1OacA8YISI1BeR+sAId5sxpgppXq8m/7qpB58+MIBOzerwp083M+K5RXy+YZ/dkR0AfBYQqloMPIDzg30LMENVN4nIUyJyrXvYPCBbRDYDC4DHVDVbVQ8Bf8YJmVXAU+42Y0wV1DW2LlPHX8Lb4/oQHhrCfVPXcN3kpSzYalN3VGd2o5wxpkIVe0qYtSadF+enkJFzgh4t6jHxivYMatfILo2tgs52masFhDHGJwqLS5i5Op3JC5yguLilExQD2lpQVCUWEMYYvyko9vBhkhMU+47k0ye+PhOHt+fSNg0tKKoACwhjjN8VFHv4YNUeJi9I4cDRAvomNODXV7SnX+uG/i4tqFlAGGOqjPwiD9NX7mbywlSyjhVwaeuGTLyiPX0TGvi7tKBkAWGMqXLyizxM/XY3ryxM5eDxAvq3bcjE4e3pHW9BUZksIIwxVdaJQg9Tv93Fq4tSOXi8kIHtGjFheHt6tarv79KCggWEMabKyyss5r3lu3gtMY1DuYUMbh/DxCva06NFmWffMefBAsIYU23kFhTzn+W7mJKYyuG8IoZ2cIKiW5wFhS9YQBhjqp3jBcW8u2wnry9OIyeviOGdGjNheHu6xtb1d2kBxQLCGFNtHcsv4p2lTlAczS/mis5NmDC8HV2aW1BUBAsIY0y1dzS/iLeX7OSNJWkcyy9mZJcmTBjenk7N6vi7tGrNAsIYEzCOnCjizSU7eHvJDo4VFDP6oqY8fHl7OjSN9ndp1ZIFhDEm4OTkFTpBsXQnuYXFjL6oGRMub0e7JhYU5WEBYYwJWIdzC3ljSRrvLN1JXpGHq7s151dD2ljXUxlZQBhjAt6h3EKmJKbx3vKd5BZ6GNaxMfcPbUOvVnZn9tlYQBhjgsaRvCLeXb6Tt5fu4HBeEX0TGnD/0La2HsUZWEAYY4JOXmEx01bu4fXENPYfzadrbB1+NaQtI7s0JTTEguIkCwhjTNAqLC7h47UZvLIolR0Hc2ndqBb3Dm7DdT1jiQjz2arL1cbZAsKnfzsiMkpEtolIiohMKmX/OBHJEpFk9zHea5/Ha/scX9ZpjAlcEWEh3NSnBV//ejCTb72YyPBQfjNrPYOfXcBbS3aQV1js7xKrLJ+1IEQkFNgOXAGkA6uAsaq62euYcUBvVX2glPOPq2rtsn6etSCMMWWhqizansXLC1JZufMQDWpFcNdl8fzs0njqRoX7u7xKd7YWRJgPP7cvkKKqaW4R04ExwOaznmWMMT4kIgzp0JghHRqTtPMQLy9M5Z9fbee1xDRu69eSXwxIoHF0pL/LrBJ82cUUC+zxep3ubjvdDSKyXkRmikgLr+2RIpIkIitE5LrSPkBE7nGPScrKyqrA0o0xwaB3fAPeGteHuQ8NZGjHxryemMaAvy3g9x9tYHd2nr/L8ztfdjHdCIxS1fHu6zuAS7y7k0SkIXBcVQtE5JfAzao6zN0Xq6oZItIa+Aa4XFVTz/R51sVkjLlQOw/m8lpiKrNWZ+BR5ZpuzbhvSNuAnsbDX4PUGYB3iyDO3fY9Vc1W1QL35RtAL699Ge6facBCoKcPazXGGOIb1eLp67uR+Juh3HVZPF9uPsDI5xMZ/24Sa3Yf9nd5lc6XAbEKaCciCSISAdwC/OBqJBFp5vXyWmCLu72+iNRwnzcC+mNjF8aYStK0biR/uLozS387jAnD27Fq5yGuf3kZY6esYPF3WQTK7QHn4rNBalUtFpEHgHlAKPCWqm4SkaeAJFWdAzwkItcCxcAhYJx7eifgNREpwQmxZ7yvfjLGmMpQv1YEE4a3Z/zA1kz7djevL07jjjdX0i2uLr8a0oYRnZsSEsA33dmNcsYYU0YFxR5mr8ng1UWp7MrOo01MLe4b0pYxPZoTHlo9b7qzO6mNMaYCFXtKmLtxPy8vSGHr/mPE1qvJPYNac3OfFkSGh/q7vHKxgDDGGB9QVRZsy+TlBakk7TpMo9oR3NU/gTsubUWdyOpx050FhDHG+Ni3adm8vDCVRduziK4Rxh2XtuLnAxJoVLuGv0s7KwsIY4ypJBszjvDKwlTmbtxHRGgIt/Rpwd2DWhNXP8rfpZXKAsIYYypZatZxXluUyuw1zu1f1/WM5d7BbWjbuMxTzFUKCwhjjPGTjJwTvJ6YxvRVuykoLmFUl6b8akhbLoqr6+/SAAsIY4zxu+zjBby9dCfvLt/JsfxiBrZrxP1D23JJQgO/rnRnAWGMMVXE0fwipq7YzZtL0jh4vJBerepz/9A2DO3Q2C9BYQFhjDFVTH6RhxlJe3htURoZOSfo2DSaXw1ty1UXNavUJVEtIIwxpooq8pQwJ3kvryxKJSXzOK0aRnHv4DZcf3EsNcJ8f9OdBYQxxlRxJSXKl5sP8PLCFNanH6FJnRrcPbA1Y/u2pFYN363tZgFhjDHVhKqyNCWbyQtSWJ6WTb2ocO66LIE7L2tFvaiICv88CwhjjKmG1uw+zMsLUvl6ywFqRYRyW79WjB+QQOM6FbckqgWEMcZUY1v3H+WVhal8um4vYSEh3Ng7jnsHtaFlwwu/O9sCwhhjAsCu7FxeS0xjZlI6xSUlXNu9+QUviWoBYYwxAeTA0XzeXLKDqSt2kVvo4apuzXhpbM/zuo/ibAHhu6FxY4wxPtGkTiS/G92JXw1pw7vLdlHo8fjkJjsLCGOMqabqRUXw8PB2Pnt/n66RJyKjRGSbiKSIyKRS9o8TkSwRSXYf47323Ski37mPO31ZpzHGmB/zWQtCREKBycAVQDqwSkTmqOrm0w79QFUfOO3cBsATQG9AgdXuuYd9Va8xxpgf8mULoi+QoqppqloITAfGlPHckcBXqnrIDYWvgFE+qtMYY0wpfBkQscAer9fp7rbT3SAi60Vkpoi0KM+5InKPiCSJSFJWVlZF1W2MMQYfj0GUwadAvKp2w2klvFuek1V1iqr2VtXeMTExPinQGGOClS8DIgNo4fU6zt32PVXNVtUC9+UbQK+ynmuMMca3fBkQq4B2IpIgIhHALcAc7wNEpJnXy2uBLe7zecAIEakvIvWBEe42Y4wxlcRnVzGparGIPIDzgz0UeEtVN4nIU0CSqs4BHhKRa4Fi4BAwzj33kIj8GSdkAJ5S1UO+qtUYY8yPBcxUGyKSBey6gLdoBBysoHKqi2D7zsH2fcG+c7C4kO/cSlVLHcQNmIC4UCKSdKb5SAJVsH3nYPu+YN85WPjqO/v7KiZjjDFVlAWEMcaYUllAnDLF3wX4QbB952D7vmDfOVj45DvbGIQxxphSWQvCGGNMqSwgjDHGlCroA+Jca1YEGhFpISILRGSziGwSkYf9XVNlEZFQEVkrIv/zdy2VQUTquZNgbhWRLSJyqb9r8jURmej+f71RRKaJSKS/a6poIvKWiGSKyEavbQ1E5Ct3/Zyv3BkoLlhQB4TXmhVXAp2BsSLS2b9V+Vwx8Iiqdgb6AfcHwXc+6WFOTecSDF4AvlDVjkB3Avy7i0gs8BDQW1W74szgcIt/q/KJd/jx8geTgPmq2g6Y776+YEEdEFzYmhXVkqruU9U17vNjOD80SpuGPaCISBxwFc6kkAFPROoCg4A3AVS1UFVz/FtVpQgDaopIGBAF7PVzPRVOVRNxpibyNoZTs2G/C1xXEZ8V7AFR1jUrApKIxAM9gW/9W0mleB74DVDi70IqSQKQBbztdqu9ISK1/F2UL6lqBvAPYDewDziiql/6t6pK00RV97nP9wNNKuJNgz0ggpaI1AZmARNU9ai/6/ElEbkayFTV1f6upRKFARcDr6hqTyCXCup2qKrcfvcxOOHYHKglIrf7t6rKp869CxVy/0KwB0RQrjshIuE44TBVVWf7u55K0B+4VkR24nQjDhOR//q3JJ9LB9JV9WTrcCZOYASy4cAOVc1S1SJgNnCZn2uqLAdOLp/g/plZEW8a7AFxzjUrAo2ICE6/9BZV/Ze/66kMqvq4qsapajzOf+NvVDWgf7NU1f3AHhHp4G66HNjsx5Iqw26gn4hEuf+fX06AD8x7mQPc6T6/E/ikIt7UZ+tBVAdnWrPCz2X5Wn/gDmCDiCS7236nqnP9WJPxjQeBqe4vP2nAXX6ux6dU9VsRmQmswblaby0BOO2GiEwDhgCNRCQdeAJ4BpghIr/AWfbgpgr5LJtqwxhjTGmCvYvJGGPMGVhAGGOMKZUFhDHGmFJZQBhjjCmVBYQxxphSWUAYUwWIyJBgmWXWVB8WEMYYY0plAWFMOYjI7SKyUkSSReQ1d42J4yLynLsOwXwRiXGP7SEiK0RkvYh8dHKOfhFpKyJfi8g6EVkjIm3ct6/ttX7DVPduYGP8xgLCmDISkU7AzUB/Ve0BeIDbgFpAkqp2ARbh3NkK8B/gt6raDdjgtX0qMFlVu+PMFXRyFs6ewASctUla49z1bozfBPVUG8aU0+VAL2CV+8t9TZxJ0UqAD9xj/gvMdtdjqKeqi9zt7wIfikg0EKuqHwGoaj6A+34rVTXdfZ0MxANLfP+1jCmdBYQxZSfAu6r6+A82ivzfaced7/w1BV7PPdi/T+Nn1sVkTNnNB24Ukcbw/TrArXD+Hd3oHnMrsERVjwCHRWSgu/0OYJG7il+6iFznvkcNEYmq1G9hTBnZbyjGlJGqbhaRPwBfikgIUATcj7MYT193XybOOAU40y6/6gaA92yqdwCvichT7nv8tBK/hjFlZrO5GnOBROS4qtb2dx3GVDTrYjLGGFMqa0EYY4wplbUgjDHGlMoCwhhjTKksIIwxxpTKAsIYY0ypLCCMMcaU6v8DMcEcWn+fG/wAAAAASUVORK5CYII=\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(model.evaluate(x= {\n",
        "    'summary': x_test_text},\n",
        "  y= {'price': df_test['price'],\n",
        "    'type':df_test['type_cat']}))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3z05wJ4og21E",
        "outputId": "d577008a-13cf-4f6e-82b3-e3a315dda230"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "46/46 [==============================] - 0s 6ms/step - loss: 0.7909 - price_loss: 0.7189 - type_loss: 0.8628 - price_sparse_categorical_accuracy: 0.6739 - type_sparse_categorical_accuracy: 0.7735\n",
            "[0.7908505797386169, 0.7189457416534424, 0.8627556562423706, 0.6739426851272583, 0.7735334038734436]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "the result is very good for the first trail so i tried to submit on kaggel and got this score\n",
        "kaggel Score: 0.67744\n",
        "\n",
        "now we will try to improve the acc using more complix models like lstm and gru"
      ],
      "metadata": {
        "id": "yVuqkgbsixO7"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Trial 2\n",
        "working with text input only (LSTM)\n",
        "\n",
        "using text as input and multi-task learning to predict (price and type) and replace the reduce_mean layer with an LSTM layer\n",
        "\n",
        "LSTM may capture more relations between words and imporve the score"
      ],
      "metadata": {
        "id": "q4ADrGXPi3jR"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "keras.backend.clear_session()"
      ],
      "metadata": {
        "id": "JbRxJ6o0pP1x"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8130b7d8-96b8-4b16-e219-cdbf9fb8dc4f",
        "id": "XuCSXfZbi3jS"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model_1\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_2 (InputLayer)           [(None, 100)]        0           []                               \n",
            "                                                                                                  \n",
            " embedding_1 (Embedding)        (None, 100, 100)     4000000     ['input_2[0][0]']                \n",
            "                                                                                                  \n",
            " lstm_1 (LSTM)                  (None, 5)            2120        ['embedding_1[0][0]']            \n",
            "                                                                                                  \n",
            " price (Dense)                  (None, 3)            18          ['lstm_1[0][0]']                 \n",
            "                                                                                                  \n",
            " type (Dense)                   (None, 24)           144         ['lstm_1[0][0]']                 \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 4,002,282\n",
            "Trainable params: 4,002,282\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "# input for text.\n",
        "in_text = keras.Input(batch_shape=(None, max_len))\n",
        "\n",
        "# text part\n",
        "embedded = keras.layers.Embedding(tokenizer.num_words, 100)(in_text)\n",
        "LSTM_layer = keras.layers.LSTM(5 , dropout= 0.2)(embedded)\n",
        "\n",
        "\n",
        "# multi-task learning (each is a multi-class classification)\n",
        "# one dense layer for each task\n",
        "p_price = Dense(len_price, activation='softmax', name='price')(LSTM_layer)\n",
        "p_type = Dense(len_type, activation='softmax', name='type')(LSTM_layer)\n",
        "\n",
        "\n",
        "# define model input/output using keys.\n",
        "model = keras.Model(\n",
        "    inputs={\n",
        "        'summary': in_text,\n",
        "    },\n",
        "    outputs={\n",
        "        'price': p_price,\n",
        "        'type': p_type,\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "# compile model with optimizer, loss values for each task, loss \n",
        "# weights for each task.\n",
        "model.compile(\n",
        "    optimizer=Adam(),\n",
        "    loss={\n",
        "        'price': 'sparse_categorical_crossentropy',\n",
        "        'type': 'sparse_categorical_crossentropy',\n",
        "    },\n",
        "    loss_weights={\n",
        "        'price': 0.5,\n",
        "        'type': 0.5,       \n",
        "    },\n",
        "    metrics={\n",
        "        'price': ['SparseCategoricalAccuracy'],\n",
        "        'type': ['SparseCategoricalAccuracy'],\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(\n",
        "    x={\n",
        "        'summary': x_train_text_id,\n",
        "    },\n",
        "    y={\n",
        "        'price': y_train_price,\n",
        "        'type': y_train_type,\n",
        "    },\n",
        "    epochs=20,\n",
        "    batch_size=16,\n",
        "    validation_split=0.2,\n",
        "    callbacks=[\n",
        "        tf.keras.callbacks.EarlyStopping(monitor='val_price_loss', patience=3,\n",
        "                                         restore_best_weights=True)\n",
        "    ],\n",
        "    verbose=1\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d8ed3d58-0069-4390-8a4c-fe96cf9665a3",
        "id": "zeTk1EBZi3jU"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "293/293 [==============================] - 11s 31ms/step - loss: 1.4634 - price_loss: 0.8608 - type_loss: 2.0660 - price_sparse_categorical_accuracy: 0.5983 - type_sparse_categorical_accuracy: 0.6122 - val_loss: 1.0393 - val_price_loss: 0.8371 - val_type_loss: 1.2414 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 2/20\n",
            "293/293 [==============================] - 8s 29ms/step - loss: 0.9758 - price_loss: 0.8326 - type_loss: 1.1190 - price_sparse_categorical_accuracy: 0.6248 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.9320 - val_price_loss: 0.8365 - val_type_loss: 1.0276 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 3/20\n",
            "293/293 [==============================] - 9s 30ms/step - loss: 0.9337 - price_loss: 0.8321 - type_loss: 1.0354 - price_sparse_categorical_accuracy: 0.6248 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.9185 - val_price_loss: 0.8374 - val_type_loss: 0.9996 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 4/20\n",
            "293/293 [==============================] - 9s 32ms/step - loss: 0.9243 - price_loss: 0.8316 - type_loss: 1.0171 - price_sparse_categorical_accuracy: 0.6248 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.9135 - val_price_loss: 0.8366 - val_type_loss: 0.9904 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 5/20\n",
            "293/293 [==============================] - 8s 29ms/step - loss: 0.9206 - price_loss: 0.8313 - type_loss: 1.0099 - price_sparse_categorical_accuracy: 0.6250 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.9118 - val_price_loss: 0.8364 - val_type_loss: 0.9872 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 6/20\n",
            "293/293 [==============================] - 8s 29ms/step - loss: 0.9186 - price_loss: 0.8310 - type_loss: 1.0062 - price_sparse_categorical_accuracy: 0.6254 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.9109 - val_price_loss: 0.8364 - val_type_loss: 0.9854 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 7/20\n",
            "293/293 [==============================] - 9s 29ms/step - loss: 0.9175 - price_loss: 0.8308 - type_loss: 1.0041 - price_sparse_categorical_accuracy: 0.6252 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.9114 - val_price_loss: 0.8388 - val_type_loss: 0.9840 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 8/20\n",
            "293/293 [==============================] - 8s 29ms/step - loss: 0.9162 - price_loss: 0.8304 - type_loss: 1.0021 - price_sparse_categorical_accuracy: 0.6259 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.9098 - val_price_loss: 0.8364 - val_type_loss: 0.9832 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 9/20\n",
            "293/293 [==============================] - 9s 29ms/step - loss: 0.9046 - price_loss: 0.8316 - type_loss: 0.9776 - price_sparse_categorical_accuracy: 0.6261 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8946 - val_price_loss: 0.8376 - val_type_loss: 0.9515 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 10/20\n",
            "293/293 [==============================] - 9s 29ms/step - loss: 0.8549 - price_loss: 0.8319 - type_loss: 0.8779 - price_sparse_categorical_accuracy: 0.6265 - type_sparse_categorical_accuracy: 0.7545 - val_loss: 0.8808 - val_price_loss: 0.8367 - val_type_loss: 0.9249 - val_price_sparse_categorical_accuracy: 0.6143 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 11/20\n",
            "293/293 [==============================] - 9s 30ms/step - loss: 0.7989 - price_loss: 0.8277 - type_loss: 0.7701 - price_sparse_categorical_accuracy: 0.6280 - type_sparse_categorical_accuracy: 0.7545 - val_loss: 0.8848 - val_price_loss: 0.8341 - val_type_loss: 0.9356 - val_price_sparse_categorical_accuracy: 0.6212 - val_type_sparse_categorical_accuracy: 0.7628\n",
            "Epoch 12/20\n",
            "293/293 [==============================] - 9s 30ms/step - loss: 0.7518 - price_loss: 0.7933 - type_loss: 0.7102 - price_sparse_categorical_accuracy: 0.6448 - type_sparse_categorical_accuracy: 0.7858 - val_loss: 0.8786 - val_price_loss: 0.8091 - val_type_loss: 0.9481 - val_price_sparse_categorical_accuracy: 0.6399 - val_type_sparse_categorical_accuracy: 0.7381\n",
            "Epoch 13/20\n",
            "293/293 [==============================] - 9s 29ms/step - loss: 0.6727 - price_loss: 0.6965 - type_loss: 0.6489 - price_sparse_categorical_accuracy: 0.7148 - type_sparse_categorical_accuracy: 0.8148 - val_loss: 0.8838 - val_price_loss: 0.7947 - val_type_loss: 0.9729 - val_price_sparse_categorical_accuracy: 0.6493 - val_type_sparse_categorical_accuracy: 0.7244\n",
