MainCode.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "toc_visible": true,
      "gpuType": "T4"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU",
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "fac2919f7a2045a792085d01d84bfa86": {
          "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_908ba1eedfee4bdeb1f35566666609b5",
              "IPY_MODEL_46b514d202bb4959aefa38fe22bd88c0",
              "IPY_MODEL_0dbbb1a1133c44b5bbadfe2c25fea3b9"
            ],
            "layout": "IPY_MODEL_8d770239e7db43c7a63486d2dcba07ef"
          }
        },
        "908ba1eedfee4bdeb1f35566666609b5": {
          "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_877e783b6b4e4a4dbb34234990b2681e",
            "placeholder": "​",
            "style": "IPY_MODEL_5f5e3fec69004f0ab7135ce44747dd07",
            "value": "Downloading (…)lve/main/config.json: 100%"
          }
        },
        "46b514d202bb4959aefa38fe22bd88c0": {
          "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_0b645147a3bb44379ae50628b933d811",
            "max": 684,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_58d83848f06a41b3ac4c952eb913e6f8",
            "value": 684
          }
        },
        "0dbbb1a1133c44b5bbadfe2c25fea3b9": {
          "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_762ae5c6efca4a45a487e7157ccd9d1b",
            "placeholder": "​",
            "style": "IPY_MODEL_6a4c8bc5b6774bd589ec46a530851d39",
            "value": " 684/684 [00:00&lt;00:00, 15.9kB/s]"
          }
        },
        "8d770239e7db43c7a63486d2dcba07ef": {
          "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
          }
        },
        "877e783b6b4e4a4dbb34234990b2681e": {
          "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
          }
        },
        "5f5e3fec69004f0ab7135ce44747dd07": {
          "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": ""
          }
        },
        "0b645147a3bb44379ae50628b933d811": {
          "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
          }
        },
        "58d83848f06a41b3ac4c952eb913e6f8": {
          "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": ""
          }
        },
        "762ae5c6efca4a45a487e7157ccd9d1b": {
          "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
          }
        },
        "6a4c8bc5b6774bd589ec46a530851d39": {
          "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": ""
          }
        },
        "13e8d36217a946c9b6b4420f6ca2a19c": {
          "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_426de396581945e093160bef6a6d4bd9",
              "IPY_MODEL_d9d46af007b449f79f309bb158dc88a5",
              "IPY_MODEL_b898b80900504a9491e0631351e30188"
            ],
            "layout": "IPY_MODEL_c4855e4955aa4840ab3d998ec12be433"
          }
        },
        "426de396581945e093160bef6a6d4bd9": {
          "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_e58dc03387334b13b567413eed33ec2d",
            "placeholder": "​",
            "style": "IPY_MODEL_f7f9629826614886b4a2ec9187293620",
            "value": "Downloading (…)ve/main/spiece.model: 100%"
          }
        },
        "d9d46af007b449f79f309bb158dc88a5": {
          "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_6f6b64b779664a14bfab0b082847a713",
            "max": 1882978,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_08429e11961447189846b97dee04186c",
            "value": 1882978
          }
        },
        "b898b80900504a9491e0631351e30188": {
          "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_821cb81683974ea5ba2488ff41aeb3f0",
            "placeholder": "​",
            "style": "IPY_MODEL_08a01159787440dda7a02041a9c05250",
            "value": " 1.88M/1.88M [00:00&lt;00:00, 18.7MB/s]"
          }
        },
        "c4855e4955aa4840ab3d998ec12be433": {
          "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
          }
        },
        "e58dc03387334b13b567413eed33ec2d": {
          "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
          }
        },
        "f7f9629826614886b4a2ec9187