            "Epoch 14/20\n",
            "293/293 [==============================] - 8s 29ms/step - loss: 0.5971 - price_loss: 0.5817 - type_loss: 0.6124 - price_sparse_categorical_accuracy: 0.7846 - type_sparse_categorical_accuracy: 0.8259 - val_loss: 0.9053 - val_price_loss: 0.8100 - val_type_loss: 1.0005 - val_price_sparse_categorical_accuracy: 0.6357 - val_type_sparse_categorical_accuracy: 0.7159\n",
            "Epoch 15/20\n",
            "293/293 [==============================] - 9s 29ms/step - loss: 0.5435 - price_loss: 0.4956 - type_loss: 0.5913 - price_sparse_categorical_accuracy: 0.8340 - type_sparse_categorical_accuracy: 0.8304 - val_loss: 0.9303 - val_price_loss: 0.8356 - val_type_loss: 1.0249 - val_price_sparse_categorical_accuracy: 0.6254 - val_type_sparse_categorical_accuracy: 0.7090\n",
            "Epoch 16/20\n",
            "293/293 [==============================] - 8s 29ms/step - loss: 0.5052 - price_loss: 0.4369 - type_loss: 0.5735 - price_sparse_categorical_accuracy: 0.8575 - type_sparse_categorical_accuracy: 0.8298 - val_loss: 0.9493 - val_price_loss: 0.8725 - val_type_loss: 1.0261 - val_price_sparse_categorical_accuracy: 0.6263 - val_type_sparse_categorical_accuracy: 0.7253\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_val_acc_plt(history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "475a399c-7706-407a-dfa9-29dad56319be",
        "id": "CHXalVJIi3jV"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "train_val_loss_plt (history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "66dfb1a2-ff15-4301-e31d-49916e52a1db",
        "id": "nzc3RH8Ki3jW"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "print(model.evaluate(x= {\n",
        "    'summary': x_test_text},\n",
        "  y= {'price': df_test['price'],\n",
        "    'type':df_test['type_cat']}))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2a09384b-bbda-4b10-dedb-f3dbc908a3f5",
        "id": "Tky-x3D-i3jY"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "46/46 [==============================] - 1s 10ms/step - loss: 0.8731 - price_loss: 0.7949 - type_loss: 0.9513 - price_sparse_categorical_accuracy: 0.6241 - type_sparse_categorical_accuracy: 0.7401\n",
            "[0.873106062412262, 0.7949399352073669, 0.9512724876403809, 0.624147355556488, 0.7401091456413269]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "the result are lower that before : price_sparse_categorical_accuracy: 0.6241\n",
        "\n",
        "lets try gru and hope to improve the results."
      ],
      "metadata": {
        "id": "lKc55NHFi3jY"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Trial 3\n",
        "working with text input only (GRU)\n",
        "\n",
        "using text as input and multi-task learning to predict (price and type) and replace the reduce_mean layer with a GRU layer.\n",
        "\n",
        "let's hope the accuracy improve"
      ],
      "metadata": {
        "id": "tY_3eGugl2yc"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "keras.backend.clear_session()"
      ],
      "metadata": {
        "id": "0bZKNGskpwT8"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "13d3edba-1744-4609-e86b-17a86020ae22",
        "id": "tkvTMOR5l2y6"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_1 (InputLayer)           [(None, 100)]        0           []                               \n",
            "                                                                                                  \n",
            " embedding (Embedding)          (None, 100, 100)     4000000     ['input_1[0][0]']                \n",
            "                                                                                                  \n",
            " gru (GRU)                      (None, 5)            1605        ['embedding[0][0]']              \n",
            "                                                                                                  \n",
            " price (Dense)                  (None, 3)            18          ['gru[0][0]']                    \n",
            "                                                                                                  \n",
            " type (Dense)                   (None, 24)           144         ['gru[0][0]']                    \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 4,001,767\n",
            "Trainable params: 4,001,767\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "# input for text.\n",
        "in_text = keras.Input(batch_shape=(None, max_len))\n",
        "\n",
        "# text part\n",
        "embedded = keras.layers.Embedding(tokenizer.num_words, 100)(in_text)\n",
        "# averaged = tf.reduce_mean(embedded, axis=1)\n",
        "GRU_layer = keras.layers.GRU(5 , dropout= 0.2)(embedded)\n",
        "\n",
        "\n",
        "# multi-task learning (each is a multi-class classification)\n",
        "# one dense layer for each task\n",
        "p_price = Dense(len_price, activation='softmax', name='price')(GRU_layer)\n",
        "p_type = Dense(len_type, activation='softmax', name='type')(GRU_layer)\n",
        "\n",
        "\n",
        "# define model input/output using keys.\n",
        "model = keras.Model(\n",
        "    inputs={\n",
        "        'summary': in_text,\n",
        "    },\n",
        "    outputs={\n",
        "        'price': p_price,\n",
        "        'type': p_type,\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "# compile model with optimizer, loss values for each task, loss \n",
        "# weights for each task.\n",
        "model.compile(\n",
        "    optimizer=Adam(),\n",
        "    loss={\n",
        "        'price': 'sparse_categorical_crossentropy',\n",
        "        'type': 'sparse_categorical_crossentropy',\n",
        "    },\n",
        "    loss_weights={\n",
        "        'price': 0.5,\n",
        "        'type': 0.5,       \n",
        "    },\n",
        "    metrics={\n",
        "        'price': ['SparseCategoricalAccuracy'],\n",
        "        'type': ['SparseCategoricalAccuracy'],\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(\n",
        "    x={\n",
        "        'summary': x_train_text_id,\n",
        "    },\n",
        "    y={\n",
        "        'price': y_train_price,\n",
        "        'type': y_train_type,\n",
        "    },\n",
        "    epochs=20,\n",
        "    batch_size=16,\n",
        "    validation_split=0.2,\n",
        "    callbacks=[\n",
        "        tf.keras.callbacks.EarlyStopping(monitor='val_price_loss', patience=3,\n",
        "                                         restore_best_weights=True)\n",
        "    ],\n",
        "    verbose=1\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9c088064-d131-413b-bb2c-4c8989e22332",
        "id": "1RLg8oail2y8"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "293/293 [==============================] - 12s 32ms/step - loss: 1.3074 - price_loss: 0.8949 - type_loss: 1.7199 - price_sparse_categorical_accuracy: 0.5207 - type_sparse_categorical_accuracy: 0.7504 - val_loss: 0.9709 - val_price_loss: 0.8403 - val_type_loss: 1.1016 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 2/20\n",
            "293/293 [==============================] - 8s 29ms/step - loss: 0.9512 - price_loss: 0.8327 - type_loss: 1.0696 - price_sparse_categorical_accuracy: 0.6248 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.9241 - val_price_loss: 0.8360 - val_type_loss: 1.0121 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 3/20\n",
            "293/293 [==============================] - 8s 29ms/step - loss: 0.9287 - price_loss: 0.8315 - type_loss: 1.0258 - price_sparse_categorical_accuracy: 0.6248 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.9159 - val_price_loss: 0.8363 - val_type_loss: 0.9955 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 4/20\n",
            "293/293 [==============================] - 9s 29ms/step - loss: 0.9216 - price_loss: 0.8304 - type_loss: 1.0128 - price_sparse_categorical_accuracy: 0.6248 - type_sparse_categorical_accuracy: 0.7545 - val_loss: 0.9141 - val_price_loss: 0.8398 - val_type_loss: 0.9885 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 5/20\n",
            "293/293 [==============================] - 8s 29ms/step - loss: 0.9085 - price_loss: 0.8205 - type_loss: 0.9966 - price_sparse_categorical_accuracy: 0.6263 - type_sparse_categorical_accuracy: 0.7545 - val_loss: 0.9012 - val_price_loss: 0.8252 - val_type_loss: 0.9772 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 6/20\n",
            "293/293 [==============================] - 9s 29ms/step - loss: 0.8641 - price_loss: 0.7790 - type_loss: 0.9492 - price_sparse_categorical_accuracy: 0.6329 - type_sparse_categorical_accuracy: 0.7549 - val_loss: 0.8946 - val_price_loss: 0.8122 - val_type_loss: 0.9771 - val_price_sparse_categorical_accuracy: 0.6254 - val_type_sparse_categorical_accuracy: 0.7637\n",
            "Epoch 7/20\n",
            "293/293 [==============================] - 9s 29ms/step - loss: 0.8021 - price_loss: 0.7097 - type_loss: 0.8946 - price_sparse_categorical_accuracy: 0.6822 - type_sparse_categorical_accuracy: 0.7545 - val_loss: 0.8982 - val_price_loss: 0.8147 - val_type_loss: 0.9818 - val_price_sparse_categorical_accuracy: 0.6186 - val_type_sparse_categorical_accuracy: 0.7628\n",
            "Epoch 8/20\n",
            "293/293 [==============================] - 8s 29ms/step - loss: 0.7376 - price_loss: 0.6306 - type_loss: 0.8446 - price_sparse_categorical_accuracy: 0.7425 - type_sparse_categorical_accuracy: 0.7570 - val_loss: 0.9036 - val_price_loss: 0.8186 - val_type_loss: 0.9887 - val_price_sparse_categorical_accuracy: 0.6075 - val_type_sparse_categorical_accuracy: 0.7619\n",
            "Epoch 9/20\n",
            "293/293 [==============================] - 9s 29ms/step - loss: 0.6766 - price_loss: 0.5473 - type_loss: 0.8060 - price_sparse_categorical_accuracy: 0.7869 - type_sparse_categorical_accuracy: 0.7575 - val_loss: 0.9258 - val_price_loss: 0.8392 - val_type_loss: 1.0123 - val_price_sparse_categorical_accuracy: 0.6297 - val_type_sparse_categorical_accuracy: 0.7568\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_val_acc_plt(history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "N8LgBokfl2y9",
        "outputId": "71f6465b-3efa-45b2-ccd3-bc22928ed4ec"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "train_val_loss_plt (history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "q1fAbbZPl2y9",
        "outputId": "85901bff-a847-40ea-cfcb-e6b64664abc2"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "print(model.evaluate(x= {\n",
        "    'summary': x_test_text},\n",
        "  y= {'price': df_test['price'],\n",
        "    'type':df_test['type_cat']}))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DR82Kw27l2y-",
        "outputId": "02f28d8e-c5d7-4ea7-e430-06608fdffc82"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "46/46 [==============================] - 0s 10ms/step - loss: 0.8956 - price_loss: 0.8117 - type_loss: 0.9794 - price_sparse_categorical_accuracy: 0.6207 - type_sparse_categorical_accuracy: 0.7674\n",
            "[0.8955575227737427, 0.811715841293335, 0.9793990850448608, 0.6207367181777954, 0.7673942446708679]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "the result is lower that before even worst that lstm : price_sparse_categorical_accuracy: 0.6207\n",
        "\n",
        "this means gru is not suitable for our data and we will use lstm or embedding only.\n",
        "\n",
        "for a last try let's add bigdirectional layer to lstm and gru"
      ],
      "metadata": {
        "id": "ltRxxreRl2y-"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Trial 4\n",
        "working with text input only (BiDirectional layer)\n",
        "\n",
        "using simple model with text as input and multi-task learning to predict (price and type)\n",
        "\n",
        "try BiDirectional with lstm and gru and compare the results"
      ],
      "metadata": {
        "id": "L9LnuixMyd2g"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "###Trial 4.1\n",
        "BiDirectional with gru"
      ],
      "metadata": {
        "id": "Yvj20mIEzhNm"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "keras.backend.clear_session()"
      ],
      "metadata": {
        "id": "00vD-SLvyd27"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e4a38ebc-c7c5-4aae-dfef-f546bda61a78",
        "id": "4rFDCR5Myd28"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_1 (InputLayer)           [(None, 100)]        0           []                               \n",
            "                                                                                                  \n",
            " embedding (Embedding)          (None, 100, 100)     4000000     ['input_1[0][0]']                \n",
            "                                                                                                  \n",
            " bidirectional (Bidirectional)  (None, 10)           3210        ['embedding[0][0]']              \n",
            "                                                                                                  \n",
            " price (Dense)                  (None, 3)            33          ['bidirectional[0][0]']          \n",
            "                                                                                                  \n",
            " type (Dense)                   (None, 24)           264         ['bidirectional[0][0]']          \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 4,003,507\n",
            "Trainable params: 4,003,507\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "# input for text.\n",
        "in_text = keras.Input(batch_shape=(None, max_len))\n",
        "\n",
        "# text part\n",
        "embedded = keras.layers.Embedding(tokenizer.num_words, 100)(in_text)\n",
        "GRU_layer_biD = keras.layers.Bidirectional(keras.layers.GRU(5 , dropout= 0.2))(embedded)\n",
        "\n",
        "\n",
        "# multi-task learning (each is a multi-class classification)\n",
        "# one dense layer for each task\n",
        "p_price = Dense(len_price, activation='softmax', name='price')(GRU_layer_biD)\n",
        "p_type = Dense(len_type, activation='softmax', name='type')(GRU_layer_biD)\n",
        "\n",
        "\n",
        "# define model input/output using keys.\n",
        "model = keras.Model(\n",
        "    inputs={\n",
        "        'summary': in_text,\n",
        "    },\n",
        "    outputs={\n",
        "        'price': p_price,\n",
        "        'type': p_type,\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "# compile model with optimizer, loss values for each task, loss \n",
        "# weights for each task.\n",
        "model.compile(\n",
        "    optimizer=Adam(),\n",
        "    loss={\n",
        "        'price': 'sparse_categorical_crossentropy',\n",
        "        'type': 'sparse_categorical_crossentropy',\n",
        "    },\n",
        "    loss_weights={\n",
        "        'price': 0.5,\n",
        "        'type': 0.5,       \n",
        "    },\n",
        "    metrics={\n",
        "        'price': ['SparseCategoricalAccuracy'],\n",
        "        'type': ['SparseCategoricalAccuracy'],\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(\n",
        "    x={\n",
        "        'summary': x_train_text_id,\n",
        "    },\n",
        "    y={\n",
        "        'price': y_train_price,\n",
        "        'type': y_train_type,\n",
        "    },\n",
        "    epochs=20,\n",
        "    batch_size=16,\n",
        "    validation_split=0.2,\n",
        "    callbacks=[\n",
        "        tf.keras.callbacks.EarlyStopping(monitor='val_price_loss', patience=3,\n",
        "                                         restore_best_weights=True)\n",
        "    ],\n",
        "    verbose=1\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8225a6c9-2da6-43dd-c57c-1bc8321c3919",
        "id": "4hWuPEDiyd2-"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "293/293 [==============================] - 21s 59ms/step - loss: 1.1710 - price_loss: 0.8534 - type_loss: 1.4886 - price_sparse_categorical_accuracy: 0.6165 - type_sparse_categorical_accuracy: 0.7370 - val_loss: 0.9009 - val_price_loss: 0.8156 - val_type_loss: 0.9863 - val_price_sparse_categorical_accuracy: 0.6135 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 2/20\n",
            "293/293 [==============================] - 17s 57ms/step - loss: 0.8749 - price_loss: 0.7887 - type_loss: 0.9612 - price_sparse_categorical_accuracy: 0.6241 - type_sparse_categorical_accuracy: 0.7570 - val_loss: 0.8539 - val_price_loss: 0.7884 - val_type_loss: 0.9193 - val_price_sparse_categorical_accuracy: 0.6143 - val_type_sparse_categorical_accuracy: 0.7705\n",
            "Epoch 3/20\n",
            "293/293 [==============================] - 17s 57ms/step - loss: 0.7980 - price_loss: 0.7363 - type_loss: 0.8598 - price_sparse_categorical_accuracy: 0.6529 - type_sparse_categorical_accuracy: 0.7698 - val_loss: 0.8345 - val_price_loss: 0.7699 - val_type_loss: 0.8992 - val_price_sparse_categorical_accuracy: 0.6374 - val_type_sparse_categorical_accuracy: 0.7696\n",
            "Epoch 4/20\n",
            "293/293 [==============================] - 17s 57ms/step - loss: 0.7295 - price_loss: 0.6773 - type_loss: 0.7817 - price_sparse_categorical_accuracy: 0.6973 - type_sparse_categorical_accuracy: 0.7803 - val_loss: 0.8185 - val_price_loss: 0.7661 - val_type_loss: 0.8708 - val_price_sparse_categorical_accuracy: 0.6348 - val_type_sparse_categorical_accuracy: 0.7713\n",
            "Epoch 5/20\n",
            "293/293 [==============================] - 16s 55ms/step - loss: 0.6652 - price_loss: 0.6178 - type_loss: 0.7126 - price_sparse_categorical_accuracy: 0.7325 - type_sparse_categorical_accuracy: 0.7882 - val_loss: 0.8387 - val_price_loss: 0.7865 - val_type_loss: 0.8909 - val_price_sparse_categorical_accuracy: 0.6416 - val_type_sparse_categorical_accuracy: 0.7654\n",
            "Epoch 6/20\n",
            "293/293 [==============================] - 17s 56ms/step - loss: 0.6097 - price_loss: 0.5605 - type_loss: 0.6589 - price_sparse_categorical_accuracy: 0.7688 - type_sparse_categorical_accuracy: 0.8010 - val_loss: 0.8512 - val_price_loss: 0.8095 - val_type_loss: 0.8929 - val_price_sparse_categorical_accuracy: 0.6280 - val_type_sparse_categorical_accuracy: 0.7628\n",
            "Epoch 7/20\n",
            "293/293 [==============================] - 17s 57ms/step - loss: 0.5579 - price_loss: 0.5073 - type_loss: 0.6086 - price_sparse_categorical_accuracy: 0.7982 - type_sparse_categorical_accuracy: 0.8131 - val_loss: 0.8873 - val_price_loss: 0.8405 - val_type_loss: 0.9341 - val_price_sparse_categorical_accuracy: 0.6476 - val_type_sparse_categorical_accuracy: 0.7491\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_val_acc_plt(history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "b2e43a6d-d5f1-4319-84b9-d2f4606a9381",
        "id": "cqjUzfXNyd2-"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "train_val_loss_plt (history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "2c03ef4a-7570-46b6-9f59-6a28bc3f2237",
        "id": "h4R9hCzOyd2_"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "print(model.evaluate(x= {\n",
        "    'summary': x_test_text},\n",
        "  y= {'price': df_test['price'],\n",
        "    'type':df_test['type_cat']}))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0686589c-79e4-43e0-fb52-d44ac5da2c9e",
        "id": "c_8BfQECyd3A"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "46/46 [==============================] - 1s 19ms/step - loss: 0.8338 - price_loss: 0.7918 - type_loss: 0.8757 - price_sparse_categorical_accuracy: 0.6180 - type_sparse_categorical_accuracy: 0.7797\n",
            "[0.8337778449058533, 0.7918071746826172, 0.8757489919662476, 0.6180081963539124, 0.7796725630760193]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "###Trial 4.2\n",
        "BiDirectional with lstm"
      ],
      "metadata": {
        "id": "wJEwVJIuzXV3"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "keras.backend.clear_session()"
      ],
      "metadata": {
        "id": "cW6-KK3XzCon"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "bf708e4f-ecc6-4d0f-80a1-52b57fd05e3c",
        "id": "5CznMnbGzCon"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_1 (InputLayer)           [(None, 100)]        0           []                               \n",
            "                                                                                                  \n",
            " embedding (Embedding)          (None, 100, 100)     4000000     ['input_1[0][0]']                \n",
            "                                                                                                  \n",
            " bidirectional (Bidirectional)  (None, 10)           4240        ['embedding[0][0]']              \n",
            "                                                                                                  \n",
            " price (Dense)                  (None, 3)            33          ['bidirectional[0][0]']          \n",
            "                                                                                                  \n",
            " type (Dense)                   (None, 24)           264         ['bidirectional[0][0]']          \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 4,004,537\n",
            "Trainable params: 4,004,537\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "# input for text.\n",
        "in_text = keras.Input(batch_shape=(None, max_len))\n",
        "\n",
        "# text part\n",
        "embedded = keras.layers.Embedding(tokenizer.num_words, 100)(in_text)\n",
        "LSTM_layer_biD = keras.layers.Bidirectional(keras.layers.LSTM(5 , dropout= 0.2))(embedded)\n",
        "\n",
        "\n",
        "# multi-task learning (each is a multi-class classification)\n",
        "# one dense layer for each task\n",
        "p_price = Dense(len_price, activation='softmax', name='price')(LSTM_layer_biD)\n",
        "p_type = Dense(len_type, activation='softmax', name='type')(LSTM_layer_biD)\n",
        "\n",
        "\n",
        "# define model input/output using keys.\n",
        "model = keras.Model(\n",
        "    inputs={\n",
        "        'summary': in_text,\n",
        "    },\n",
        "    outputs={\n",
        "        'price': p_price,\n",
        "        'type': p_type,\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "# compile model with optimizer, loss values for each task, loss \n",
        "# weights for each task.\n",
        "model.compile(\n",
        "    optimizer=Adam(),\n",
        "    loss={\n",
        "        'price': 'sparse_categorical_crossentropy',\n",
        "        'type': 'sparse_categorical_crossentropy',\n",
        "    },\n",
        "    loss_weights={\n",
        "        'price': 0.5,\n",
        "        'type': 0.5,       \n",
        "    },\n",
        "    metrics={\n",
        "        'price': ['SparseCategoricalAccuracy'],\n",
        "        'type': ['SparseCategoricalAccuracy'],\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(\n",
        "    x={\n",
        "        'summary': x_train_text_id,\n",
        "    },\n",
        "    y={\n",
        "        'price': y_train_price,\n",
        "        'type': y_train_type,\n",
        "    },\n",
        "    epochs=20,\n",
        "    batch_size=16,\n",
        "    validation_split=0.2,\n",
        "    callbacks=[\n",
        "        tf.keras.callbacks.EarlyStopping(monitor='val_price_loss', patience=3,\n",
        "                                         restore_best_weights=True)\n",
        "    ],\n",
        "    verbose=1\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e0e87582-64f1-4fdd-930f-ee48282fb8b1",
        "id": "ktfGmUM5zCoo"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "293/293 [==============================] - 23s 64ms/step - loss: 1.1933 - price_loss: 0.8641 - type_loss: 1.5226 - price_sparse_categorical_accuracy: 0.6165 - type_sparse_categorical_accuracy: 0.7483 - val_loss: 0.9151 - val_price_loss: 0.8194 - val_type_loss: 1.0107 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 2/20\n",
            "293/293 [==============================] - 16s 56ms/step - loss: 0.8771 - price_loss: 0.7783 - type_loss: 0.9760 - price_sparse_categorical_accuracy: 0.6357 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8553 - val_price_loss: 0.7776 - val_type_loss: 0.9330 - val_price_sparse_categorical_accuracy: 0.6510 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 3/20\n",
            "293/293 [==============================] - 16s 56ms/step - loss: 0.7841 - price_loss: 0.6976 - type_loss: 0.8707 - price_sparse_categorical_accuracy: 0.6977 - type_sparse_categorical_accuracy: 0.7541 - val_loss: 0.8356 - val_price_loss: 0.7660 - val_type_loss: 0.9052 - val_price_sparse_categorical_accuracy: 0.6613 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 4/20\n",
            "293/293 [==============================] - 16s 56ms/step - loss: 0.7020 - price_loss: 0.6151 - type_loss: 0.7890 - price_sparse_categorical_accuracy: 0.7568 - type_sparse_categorical_accuracy: 0.7607 - val_loss: 0.8481 - val_price_loss: 0.7837 - val_type_loss: 0.9125 - val_price_sparse_categorical_accuracy: 0.6442 - val_type_sparse_categorical_accuracy: 0.7688\n",
            "Epoch 5/20\n",
            "293/293 [==============================] - 17s 57ms/step - loss: 0.6239 - price_loss: 0.5222 - type_loss: 0.7255 - price_sparse_categorical_accuracy: 0.8014 - type_sparse_categorical_accuracy: 0.7824 - val_loss: 0.8815 - val_price_loss: 0.8367 - val_type_loss: 0.9262 - val_price_sparse_categorical_accuracy: 0.6297 - val_type_sparse_categorical_accuracy: 0.7705\n",
            "Epoch 6/20\n",
            "293/293 [==============================] - 16s 56ms/step - loss: 0.5547 - price_loss: 0.4488 - type_loss: 0.6605 - price_sparse_categorical_accuracy: 0.8398 - type_sparse_categorical_accuracy: 0.8108 - val_loss: 0.9189 - val_price_loss: 0.8882 - val_type_loss: 0.9496 - val_price_sparse_categorical_accuracy: 0.6305 - val_type_sparse_categorical_accuracy: 0.7662\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_val_acc_plt(history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "bc932239-5c77-400c-f02c-6a6f93914ca0",
        "id": "-yHd_CV1zCoo"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "train_val_loss_plt (history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "f71a2738-bfa0-4653-bfcd-6019adff445a",
        "id": "aODe9UaTzCop"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "print(model.evaluate(x= {\n",
        "    'summary': x_test_text},\n",
        "  y= {'price': df_test['price'],\n",
        "    'type':df_test['type_cat']}))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "18f5a316-6b4c-4f85-9fd0-111e060c85d2",
        "id": "XgfQLQO9zCop"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "46/46 [==============================] - 1s 17ms/step - loss: 0.8390 - price_loss: 0.7664 - type_loss: 0.9117 - price_sparse_categorical_accuracy: 0.6494 - type_sparse_categorical_accuracy: 0.7694\n",
            "[0.8390364050865173, 0.7663729190826416, 0.9116994738578796, 0.6493861079216003, 0.7694406509399414]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "for gru => price_sparse_categorical_accuracy: 0.6180\n",
        "\n",
        "for lstm => price_sparse_categorical_accuracy: 0.6494\n",
        "\n",
        "for lstm the results improved alot so i tried to submit on kaggel and got this score kaggel Score: 0.65380"
      ],
      "metadata": {
        "id": "TQakqBbaL_O3"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Trial 5\n",
        "working with imge input only\n",
        "\n",
        "no we will work on the image part (cnn)\n",
        "first let's start with simple cnn network with consists of one layer ofc conv2D followed by a maxpooling layer the a flatten layer to flatten the network output before passing it to the dense layers"
      ],
      "metadata": {
        "id": "cfa9L7tVqrNu"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "keras.backend.clear_session()"
      ],
      "metadata": {
        "id": "rVUOWeGaqrOG"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a38fcd79-1972-4243-e05c-bdae34920fa5",
        "id": "OipyGEx_qrOG"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model_5\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_6 (InputLayer)           [(None, 64, 64, 1)]  0           []                               \n",
            "                                                                                                  \n",
            " conv2d_4 (Conv2D)              (None, 62, 62, 32)   320         ['input_6[0][0]']                \n",
            "                                                                                                  \n",
            " max_pooling2d_4 (MaxPooling2D)  (None, 3, 3, 32)    0           ['conv2d_4[0][0]']               \n",
            "                                                                                                  \n",
            " flatten_4 (Flatten)            (None, 288)          0           ['max_pooling2d_4[0][0]']        \n",
            "                                                                                                  \n",
            " price (Dense)                  (None, 3)            867         ['flatten_4[0][0]']              \n",
            "                                                                                                  \n",
            " type (Dense)                   (None, 24)           6936        ['flatten_4[0][0]']              \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 8,123\n",
            "Trainable params: 8,123\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "# input for image\n",
        "in_image = keras.Input(batch_shape=(None, 64, 64,1))\n",
        "\n",
        "\n",
        "cov = Conv2D(32, (3, 3),activation='relu')(in_image)\n",
        "pl = MaxPool2D((16,16))(cov)\n",
        "flattened = Flatten()(pl)\n",
        "\n",
        "p_price = Dense(len_price, activation='softmax', name='price')(flattened)\n",
        "p_type = Dense(len_type, activation='softmax', name='type')(flattened)\n",
        "\n",
        "# define model input/output using keys.\n",
        "model = keras.Model(\n",
        "    inputs={\n",
        "        'image': in_image\n",
        "    },\n",
        "    outputs={\n",
        "        'price': p_price,\n",
        "        'type': p_type,\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "# compile model with optimizer, loss values for each task, loss \n",
        "# weights for each task.\n",
        "model.compile(\n",
        "    optimizer=Adam(),\n",
        "    loss={\n",
        "        'price': 'sparse_categorical_crossentropy',\n",
        "        'type': 'sparse_categorical_crossentropy',\n",
        "    },\n",
        "    loss_weights={\n",
        "        'price': 0.5,\n",
        "        'type': 0.5,       \n",
        "    },\n",
        "    metrics={\n",
        "        'price': ['SparseCategoricalAccuracy'],\n",
        "        'type': ['SparseCategoricalAccuracy'],\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(\n",
        "    x={\n",
        "        'image': x_train_image,\n",
        "    },\n",
        "    y={\n",
        "        'price': y_train_price,\n",
        "        'type': y_train_type,\n",
        "    },\n",
        "    epochs=20,\n",
        "    batch_size=16,\n",
        "    validation_split=0.2,\n",
        "    callbacks=[\n",
        "        tf.keras.callbacks.EarlyStopping(monitor='val_price_loss', patience=3,\n",
        "                                         restore_best_weights=True)\n",
        "    ],\n",
        "    verbose=1\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1f8d7980-11c7-40d7-840e-47c5d2d2b653",
        "id": "8Sw-fJoZqrOJ"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "293/293 [==============================] - 3s 8ms/step - loss: 4.8675 - price_loss: 4.2483 - type_loss: 5.4868 - price_sparse_categorical_accuracy: 0.4957 - type_sparse_categorical_accuracy: 0.5894 - val_loss: 1.7269 - val_price_loss: 1.5630 - val_type_loss: 1.8908 - val_price_sparse_categorical_accuracy: 0.5051 - val_type_sparse_categorical_accuracy: 0.6425\n",
            "Epoch 2/20\n",
            "293/293 [==============================] - 2s 9ms/step - loss: 1.3574 - price_loss: 1.2571 - type_loss: 1.4577 - price_sparse_categorical_accuracy: 0.5286 - type_sparse_categorical_accuracy: 0.6820 - val_loss: 1.2115 - val_price_loss: 1.1336 - val_type_loss: 1.2894 - val_price_sparse_categorical_accuracy: 0.6067 - val_type_sparse_categorical_accuracy: 0.7005\n",
            "Epoch 3/20\n",
            "293/293 [==============================] - 2s 8ms/step - loss: 1.0217 - price_loss: 0.9385 - type_loss: 1.1050 - price_sparse_categorical_accuracy: 0.5723 - type_sparse_categorical_accuracy: 0.7355 - val_loss: 1.0836 - val_price_loss: 0.9604 - val_type_loss: 1.2068 - val_price_sparse_categorical_accuracy: 0.5452 - val_type_sparse_categorical_accuracy: 0.7406\n",
            "Epoch 4/20\n",
            "293/293 [==============================] - 2s 7ms/step - loss: 0.9351 - price_loss: 0.8760 - type_loss: 0.9942 - price_sparse_categorical_accuracy: 0.5898 - type_sparse_categorical_accuracy: 0.7545 - val_loss: 1.0641 - val_price_loss: 0.9373 - val_type_loss: 1.1909 - val_price_sparse_categorical_accuracy: 0.5904 - val_type_sparse_categorical_accuracy: 0.7585\n",
            "Epoch 5/20\n",
            "293/293 [==============================] - 2s 7ms/step - loss: 0.8854 - price_loss: 0.8338 - type_loss: 0.9371 - price_sparse_categorical_accuracy: 0.6113 - type_sparse_categorical_accuracy: 0.7605 - val_loss: 1.0652 - val_price_loss: 0.9375 - val_type_loss: 1.1929 - val_price_sparse_categorical_accuracy: 0.5836 - val_type_sparse_categorical_accuracy: 0.7500\n",
            "Epoch 6/20\n",
            "293/293 [==============================] - 2s 7ms/step - loss: 0.8567 - price_loss: 0.8219 - type_loss: 0.8914 - price_sparse_categorical_accuracy: 0.6137 - type_sparse_categorical_accuracy: 0.7666 - val_loss: 1.0978 - val_price_loss: 0.9585 - val_type_loss: 1.2371 - val_price_sparse_categorical_accuracy: 0.6032 - val_type_sparse_categorical_accuracy: 0.7500\n",
            "Epoch 7/20\n",
            "293/293 [==============================] - 2s 7ms/step - loss: 0.8240 - price_loss: 0.8014 - type_loss: 0.8467 - price_sparse_categorical_accuracy: 0.6340 - type_sparse_categorical_accuracy: 0.7741 - val_loss: 1.1699 - val_price_loss: 1.0047 - val_type_loss: 1.3351 - val_price_sparse_categorical_accuracy: 0.5862 - val_type_sparse_categorical_accuracy: 0.7423\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_val_acc_plt(history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "r6HdmpfSqrOJ",
        "outputId": "cbf1c629-da29-4f13-cfb4-9222fc44e6c9"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "train_val_loss_plt (history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "7z5djqO2qrOJ",
        "outputId": "55084d1b-1091-41b2-d6c3-48949c71d082"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "print(model.evaluate(x= {\n",
        "    'image': x_test_image},\n",
        "  y= {'price': df_test['price'],\n",
        "    'type':df_test['type_cat']}))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GW1UAZ2KqrOK",
        "outputId": "69d6dcc2-73e1-499b-cb9e-190bad261808"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "46/46 [==============================] - 1s 13ms/step - loss: 1.1190 - price_loss: 0.9781 - type_loss: 1.2600 - price_sparse_categorical_accuracy: 0.5628 - type_sparse_categorical_accuracy: 0.7558\n",
            "[1.119046688079834, 0.9780516028404236, 1.260041356086731, 0.562755823135376, 0.7557981014251709]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "result is price_sparse_categorical_accuracy: 0.5628\n",
        "\n",
        "lower than all previous trials but not bad for simple network\n",
        "\n",
        "at the next trial we will try to improve the performance of this model"
      ],
      "metadata": {
        "id": "bk83Yt17qrOK"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Trial 6\n",
        "working with imge input only (add drop out)\n",
        "\n",
        "let's make the model more complex and add dropout layer\n",
        "\n",
        "we will use 2 layers on Conv2D instead of one followed by a drop out layer\n",
        "\n",
        "the results showed be imporved as there is no sign of overfitting from the previous graphs so increasing the model complexity should improve acc"
      ],
      "metadata": {
        "id": "QEOplGYnxWXM"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "keras.backend.clear_session()"
      ],
      "metadata": {
        "id": "GiV0wOlcxWXl"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "70e4e91d-1d48-476e-8710-d989ab24ef3c",
        "id": "ctogsLFvxWXm"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model_6\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_7 (InputLayer)           [(None, 64, 64, 1)]  0           []                               \n",
            "                                                                                                  \n",
            " conv2d_5 (Conv2D)              (None, 62, 62, 64)   640         ['input_7[0][0]']                \n",
            "                                                                                                  \n",
            " conv2d_6 (Conv2D)              (None, 60, 60, 32)   18464       ['conv2d_5[0][0]']               \n",
            "                                                                                                  \n",
            " spatial_dropout2d (SpatialDrop  (None, 60, 60, 32)  0           ['conv2d_6[0][0]']               \n",
            " out2D)                                                                                           \n",
            "                                                                                                  \n",
            " max_pooling2d_5 (MaxPooling2D)  (None, 3, 3, 32)    0           ['spatial_dropout2d[0][0]']      \n",
            "                                                                                                  \n",
            " flatten_5 (Flatten)            (None, 288)          0           ['max_pooling2d_5[0][0]']        \n",
            "                                                                                                  \n",
            " price (Dense)                  (None, 3)            867         ['flatten_5[0][0]']              \n",
            "                                                                                                  \n",
            " type (Dense)                   (None, 24)           6936        ['flatten_5[0][0]']              \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 26,907\n",
            "Trainable params: 26,907\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "# input for image\n",
        "in_image = keras.Input(batch_shape=(None, 64, 64,1))\n",
        "\n",
        "\n",
        "cov_1 = Conv2D(64, (3, 3),activation='relu')(in_image)\n",
        "cov_2 = Conv2D(32, (3, 3),activation='relu')(cov_1)\n",
        "droup_out_layer = keras.layers.SpatialDropout2D(0.5)(cov_2)\n",
        "pl = MaxPool2D((16,16))(droup_out_layer)\n",
        "\n",
        "flattened = Flatten()(pl)\n",
        "\n",
        "p_price = Dense(len_price, activation='softmax', name='price')(flattened)\n",
        "p_type = Dense(len_type, activation='softmax', name='type')(flattened)\n",
        "\n",
        "# define model input/output using keys.\n",
        "model = keras.Model(\n",
        "    inputs={\n",
        "        'image': in_image\n",
        "    },\n",
        "    outputs={\n",
        "        'price': p_price,\n",
        "        'type': p_type,\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "# compile model with optimizer, loss values for each task, loss \n",
        "# weights for each task.\n",
        "model.compile(\n",
        "    optimizer=Adam(),\n",
        "    loss={\n",
        "        'price': 'sparse_categorical_crossentropy',\n",
        "        'type': 'sparse_categorical_crossentropy',\n",
        "    },\n",
        "    loss_weights={\n",
        "        'price': 0.5,\n",
        "        'type': 0.5,       \n",
        "    },\n",
        "    metrics={\n",
        "        'price': ['SparseCategoricalAccuracy'],\n",
        "        'type': ['SparseCategoricalAccuracy'],\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(\n",
        "    x={\n",
        "        'image': x_train_image,\n",
        "    },\n",
        "    y={\n",
        "        'price': y_train_price,\n",
        "        'type': y_train_type,\n",
        "    },\n",
        "    epochs=20,\n",
        "    batch_size=16,\n",
        "    validation_split=0.2,\n",
        "    callbacks=[\n",
        "        tf.keras.callbacks.EarlyStopping(monitor='val_price_loss', patience=3,\n",
        "                                         restore_best_weights=True)\n",
        "    ],\n",
        "    verbose=1\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "fb8c537e-2a29-49e6-998f-627d331e2b62",
        "id": "932cq1cfxWXn"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "293/293 [==============================] - 6s 18ms/step - loss: 1.8208 - price_loss: 1.3532 - type_loss: 2.2883 - price_sparse_categorical_accuracy: 0.5230 - type_sparse_categorical_accuracy: 0.6570 - val_loss: 1.0587 - val_price_loss: 0.9005 - val_type_loss: 1.2168 - val_price_sparse_categorical_accuracy: 0.5580 - val_type_sparse_categorical_accuracy: 0.7602\n",
            "Epoch 2/20\n",
            "293/293 [==============================] - 5s 18ms/step - loss: 1.2457 - price_loss: 0.9489 - type_loss: 1.5426 - price_sparse_categorical_accuracy: 0.5651 - type_sparse_categorical_accuracy: 0.7376 - val_loss: 0.9585 - val_price_loss: 0.8705 - val_type_loss: 1.0466 - val_price_sparse_categorical_accuracy: 0.6049 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 3/20\n",
            "293/293 [==============================] - 5s 17ms/step - loss: 1.1673 - price_loss: 0.9222 - type_loss: 1.4123 - price_sparse_categorical_accuracy: 0.5779 - type_sparse_categorical_accuracy: 0.7479 - val_loss: 0.9564 - val_price_loss: 0.8732 - val_type_loss: 1.0397 - val_price_sparse_categorical_accuracy: 0.5700 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 4/20\n",
            "293/293 [==============================] - 5s 16ms/step - loss: 1.0853 - price_loss: 0.8788 - type_loss: 1.2918 - price_sparse_categorical_accuracy: 0.5990 - type_sparse_categorical_accuracy: 0.7526 - val_loss: 0.9400 - val_price_loss: 0.8555 - val_type_loss: 1.0244 - val_price_sparse_categorical_accuracy: 0.6015 - val_type_sparse_categorical_accuracy: 0.7637\n",
            "Epoch 5/20\n",
            "293/293 [==============================] - 5s 16ms/step - loss: 1.0580 - price_loss: 0.8692 - type_loss: 1.2468 - price_sparse_categorical_accuracy: 0.6139 - type_sparse_categorical_accuracy: 0.7517 - val_loss: 0.9398 - val_price_loss: 0.8478 - val_type_loss: 1.0317 - val_price_sparse_categorical_accuracy: 0.6092 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 6/20\n",
            "293/293 [==============================] - 5s 16ms/step - loss: 1.0052 - price_loss: 0.8533 - type_loss: 1.1571 - price_sparse_categorical_accuracy: 0.6160 - type_sparse_categorical_accuracy: 0.7538 - val_loss: 0.9333 - val_price_loss: 0.8445 - val_type_loss: 1.0222 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 7/20\n",
            "293/293 [==============================] - 5s 17ms/step - loss: 0.9808 - price_loss: 0.8455 - type_loss: 1.1161 - price_sparse_categorical_accuracy: 0.6216 - type_sparse_categorical_accuracy: 0.7536 - val_loss: 0.9296 - val_price_loss: 0.8479 - val_type_loss: 1.0113 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 8/20\n",
            "293/293 [==============================] - 5s 17ms/step - loss: 0.9743 - price_loss: 0.8402 - type_loss: 1.1085 - price_sparse_categorical_accuracy: 0.6203 - type_sparse_categorical_accuracy: 0.7545 - val_loss: 0.9272 - val_price_loss: 0.8452 - val_type_loss: 1.0092 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 9/20\n",
            "293/293 [==============================] - 5s 18ms/step - loss: 0.9539 - price_loss: 0.8364 - type_loss: 1.0714 - price_sparse_categorical_accuracy: 0.6241 - type_sparse_categorical_accuracy: 0.7547 - val_loss: 0.9240 - val_price_loss: 0.8457 - val_type_loss: 1.0023 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_val_acc_plt(history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "5b665722-6766-4d42-fb94-0e59cfadf2dc",
        "id": "Wqdnt29YxWXo"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "train_val_loss_plt (history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "05233c2d-8375-459b-fce4-520b96bad01b",
        "id": "5Nwlxi13xWXp"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "print(model.evaluate(x= {\n",
        "    'image': x_test_image},\n",
        "  y= {'price': df_test['price'],\n",
        "    'type':df_test['type_cat']}))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "97cf89f8-a90f-4b60-a08a-1eb248859f2b",
        "id": "ooFYawRMxWXq"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "46/46 [==============================] - 1s 14ms/step - loss: 0.9532 - price_loss: 0.8524 - type_loss: 1.0540 - price_sparse_categorical_accuracy: 0.6119 - type_sparse_categorical_accuracy: 0.7674\n",
            "[0.9531843662261963, 0.8523790240287781, 1.0539900064468384, 0.6118690371513367, 0.7673942446708679]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "results imporved price_sparse_categorical_accuracy: 0.6119\n",
        "\n",
        "now let's but all together"
      ],
      "metadata": {
        "id": "gpwA4ngmMqkW"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Trial 7\n",
        "putting all together (Multi-modality learning.)\n",
        "\n",
        "use both image and text data\n",
        "\n",
        "now we will add the two networks together (the cnn network with the image input and the rnn network with the text input) then we concat the output of both models then pass it to the output layer (2 dense layers).\n",
        "\n",
        "the accuracy showed improve as we compining the efforts of 2 model to do the same task."
      ],
      "metadata": {
        "id": "lizW93mZ0rys"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "keras.backend.clear_session()"
      ],
      "metadata": {
        "id": "FYrOyOuM0rzG"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "93fa3051-814d-4614-b37a-5ef89fc2065c",
        "id": "aJA_J3Fw0rzH"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model_1\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_4 (InputLayer)           [(None, 64, 64, 1)]  0           []                               \n",
            "                                                                                                  \n",
            " conv2d_2 (Conv2D)              (None, 62, 62, 64)   640         ['input_4[0][0]']                \n",
            "                                                                                                  \n",
            " conv2d_3 (Conv2D)              (None, 60, 60, 32)   18464       ['conv2d_2[0][0]']               \n",
            "                                                                                                  \n",
            " input_3 (InputLayer)           [(None, 100)]        0           []                               \n",
            "                                                                                                  \n",
            " spatial_dropout2d_1 (SpatialDr  (None, 60, 60, 32)  0           ['conv2d_3[0][0]']               \n",
            " opout2D)                                                                                         \n",
            "                                                                                                  \n",
            " embedding_1 (Embedding)        (None, 100, 100)     4000000     ['input_3[0][0]']                \n",
            "                                                                                                  \n",
            " max_pooling2d_1 (MaxPooling2D)  (None, 7, 7, 32)    0           ['spatial_dropout2d_1[0][0]']    \n",
            "                                                                                                  \n",
            " bidirectional_1 (Bidirectional  (None, 10)          4240        ['embedding_1[0][0]']            \n",
            " )                                                                                                \n",
            "                                                                                                  \n",
            " flatten_1 (Flatten)            (None, 1568)         0           ['max_pooling2d_1[0][0]']        \n",
            "                                                                                                  \n",
            " tf.concat_1 (TFOpLambda)       (None, 1578)         0           ['bidirectional_1[0][0]',        \n",
            "                                                                  'flatten_1[0][0]']              \n",
            "                                                                                                  \n",
            " price (Dense)                  (None, 3)            4737        ['tf.concat_1[0][0]']            \n",
            "                                                                                                  \n",
            " type (Dense)                   (None, 24)           37896       ['tf.concat_1[0][0]']            \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 4,065,977\n",
            "Trainable params: 4,065,977\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "# here we have two inputs. one for image and the other for text.\n",
        "in_text = keras.Input(batch_shape=(None, max_len))\n",
        "in_image = keras.Input(batch_shape=(None, 64, 64,1))\n",
        "\n",
        "#imge network\n",
        "cov_1 = Conv2D(64, (3, 3),activation='relu')(in_image)\n",
        "cov_2 = Conv2D(32, (3, 3),activation='relu')(cov_1)\n",
        "droup_out_layer = keras.layers.SpatialDropout2D(0.5)(cov_2)\n",
        "pl = MaxPool2D((8,8))(droup_out_layer)\n",
        "flattened = Flatten()(pl)\n",
        "\n",
        "#text network\n",
        "embedded = keras.layers.Embedding(tokenizer.num_words, 100)(in_text)\n",
        "LSTM_layer_biD = keras.layers.Bidirectional(keras.layers.LSTM(5 , dropout= 0.5))(embedded)\n",
        "\n",
        "fused = tf.concat([LSTM_layer_biD, flattened], axis=-1)\n",
        "\n",
        "\n",
        "p_price = Dense(len_price, activation='softmax', name='price')(fused)\n",
        "p_type = Dense(len_type, activation='softmax', name='type')(fused)\n",
        "\n",
        "# define model input/output using keys.\n",
        "model = keras.Model(\n",
        "    inputs={\n",
        "        'summary': in_text,\n",
        "        'image': in_image\n",
        "    },\n",
        "    outputs={\n",
        "        'price': p_price,\n",
        "        'type': p_type,\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "# compile model with optimizer, loss values for each task, loss \n",
        "# weights for each task.\n",
        "model.compile(\n",
        "    optimizer=Adam(),\n",
        "    loss={\n",
        "        'price': 'sparse_categorical_crossentropy',\n",
        "        'type': 'sparse_categorical_crossentropy',\n",
        "    },\n",
        "    loss_weights={\n",
        "        'price': 0.5,\n",
        "        'type': 0.5,       \n",
        "    },\n",
        "    metrics={\n",
        "        'price': ['SparseCategoricalAccuracy'],\n",
        "        'type': ['SparseCategoricalAccuracy'],\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(\n",
        "    x={\n",
        "        'summary': x_train_text_id,\n",
        "        'image': x_train_image\n",
        "    },\n",
        "    y={\n",
        "        'price': y_train_price,\n",
        "        'type': y_train_type,\n",
        "    },\n",
        "    epochs=20,\n",
        "    batch_size=16,\n",
        "    validation_split=0.2,\n",
        "    callbacks=[\n",
        "        tf.keras.callbacks.EarlyStopping(monitor='val_price_loss', patience=3,\n",
        "                                         restore_best_weights=True)\n",
        "    ],\n",
        "    verbose=1\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f4bd154e-b41f-4a86-d733-d4501f26901f",
        "id": "4K7qP3xp0rzI"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "293/293 [==============================] - 28s 81ms/step - loss: 1.8999 - price_loss: 1.3022 - type_loss: 2.4977 - price_sparse_categorical_accuracy: 0.5862 - type_sparse_categorical_accuracy: 0.7035 - val_loss: 0.9242 - val_price_loss: 0.8386 - val_type_loss: 1.0098 - val_price_sparse_categorical_accuracy: 0.6143 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 2/20\n",
            "293/293 [==============================] - 23s 77ms/step - loss: 0.9198 - price_loss: 0.8311 - type_loss: 1.0085 - price_sparse_categorical_accuracy: 0.6237 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.9018 - val_price_loss: 0.8359 - val_type_loss: 0.9677 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 3/20\n",
            "293/293 [==============================] - 22s 77ms/step - loss: 0.8766 - price_loss: 0.8151 - type_loss: 0.9381 - price_sparse_categorical_accuracy: 0.6248 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8701 - val_price_loss: 0.8179 - val_type_loss: 0.9223 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 4/20\n",
            "293/293 [==============================] - 23s 77ms/step - loss: 0.8207 - price_loss: 0.7910 - type_loss: 0.8503 - price_sparse_categorical_accuracy: 0.6261 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8645 - val_price_loss: 0.8220 - val_type_loss: 0.9070 - val_price_sparse_categorical_accuracy: 0.6169 - val_type_sparse_categorical_accuracy: 0.7654\n",
            "Epoch 5/20\n",
            "293/293 [==============================] - 23s 77ms/step - loss: 0.7728 - price_loss: 0.7619 - type_loss: 0.7836 - price_sparse_categorical_accuracy: 0.6438 - type_sparse_categorical_accuracy: 0.7615 - val_loss: 0.8648 - val_price_loss: 0.8226 - val_type_loss: 0.9069 - val_price_sparse_categorical_accuracy: 0.6186 - val_type_sparse_categorical_accuracy: 0.7688\n",
            "Epoch 6/20\n",
            "293/293 [==============================] - 23s 77ms/step - loss: 0.7286 - price_loss: 0.7296 - type_loss: 0.7277 - price_sparse_categorical_accuracy: 0.6689 - type_sparse_categorical_accuracy: 0.7737 - val_loss: 0.8673 - val_price_loss: 0.8114 - val_type_loss: 0.9232 - val_price_sparse_categorical_accuracy: 0.6109 - val_type_sparse_categorical_accuracy: 0.7654\n",
            "Epoch 7/20\n",
            "293/293 [==============================] - 23s 77ms/step - loss: 0.6780 - price_loss: 0.6774 - type_loss: 0.6785 - price_sparse_categorical_accuracy: 0.7133 - type_sparse_categorical_accuracy: 0.7959 - val_loss: 0.8767 - val_price_loss: 0.8061 - val_type_loss: 0.9474 - val_price_sparse_categorical_accuracy: 0.6297 - val_type_sparse_categorical_accuracy: 0.7602\n",
            "Epoch 8/20\n",
            "293/293 [==============================] - 23s 78ms/step - loss: 0.6276 - price_loss: 0.6192 - type_loss: 0.6360 - price_sparse_categorical_accuracy: 0.7619 - type_sparse_categorical_accuracy: 0.8140 - val_loss: 0.8799 - val_price_loss: 0.8045 - val_type_loss: 0.9553 - val_price_sparse_categorical_accuracy: 0.6365 - val_type_sparse_categorical_accuracy: 0.7594\n",
            "Epoch 9/20\n",
            "293/293 [==============================] - 23s 78ms/step - loss: 0.5784 - price_loss: 0.5597 - type_loss: 0.5971 - price_sparse_categorical_accuracy: 0.7892 - type_sparse_categorical_accuracy: 0.8272 - val_loss: 0.9065 - val_price_loss: 0.8296 - val_type_loss: 0.9834 - val_price_sparse_categorical_accuracy: 0.6596 - val_type_sparse_categorical_accuracy: 0.7568\n",
            "Epoch 10/20\n",
            "293/293 [==============================] - 23s 77ms/step - loss: 0.5406 - price_loss: 0.5226 - type_loss: 0.5586 - price_sparse_categorical_accuracy: 0.8104 - type_sparse_categorical_accuracy: 0.8372 - val_loss: 0.9155 - val_price_loss: 0.8266 - val_type_loss: 1.0045 - val_price_sparse_categorical_accuracy: 0.6485 - val_type_sparse_categorical_accuracy: 0.7534\n",
            "Epoch 11/20\n",
            "293/293 [==============================] - 23s 77ms/step - loss: 0.5014 - price_loss: 0.4745 - type_loss: 0.5284 - price_sparse_categorical_accuracy: 0.8317 - type_sparse_categorical_accuracy: 0.8445 - val_loss: 0.9440 - val_price_loss: 0.8624 - val_type_loss: 1.0256 - val_price_sparse_categorical_accuracy: 0.6553 - val_type_sparse_categorical_accuracy: 0.7483\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_val_acc_plt(history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "b9e18c3c-1eeb-44a1-fcc1-baa4b51a403f",
        "id": "KtTw6JRt0rzI"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "train_val_loss_plt (history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "c92fb4e1-c34d-4f3d-cc9b-d7290aebffb4",
        "id": "hpGAeaFD0rzI"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "print(model.evaluate(x= {\n",
        "        'summary': x_test_text,\n",
        "        'image': x_test_image\n",
        "    },\n",
        "    y= {'price': df_test['price'],\n",
        "        'type':df_test['type_cat']}))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1440c1c2-8cff-4f22-d8a2-919fb63f5bdf",
        "id": "1sSEvZlK0rzJ"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "46/46 [==============================] - 1s 28ms/step - loss: 0.8539 - price_loss: 0.7894 - type_loss: 0.9184 - price_sparse_categorical_accuracy: 0.6446 - type_sparse_categorical_accuracy: 0.7749\n",
            "[0.8538644909858704, 0.7893632650375366, 0.9183658361434937, 0.6446111798286438, 0.7748976945877075]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "reuslt not as i wished price_sparse_categorical_accuracy: 0.6446\n",
        "so now we will make anothor try but with different parameters.\n",
        "and we will make the model simpler as there is a big gap between the training loss and validation loss as we see in the second graph."
      ],
      "metadata": {
        "id": "RY1Zk9MHM40W"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Trial 8\n",
        "putting all together (Multi-modality learning.)\n",
        "\n",
        "this is a big gap between train and validation curve and that means over fitting so we will make our model simpler.\n",
        "\n",
        "for the rnn part we will use embedding layer as the first trial (which achieved the hightest acc up till now)"
      ],
      "metadata": {
        "id": "MNFH8-WG4PU3"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "keras.backend.clear_session()"
      ],
      "metadata": {
        "id": "Erf4I1rO4PVB"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e2877c7e-3678-4adb-e638-c6d0a98767bb",
        "id": "R3AVRTPA4PVC"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_2 (InputLayer)           [(None, 64, 64, 1)]  0           []                               \n",
            "                                                                                                  \n",
            " conv2d (Conv2D)                (None, 62, 62, 64)   640         ['input_2[0][0]']                \n",
            "                                                                                                  \n",
            " conv2d_1 (Conv2D)              (None, 60, 60, 32)   18464       ['conv2d[0][0]']                 \n",
            "                                                                                                  \n",
            " input_1 (InputLayer)           [(None, 100)]        0           []                               \n",
            "                                                                                                  \n",
            " spatial_dropout2d (SpatialDrop  (None, 60, 60, 32)  0           ['conv2d_1[0][0]']               \n",
            " out2D)                                                                                           \n",
            "                                                                                                  \n",
            " embedding (Embedding)          (None, 100, 100)     4000000     ['input_1[0][0]']                \n",
            "                                                                                                  \n",
            " max_pooling2d (MaxPooling2D)   (None, 7, 7, 32)     0           ['spatial_dropout2d[0][0]']      \n",
            "                                                                                                  \n",
            " tf.math.reduce_mean (TFOpLambd  (None, 100)         0           ['embedding[0][0]']              \n",
            " a)                                                                                               \n",
            "                                                                                                  \n",
            " flatten (Flatten)              (None, 1568)         0           ['max_pooling2d[0][0]']          \n",
            "                                                                                                  \n",
            " tf.concat (TFOpLambda)         (None, 1668)         0           ['tf.math.reduce_mean[0][0]',    \n",
            "                                                                  'flatten[0][0]']                \n",
            "                                                                                                  \n",
            " price (Dense)                  (None, 3)            5007        ['tf.concat[0][0]']              \n",
            "                                                                                                  \n",
            " type (Dense)                   (None, 24)           40056       ['tf.concat[0][0]']              \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 4,064,167\n",
            "Trainable params: 4,064,167\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "# here we have two inputs. one for image and the other for text.\n",
        "in_text = keras.Input(batch_shape=(None, max_len))\n",
        "in_image = keras.Input(batch_shape=(None, 64, 64,1))\n",
        "\n",
        "#imge network\n",
        "cov_1 = Conv2D(64, (3, 3),activation='relu')(in_image)\n",
        "cov_2 = Conv2D(32, (3, 3),activation='relu')(cov_1)\n",
        "droup_out_layer = keras.layers.SpatialDropout2D(0.5)(cov_2)\n",
        "pl = MaxPool2D((8,8))(droup_out_layer)\n",
        "flattened = Flatten()(pl)\n",
        "\n",
        "#text network\n",
        "embedded = keras.layers.Embedding(tokenizer.num_words, 100)(in_text)\n",
        "averaged = tf.reduce_mean(embedded, axis=1)\n",
        "\n",
        "fused = tf.concat([averaged, flattened], axis=-1)\n",
        "\n",
        "\n",
        "p_price = Dense(len_price, activation='softmax', name='price')(fused)\n",
        "p_type = Dense(len_type, activation='softmax', name='type')(fused)\n",
        "\n",
        "# define model input/output using keys.\n",
        "model = keras.Model(\n",
        "    inputs={\n",
        "        'summary': in_text,\n",
        "        'image': in_image\n",
        "    },\n",
        "    outputs={\n",
        "        'price': p_price,\n",
        "        'type': p_type,\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "# compile model with optimizer, loss values for each task, loss \n",
        "# weights for each task.\n",
        "model.compile(\n",
        "    optimizer=Adam(),\n",
        "    loss={\n",
        "        'price': 'sparse_categorical_crossentropy',\n",
        "        'type': 'sparse_categorical_crossentropy',\n",
        "    },\n",
        "    loss_weights={\n",
        "        'price': 0.5,\n",
        "        'type': 0.5,       \n",
        "    },\n",
        "    metrics={\n",
        "        'price': ['SparseCategoricalAccuracy'],\n",
        "        'type': ['SparseCategoricalAccuracy'],\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(\n",
        "    x={\n",
        "        'summary': x_train_text_id,\n",
        "        'image': x_train_image\n",
        "    },\n",
        "    y={\n",
        "        'price': y_train_price,\n",
        "        'type': y_train_type,\n",
        "    },\n",
        "    epochs=20,\n",
        "    batch_size=16,\n",
        "    validation_split=0.2,\n",
        "    callbacks=[\n",
        "        tf.keras.callbacks.EarlyStopping(monitor='val_price_loss', patience=3,\n",
        "                                         restore_best_weights=True)\n",
        "    ],\n",
        "    verbose=1\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "421f7224-ffd7-4385-adfa-6b6597f8645e",
        "id": "LipJLnXz4PVC"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "293/293 [==============================] - 8s 24ms/step - loss: 1.8903 - price_loss: 1.6215 - type_loss: 2.1592 - price_sparse_categorical_accuracy: 0.5644 - type_sparse_categorical_accuracy: 0.6653 - val_loss: 0.9032 - val_price_loss: 0.8322 - val_type_loss: 0.9743 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 2/20\n",
            "293/293 [==============================] - 6s 21ms/step - loss: 0.8974 - price_loss: 0.8172 - type_loss: 0.9776 - price_sparse_categorical_accuracy: 0.6241 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8868 - val_price_loss: 0.8227 - val_type_loss: 0.9508 - val_price_sparse_categorical_accuracy: 0.6152 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 3/20\n",
            "293/293 [==============================] - 6s 22ms/step - loss: 0.8730 - price_loss: 0.8011 - type_loss: 0.9450 - price_sparse_categorical_accuracy: 0.6252 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8767 - val_price_loss: 0.8119 - val_type_loss: 0.9415 - val_price_sparse_categorical_accuracy: 0.6212 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 4/20\n",
            "293/293 [==============================] - 6s 21ms/step - loss: 0.8501 - price_loss: 0.7811 - type_loss: 0.9190 - price_sparse_categorical_accuracy: 0.6333 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8653 - val_price_loss: 0.8034 - val_type_loss: 0.9273 - val_price_sparse_categorical_accuracy: 0.6254 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 5/20\n",
            "293/293 [==============================] - 6s 21ms/step - loss: 0.8236 - price_loss: 0.7541 - type_loss: 0.8931 - price_sparse_categorical_accuracy: 0.6542 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8503 - val_price_loss: 0.7844 - val_type_loss: 0.9161 - val_price_sparse_categorical_accuracy: 0.6425 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 6/20\n",
            "293/293 [==============================] - 6s 21ms/step - loss: 0.7940 - price_loss: 0.7210 - type_loss: 0.8669 - price_sparse_categorical_accuracy: 0.6713 - type_sparse_categorical_accuracy: 0.7549 - val_loss: 0.8342 - val_price_loss: 0.7671 - val_type_loss: 0.9014 - val_price_sparse_categorical_accuracy: 0.6536 - val_type_sparse_categorical_accuracy: 0.7654\n",
            "Epoch 7/20\n",
            "293/293 [==============================] - 6s 21ms/step - loss: 0.7603 - price_loss: 0.6838 - type_loss: 0.8369 - price_sparse_categorical_accuracy: 0.6922 - type_sparse_categorical_accuracy: 0.7560 - val_loss: 0.8224 - val_price_loss: 0.7573 - val_type_loss: 0.8876 - val_price_sparse_categorical_accuracy: 0.6621 - val_type_sparse_categorical_accuracy: 0.7654\n",
            "Epoch 8/20\n",
            "293/293 [==============================] - 6s 22ms/step - loss: 0.7301 - price_loss: 0.6514 - type_loss: 0.8088 - price_sparse_categorical_accuracy: 0.7137 - type_sparse_categorical_accuracy: 0.7590 - val_loss: 0.8163 - val_price_loss: 0.7497 - val_type_loss: 0.8828 - val_price_sparse_categorical_accuracy: 0.6672 - val_type_sparse_categorical_accuracy: 0.7696\n",
            "Epoch 9/20\n",
            "293/293 [==============================] - 6s 22ms/step - loss: 0.6972 - price_loss: 0.6185 - type_loss: 0.7759 - price_sparse_categorical_accuracy: 0.7349 - type_sparse_categorical_accuracy: 0.7622 - val_loss: 0.8044 - val_price_loss: 0.7427 - val_type_loss: 0.8661 - val_price_sparse_categorical_accuracy: 0.6724 - val_type_sparse_categorical_accuracy: 0.7730\n",
            "Epoch 10/20\n",
            "293/293 [==============================] - 6s 21ms/step - loss: 0.6687 - price_loss: 0.5923 - type_loss: 0.7452 - price_sparse_categorical_accuracy: 0.7549 - type_sparse_categorical_accuracy: 0.7675 - val_loss: 0.7995 - val_price_loss: 0.7420 - val_type_loss: 0.8570 - val_price_sparse_categorical_accuracy: 0.6732 - val_type_sparse_categorical_accuracy: 0.7730\n",
            "Epoch 11/20\n",
            "293/293 [==============================] - 6s 22ms/step - loss: 0.6380 - price_loss: 0.5631 - type_loss: 0.7128 - price_sparse_categorical_accuracy: 0.7662 - type_sparse_categorical_accuracy: 0.7767 - val_loss: 0.8022 - val_price_loss: 0.7525 - val_type_loss: 0.8519 - val_price_sparse_categorical_accuracy: 0.6706 - val_type_sparse_categorical_accuracy: 0.7713\n",
            "Epoch 12/20\n",
            "293/293 [==============================] - 6s 22ms/step - loss: 0.6134 - price_loss: 0.5406 - type_loss: 0.6862 - price_sparse_categorical_accuracy: 0.7863 - type_sparse_categorical_accuracy: 0.7839 - val_loss: 0.8207 - val_price_loss: 0.7725 - val_type_loss: 0.8689 - val_price_sparse_categorical_accuracy: 0.6758 - val_type_sparse_categorical_accuracy: 0.7756\n",
            "Epoch 13/20\n",
            "293/293 [==============================] - 6s 22ms/step - loss: 0.5823 - price_loss: 0.5135 - type_loss: 0.6511 - price_sparse_categorical_accuracy: 0.7946 - type_sparse_categorical_accuracy: 0.7944 - val_loss: 0.7957 - val_price_loss: 0.7514 - val_type_loss: 0.8399 - val_price_sparse_categorical_accuracy: 0.6732 - val_type_sparse_categorical_accuracy: 0.7747\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_val_acc_plt(history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "583f2cf5-c0c2-4e5d-d47e-86ad2874421c",
        "id": "xAcPedNT4PVC"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "train_val_loss_plt (history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "cd9dd50e-2e03-41c6-dbf0-6bf554825b0b",
        "id": "Xzml_T434PVD"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "print(model.evaluate(x= {\n",
        "        'summary': x_test_text,\n",
        "        'image': x_test_image\n",
        "    },\n",
        "    y= {'price': df_test['price'],\n",
        "        'type':df_test['type_cat']}))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "11fa0e7a-123e-408f-ee68-0b65ea234f06",
        "id": "Sr9icKSh4PVD"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "46/46 [==============================] - 1s 13ms/step - loss: 0.7909 - price_loss: 0.7217 - type_loss: 0.8601 - price_sparse_categorical_accuracy: 0.6698 - type_sparse_categorical_accuracy: 0.7708\n",
            "[0.790868878364563, 0.7216728329658508, 0.8600652813911438, 0.6698499321937561, 0.7708048820495605]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "results improved alot price_sparse_categorical_accuracy: 0.6698\n",
        "\n",
        "\n",
        "I tried to submit on kaggel and got this socre\n",
        "kaggel score:  0.67635\n",
        "\n",
        "the best score so far\n",
        "\n",
        "now for the cnn, we will try to improve this part by using a pre-traind model (transfer learning) to extract the features from the images then feed then to the output layer."
      ],
      "metadata": {
        "id": "4t4sWhOB5J3z"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Trial 9\n",
        "putting all together (using transfer learning)\n",
        "\n",
        "Transfer learning consists of taking features learned on one problem, and leveraging them on a new, similar problem.\n",
        "\n",
        "Now let's built the model.\n",
        "\n",
        "* we will take layers from a previously trained model which is VGG16\n",
        "* Freeze them, so as to avoid destroying any of the information they contain during future training rounds.\n",
        "* add some new, trainable layers on top of the frozen layers. They will learn to turn the old features into predictions on a new dataset.\n",
        "\n",
        "\n",
        "We add a Rescaling layer to scale input values (initially in the [0, 255] range) to the [-1, 1] range.\n",
        "We add a Dropout layer before the classification layer, for regularization.\n",
        "We make sure to pass training=False when calling the base model, so that it runs in inference mode, so that batchnorm statistics don't get updated even after we unfreeze the base model for fine-tuning.\n",
        "\n",
        "ref: https://keras.io/guides/transfer_learning/#an-endtoend-example-finetuning-an-image-classification-model-on-a-cats-vs-dogs-dataset"
      ],
      "metadata": {
        "id": "XcjLSGnlCLND"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "note that we need to reprocess our input images to fit the new model\n",
        "\n",
        "VGG16 requires the input image to have 3 channels so we will implemant new function to make use of the rgb channels of the images instead turning them to grayscale images then increasing the dimension."
      ],
      "metadata": {
        "id": "liHqj7Y1JBwk"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def load_image_RGB(file):\n",
        "    try:\n",
        "        image = Image.open(\n",
        "            file\n",
        "        ).convert('RGB').resize((128, 128))\n",
        "        arr = np.array(image)\n",
        "    except:\n",
        "        arr = np.zeros((128, 128,3))\n",
        "    return arr\n",
        "\n",
        "\n",
        "# loading images:\n",
        "x_train_image = np.array([load_image_RGB(i) for i in tqdm(df_train.image)])\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 49,
          "referenced_widgets": [
            "786f637c189d457aaaa1f54b88ca1cff",
            "35244a1845b6499e989864e0ebc6238d",
            "6b6f92e7daed47c5b6dae08f54b54ea9",
            "021d31b7f7dc4f798c317109a79c142b",
            "1faad506bbf84271ae7669c652edfdc0",
            "c9e35f22cd9e4561a14daaed3abca124",
            "954e6968261b4d1a9f9bbea55a0c242d",
            "86479a7791b34265a2f7dcc68d20d8bd",
            "642a5ac2f4b2427a89ef498643392651",
            "f2e003882f804290acac27ef5439f4cc",
            "7abd1b443f3f4fdd9a2b589244125b9c"
          ]
        },
        "id": "eAFB3Q-6DLFw",
        "outputId": "80d954ed-1cb2-4a7c-e861-e4174e8c6609"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/5860 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "786f637c189d457aaaa1f54b88ca1cff"
            }
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "keras.backend.clear_session()"
      ],
      "metadata": {
        "id": "xm8QOpWnCLNe"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "from keras.applications.vgg16 import VGG16\n",
        "base_model = VGG16(include_top=False, #drop the 3 fully-connected layers at the top of the network.\n",
        "                   weights=\"imagenet\",  # Load weights pre-trained on ImageNet.\n",
        "                   input_shape=(128,128,3)) # the input shape of images\n",
        "base_model.trainable = False # freeze the model"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xJbe3o9dChX-",
        "outputId": "308fdb0e-4d77-4297-f538-69658027968e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "58892288/58889256 [==============================] - 1s 0us/step\n",
            "58900480/58889256 [==============================] - 1s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "fcd95b78-eb93-482e-eb65-022b3ad0b527",
        "id": "JT4yNn7zCLNf"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model_1\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_5 (InputLayer)           [(None, 128, 128, 3  0           []                               \n",
            "                                )]                                                                \n",
            "                                                                                                  \n",
            " rescaling (Rescaling)          (None, 128, 128, 3)  0           ['input_5[0][0]']                \n",
            "                                                                                                  \n",
            " input_4 (InputLayer)           [(None, 100)]        0           []                               \n",
            "                                                                                                  \n",
            " vgg16 (Functional)             (None, 4, 4, 512)    14714688    ['rescaling[0][0]']              \n",
            "                                                                                                  \n",
            " embedding_1 (Embedding)        (None, 100, 100)     4000000     ['input_4[0][0]']                \n",
            "                                                                                                  \n",
            " global_average_pooling2d (Glob  (None, 512)         0           ['vgg16[0][0]']                  \n",
            " alAveragePooling2D)                                                                              \n",
            "                                                                                                  \n",
            " tf.math.reduce_mean_1 (TFOpLam  (None, 100)         0           ['embedding_1[0][0]']            \n",
            " bda)                                                                                             \n",
            "                                                                                                  \n",
            " dropout (Dropout)              (None, 512)          0           ['global_average_pooling2d[0][0]'\n",
            "                                                                 ]                                \n",
            "                                                                                                  \n",
            " tf.concat_1 (TFOpLambda)       (None, 612)          0           ['tf.math.reduce_mean_1[0][0]',  \n",
            "                                                                  'dropout[0][0]']                \n",
            "                                                                                                  \n",
            " price (Dense)                  (None, 3)            1839        ['tf.concat_1[0][0]']            \n",
            "                                                                                                  \n",
            " type (Dense)                   (None, 24)           14712       ['tf.concat_1[0][0]']            \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 18,731,239\n",
            "Trainable params: 4,016,551\n",
            "Non-trainable params: 14,714,688\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "# here we have two inputs. one for image and the other for text.\n",
        "in_text = keras.Input(batch_shape=(None, max_len))\n",
        "in_image = keras.Input(batch_shape=(None, 128, 128,3))\n",
        "\n",
        "#image network\n",
        "scale_layer = keras.layers.Rescaling(scale=1 / 127.5, offset=-1)(in_image)\n",
        "vgg_model = base_model(scale_layer, training=False)\n",
        "pooling = keras.layers.GlobalAveragePooling2D()(vgg_model)\n",
        "drop_layer = keras.layers.Dropout(0.2)(pooling)\n",
        "\n",
        "#text network\n",
        "embedded = keras.layers.Embedding(tokenizer.num_words, 100)(in_text)\n",
        "averaged = tf.reduce_mean(embedded, axis=1)\n",
        "\n",
        "fused = tf.concat([averaged, drop_layer], axis=-1)\n",
        "\n",
        "\n",
        "p_price = Dense(len_price, activation='softmax', name='price')(fused)\n",
        "p_type = Dense(len_type, activation='softmax', name='type')(fused)\n",
        "\n",
        "# define model input/output using keys.\n",
        "model = keras.Model(\n",
        "    inputs={\n",
        "        'summary': in_text,\n",
        "        'image': in_image\n",
        "    },\n",
        "    outputs={\n",
        "        'price': p_price,\n",
        "        'type': p_type,\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "# compile model with optimizer, loss values for each task, loss \n",
        "# weights for each task.\n",
        "model.compile(\n",
        "    optimizer=Adam(),\n",
        "    loss={\n",
        "        'price': 'sparse_categorical_crossentropy',\n",
        "        'type': 'sparse_categorical_crossentropy',\n",
        "    },\n",
        "    loss_weights={\n",
        "        'price': 0.5,\n",
        "        'type': 0.5,       \n",
        "    },\n",
        "    metrics={\n",
        "        'price': ['SparseCategoricalAccuracy'],\n",
        "        'type': ['SparseCategoricalAccuracy'],\n",
        "    },\n",
        ")\n",
        "\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(\n",
        "    x={\n",
        "        'summary': x_train_text_id,\n",
        "        'image': x_train_image\n",
        "    },\n",
        "    y={\n",
        "        'price': y_train_price,\n",
        "        'type': y_train_type,\n",
        "    },\n",
        "    epochs=20,\n",
        "    batch_size=16,\n",
        "    validation_split=0.2,\n",
        "    callbacks=[\n",
        "        tf.keras.callbacks.EarlyStopping(monitor='val_price_loss', patience=3,\n",
        "                                         restore_best_weights=True)\n",
        "    ],\n",
        "    verbose=1\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e3b9fc92-9f04-44ef-bbd8-3aa2dc981e0a",
        "id": "iie6NWEtCLNg"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "293/293 [==============================] - 26s 82ms/step - loss: 1.0079 - price_loss: 0.8552 - type_loss: 1.1606 - price_sparse_categorical_accuracy: 0.6081 - type_sparse_categorical_accuracy: 0.7259 - val_loss: 0.8967 - val_price_loss: 0.8113 - val_type_loss: 0.9821 - val_price_sparse_categorical_accuracy: 0.6331 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 2/20\n",
            "293/293 [==============================] - 23s 77ms/step - loss: 0.8738 - price_loss: 0.7756 - type_loss: 0.9721 - price_sparse_categorical_accuracy: 0.6555 - type_sparse_categorical_accuracy: 0.7543 - val_loss: 0.8701 - val_price_loss: 0.7770 - val_type_loss: 0.9632 - val_price_sparse_categorical_accuracy: 0.6502 - val_type_sparse_categorical_accuracy: 0.7645\n",
            "Epoch 3/20\n",
            "293/293 [==============================] - 24s 82ms/step - loss: 0.8208 - price_loss: 0.7312 - type_loss: 0.9103 - price_sparse_categorical_accuracy: 0.6730 - type_sparse_categorical_accuracy: 0.7547 - val_loss: 0.8342 - val_price_loss: 0.7526 - val_type_loss: 0.9157 - val_price_sparse_categorical_accuracy: 0.6621 - val_type_sparse_categorical_accuracy: 0.7671\n",
            "Epoch 4/20\n",
            "293/293 [==============================] - 24s 80ms/step - loss: 0.7608 - price_loss: 0.6760 - type_loss: 0.8456 - price_sparse_categorical_accuracy: 0.7031 - type_sparse_categorical_accuracy: 0.7592 - val_loss: 0.8136 - val_price_loss: 0.7404 - val_type_loss: 0.8869 - val_price_sparse_categorical_accuracy: 0.6689 - val_type_sparse_categorical_accuracy: 0.7696\n",
            "Epoch 5/20\n",
            "293/293 [==============================] - 23s 79ms/step - loss: 0.7017 - price_loss: 0.6243 - type_loss: 0.7792 - price_sparse_categorical_accuracy: 0.7304 - type_sparse_categorical_accuracy: 0.7666 - val_loss: 0.7949 - val_price_loss: 0.7264 - val_type_loss: 0.8634 - val_price_sparse_categorical_accuracy: 0.6809 - val_type_sparse_categorical_accuracy: 0.7696\n",
            "Epoch 6/20\n",
            "293/293 [==============================] - 23s 77ms/step - loss: 0.6530 - price_loss: 0.5835 - type_loss: 0.7225 - price_sparse_categorical_accuracy: 0.7551 - type_sparse_categorical_accuracy: 0.7760 - val_loss: 0.8001 - val_price_loss: 0.7382 - val_type_loss: 0.8620 - val_price_sparse_categorical_accuracy: 0.6706 - val_type_sparse_categorical_accuracy: 0.7790\n",
            "Epoch 7/20\n",
            "293/293 [==============================] - 23s 79ms/step - loss: 0.6071 - price_loss: 0.5451 - type_loss: 0.6690 - price_sparse_categorical_accuracy: 0.7769 - type_sparse_categorical_accuracy: 0.7944 - val_loss: 0.7804 - val_price_loss: 0.7240 - val_type_loss: 0.8368 - val_price_sparse_categorical_accuracy: 0.6732 - val_type_sparse_categorical_accuracy: 0.7790\n",
            "Epoch 8/20\n",
            "293/293 [==============================] - 23s 77ms/step - loss: 0.5665 - price_loss: 0.5122 - type_loss: 0.6208 - price_sparse_categorical_accuracy: 0.7988 - type_sparse_categorical_accuracy: 0.8078 - val_loss: 0.7843 - val_price_loss: 0.7291 - val_type_loss: 0.8395 - val_price_sparse_categorical_accuracy: 0.6715 - val_type_sparse_categorical_accuracy: 0.7756\n",
            "Epoch 9/20\n",
            "293/293 [==============================] - 23s 78ms/step - loss: 0.5259 - price_loss: 0.4825 - type_loss: 0.5693 - price_sparse_categorical_accuracy: 0.8042 - type_sparse_categorical_accuracy: 0.8244 - val_loss: 0.7833 - val_price_loss: 0.7420 - val_type_loss: 0.8245 - val_price_sparse_categorical_accuracy: 0.6655 - val_type_sparse_categorical_accuracy: 0.7841\n",
            "Epoch 10/20\n",
            "293/293 [==============================] - 23s 79ms/step - loss: 0.4899 - price_loss: 0.4528 - type_loss: 0.5270 - price_sparse_categorical_accuracy: 0.8178 - type_sparse_categorical_accuracy: 0.8377 - val_loss: 0.7814 - val_price_loss: 0.7441 - val_type_loss: 0.8186 - val_price_sparse_categorical_accuracy: 0.6749 - val_type_sparse_categorical_accuracy: 0.7824\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_val_acc_plt(history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "de07a005-39f8-4a1d-85bd-a6de0e4421fe",
        "id": "Zgsja1nDCLNh"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "train_val_loss_plt (history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "aaeaaa90-1256-4bd2-bf34-6cc648db9b3b",
        "id": "1zcshTMuCLNi"
      },
      "execution_count": null,
      "outputs": [
        {
          "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": [
        "x_test_image = np.array([load_image_RGB(i) for i in tqdm(df_test.image)])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 49,
          "referenced_widgets": [
            "03c5635e6de2450dbd9eea85d2e0d9fb",
            "9db137265a0c487a935c2969a44558d0",
            "ef212e2db02c4500aac47ecf8b770981",
            "feaddee22c7e4742a880d808c1504c52",
            "78ccfb12800b46c090fb98f9f19f2d9f",
            "e1d54376d8734b64b814806e9fef5add",
            "e76d113ae5904a04b565035a1fadfc88",
            "cb8db989e41345ea9f3ff017bdbd6ca2",
            "05f06f0b7a6c4abba0537eac94652aaa",
            "f54122d9b8df4735ab01b24745338daf",
            "258d76ec903d41ce85ed7cd695d78703"
          ]
        },
        "id": "3MJOv0bpGN1e",
        "outputId": "45df6289-795a-40fb-fb03-308ea6301f55"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/1466 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "03c5635e6de2450dbd9eea85d2e0d9fb"
            }
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(model.evaluate(x= {\n",
        "        'summary': x_test_text,\n",
        "        'image': x_test_image\n",
        "    },\n",
        "    y= {'price': df_test['price'],\n",
        "        'type':df_test['type_cat']}))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7efb91f3-30d9-4013-8394-1c3db638da5a",
        "id": "1JPszKxACLNj"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "46/46 [==============================] - 8s 134ms/step - loss: 0.7785 - price_loss: 0.7091 - type_loss: 0.8478 - price_sparse_categorical_accuracy: 0.6808 - type_sparse_categorical_accuracy: 0.7769\n",
            "[0.7784526348114014, 0.7091405987739563, 0.8477648496627808, 0.6807639598846436, 0.7769440412521362]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "results are ammazing price_sparse_categorical_accuracy: 0.6808\n",
        "\n",
        "Score on kaggel: 0.69375 unfortunately at this time the comptition closed so this is a late submission :( .\n",
        "\n",
        "with more tuning we can achieve higher score.\n",
        "\n",
        "for example: we can fine-tuning the VGG model, which consists of unfreezing the entire model, and re-training it on the new data with a very low learning rate. This can potentially achieve meaningful improvements, by incrementally adapting the pretrained features to the new data."
      ],
      "metadata": {
        "id": "u-HsYHDWCLNj"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Testing\n",
        "for testing purpose only"
      ],
      "metadata": {
        "id": "NHySoDPm34VX"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# loading images:\n",
        "x_test_image = np.array([load_image(i) for i in tqdm(df_test.image)])\n",
        "\n",
        "# loading overview: (force convert some of the non-string cell to string)\n",
        "x_test_text = _preprocess(df_test.summary.astype('str'))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 49,
          "referenced_widgets": [
            "be3d61d8fe314e81ae223d0e75158178",
            "c6a8b9ce9e5f47fcb5d204d11725b509",
            "a10dc24e06c2488ea2d0fbee58246b7b",
            "472ccd92d91f475595b2f424d859d951",
            "1207fc6f1f59478fbe4014cc758d1bb1",
            "a6d9fadd10d94e83a1d374f1f38eb42c",
            "7255f4c4a0554272bb37f38a55cb0da2",
            "1bc074406aa947758cfb3686286b1b6f",
            "6bbba4cf396f4d6581e32f8db511c6d4",
            "e761f4aba4d7443a986cd62e3c93b4b3",
            "507b5beafe374b7f92759edccec7581d"
          ]
        },
        "id": "PPHkUWH_uSm4",
        "outputId": "8a4d2254-192e-4e36-e0e2-6beea44ab8ef"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/1466 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "be3d61d8fe314e81ae223d0e75158178"
            }
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model.evaluate(x= {\n",
        "        'summary': x_test_text,\n",
        "        'image': x_test_image\n",
        "    },\n",
        "    y= {'price': df_test['price'],\n",
        "        'type':df_test['type_cat']})"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "71SOew4K14lF",
        "outputId": "1e35c9d2-a201-4cfd-fb88-e98e3cc62d55"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow/python/data/ops/structured_function.py:265: UserWarning: Even though the `tf.config.experimental_run_functions_eagerly` option is set, this option does not apply to tf.data functions. To force eager execution of tf.data functions, please use `tf.data.experimental.enable_debug_mode()`.\n",
            "  \"Even though the `tf.config.experimental_run_functions_eagerly` \"\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "46/46 [==============================] - 8s 166ms/step - loss: 0.8459 - price_loss: 0.7581 - type_loss: 0.9199 - price_sparse_categorical_accuracy: 0.6467 - type_sparse_categorical_accuracy: 0.7674\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.8458641767501831,\n",
              " 0.7580727338790894,\n",
              " 0.9199457168579102,\n",
              " 0.6466575860977173,\n",
              " 0.7673942446708679]"
            ]
          },
          "metadata": {},
          "execution_count": 20
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# we can do prediction on training set\n",
        "y_predict = model.predict(\n",
        "    {\n",
        "        'summary': x_test_text,\n",
        "        'image': x_test_image\n",
        "    }\n",
        ")\n",
        "\n",
        "\n",
        "# probabilities\n",
        "price_predicted = y_predict['price']\n",
        "print(price_predicted)\n",
        "\n",
        "# categories\n",
        "price_category_predicted = np.argmax(price_predicted, axis=1)\n",
        "print(price_category_predicted)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QBWffa0mz0kx",
        "outputId": "3c8b37ac-6f5b-4206-f470-4db4d213cbda"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow/python/data/ops/structured_function.py:265: UserWarning: Even though the `tf.config.experimental_run_functions_eagerly` option is set, this option does not apply to tf.data functions. To force eager execution of tf.data functions, please use `tf.data.experimental.enable_debug_mode()`.\n",
            "  \"Even though the `tf.config.experimental_run_functions_eagerly` \"\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[[0.7564619  0.19540115 0.04813695]\n",
            " [0.86691064 0.10713629 0.025953  ]\n",
            " [0.60202014 0.35311493 0.04486489]\n",
            " ...\n",
            " [0.36335424 0.5999236  0.03672213]\n",
            " [0.819257   0.14730147 0.03344148]\n",
            " [0.8195616  0.13328275 0.0471557 ]]\n",
            "[0 0 0 ... 1 0 0]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "u , counts = np.unique(price_category_predicted, return_counts=True)\n",
        "np.asarray((u, counts)).T\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "08sRfeV40VGK",
        "outputId": "6ae7a0f3-c6c8-4e4b-8c6e-8b17e565b4ac"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[  0, 920],\n",
              "       [  1, 546]])"
            ]
          },
          "metadata": {},
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Test set for kaggel\n"
      ],
      "metadata": {
        "id": "1ebZB2Xd1DTe"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "df_test_kaggel = pd.read_csv(\"test_x.csv\")\n",
        "# loading images:\n",
        "x_train_image = np.array([load_image_RGB(i) for i in tqdm(df_test_kaggel.image)])\n",
        "\n",
        "# loading overview: (force convert some of the non-string cell to string)\n",
        "x_test_text_kaggel = _preprocess(df_test_kaggel.summary.astype('str'))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 49,
          "referenced_widgets": [
            "8232384db7c749789535d33d54590666",
            "3f46695168144e3da72a9f524d5e1950",
            "33f53533420c40f1b70e03992a69b459",
            "649131dc159f47069eb061e4dc009445",
            "f149f3c00b0d4999a254c0dc99757191",
            "fe56a1e7d8d64a3d9f9521729810d7e1",
            "da88b3466a3b4ab4b6763e336886aae0",
            "b1082cb5983141da877f56673e442a5f",
            "c38068e182eb46639b8cd8f57bc25ccd",
            "7f02f1d13dc24d4280439847a917b8cc",
            "8a1543e994d54748bfb148719aa6f229"
          ]
        },
        "id": "RX_v3sk_0qHZ",
        "outputId": "7b871416-9511-46d9-d635-abb1cfbe6b22"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/7360 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "8232384db7c749789535d33d54590666"
            }
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# we can do prediction on training set\n",
        "y_predict_kaggel  = model.predict(\n",
        "    {\n",
        "        'summary': x_test_text_kaggel,\n",
        "        'image': x_train_image\n",
        "    }\n",
        ")\n",
        "\n",
        "\n",
        "# probabilities\n",
        "price_predicted_kaggel = y_predict_kaggel['price']\n",
        "# print(price_predicted)\n",
        "\n",
        "# categories\n",
        "price_category_predicted_kaggel  = np.argmax(price_predicted_kaggel, axis=1)\n",
        "print(price_category_predicted_kaggel)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GMaM-yva_NO9",
        "outputId": "c0f53a41-892c-4d69-e74b-59bec68fae4f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[0 0 0 ... 0 0 0]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "u , counts = np.unique(price_category_predicted_kaggel, return_counts=True)\n",
        "np.asarray((u, counts)).T"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_53qRuB9_-U6",
        "outputId": "c507ad23-c117-479e-ac07-6fb4290c453f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[   0, 5355],\n",
              "       [   1, 1989],\n",
              "       [   2,   16]])"
            ]
          },
          "metadata": {},
          "execution_count": 178
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "pd.DataFrame(\n",
        "    {'id': df_test_kaggel.id,\n",
        "     'price': price_category_predicted_kaggel}\n",
        ").to_csv('submission.csv', index=False)"
      ],
      "metadata": {
        "id": "RjSJb0LM-6le"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "!kaggle competitions submit -c cisc-873-dm-f22-a4 -f submission.csv -m \"\""
      ],
      "metadata": {
        "id": "d-kQ6z55-ulI"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "e3TIGmm2Tmag"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Questions"
      ],
      "metadata": {
        "id": "SkvdbctZm4gH"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Is fully-connected model a good one for sequential data? Why? How about for image data? Is it good? Why?\n",
        "\n",
        "* yes \n",
        "* because Fully recurrent neural networks (FRNN) connect the outputs of all neurons to the inputs of all neurons. This is the most general neural network topology because all other topologies can be represented by setting some connection weights to zero to simulate the lack of connections between those neurons.\n",
        "* not good for image data\n",
        "* why? Because fully connected neural networks are not shift invariant if the object\n",
        "that we want to detect in the image moved to a new location (ex from the\n",
        "top left corner to the bottom right one) then the model will fail to detect\n",
        "that object. (only certain neurons are trained to detect this object) so if the pixels of the same object are passed to other set of neurons the network\n",
        "will fail. To solve this problem we need to train a very large network and a\n",
        "large amount of training data to cover every case.\n"
      ],
      "metadata": {
        "id": "AldcgYWKm-QE"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "What is gradient vanishing and gradient explosion, and how GRU/LSTM tries to mitigate this problem?\n",
        "* When training a deep neural network with gradient based learning and backpropagation, we find the partial derivatives by traversing the network from the the final layer (y_hat) to the initial layer. Using the chain rule, layers that are deeper into the network go through continuous matrix multiplications in order to compute their derivatives.\n",
        "* gradient vanishing: if the derivatives are small then the gradient will decrease exponentially as we propagate through the model until it eventually vanishes\n",
        "* gradient explosion: If the derivatives are large then the gradient will increase exponentially as we propagate down the model until they eventually explode\n",
        "* GRU/LSTM tries to mitigate this problem by using a unique additive gradient structure that includes direct access to the forget gate’s activations, enabling the network to encourage desired behaviour from the error gradient using frequent gates update on every time step of the learning process so that the information that is relevant for long-term sequence modeling is retained, and information that is only required for short-term learning is \"forgotten\".\n",
        "\n",
        "ref: https://medium.datadriveninvestor.com/how-do-lstm-networks-solve-the-problem-of-vanishing-gradients-a6784971a577\n",
        "\n",
        "ref: https://towardsdatascience.com/the-vanishing-exploding-gradient-problem-in-deep-neural-networks-191358470c11"
      ],
      "metadata": {
        "id": "upvRvtS9m_y6"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "What is multi-objective/multi-task learning? What is multi-modality learning? How do you use them in this assignment?\n",
        "* multi-objective/multi-task: Multitasking refers to either the ability to pay attention to several pieces of information at the same time or the process of performing more than one task at the same time.\n",
        "\n",
        "* multi-modality: the interplay between different representational modes, for instance, between images and written/spoken word. or combining different modalities or types of information for improving performance\n",
        "\n",
        "* for multi-task: we designed only one network to predict two ouputs (the type and price)\n",
        "\n",
        "* for multi-modality: we used image and text together as input to separate networks and compine the results before the output layer (prediction layers)"
      ],
      "metadata": {
        "id": "rGONZkVOnDth"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "What is the difference among xgboost, lightgbm and catboost\n",
        "\n",
        "* XGBoost: XGBoost cannot handle categorical\n",
        "features by itself, it only accepts numerical values similar to Random Forest. Therefore one has to perform various encodings like label encoding, mean encoding or one-hot encoding before supplying categorical data to XGBoost\n",
        "* CatBoost: CatBoost determines different categories automatically with no need for preprocessing.\n",
        "\n",
        "    * CatBoost is designed for categorical data and is known to have the best performance on it, showing the state-of-the-art performance over XGBoost and LightGBM in eight datasets in its official journal article.\n",
        "\n",
        "    * Symmetric trees: CatBoost builds symmetric (balanced) trees, unlike XGBoost and LightGBM. In every step, leaves from the previous tree are split using the same condition. The feature-split pair that accounts for the lowest loss is selected and used for all the level’s nodes. This balanced tree architecture aids in efficient CPU implementation, decreases prediction time, makes swift model appliers, and controls overfitting as the structure serves as regularization.\n",
        "\n",
        "    * Ordered boosting:  Classic boosting algorithms are prone to overfitting on small/noisy datasets due to a problem known as prediction shift.\n",
        "\n",
        "    * Native feature support: CatBoost supports all kinds of features be it numeric, categorical, or text and saves time and effort of preprocessing.\n",
        "\n",
        "    * CatBoost has the flexibility of giving indices of categorical columns so that it can be encoded as one-hot encoding using one_hot_max_size (Use one-hot encoding for all features with number of different values less than or equal to the given parameter value).\n",
        "If you don’t pass any anything in cat_features argument, CatBoost will treat all the columns as numerical variables.\n",
        "\n",
        "\n",
        "* LightGBM: Similar to CatBoost, LightGBM can also handle categorical features by taking the input of feature names. It does not convert to one-hot coding, and is much faster than one-hot coding. LGBM uses a special algorithm to find the split value of categorical features\n",
        "\n",
        "\n",
        "ref:https://towardsdatascience.com/catboost-vs-light-gbm-vs-xgboost-5f93620723db\n",
        "\n",
        "ref: https://neptune.ai/blog/when-to-choose-catboost-over-xgboost-or-lightgbm\n"
      ],
      "metadata": {
        "id": "LTKkW32YnG9S"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "cZ59SX7Bm61X"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}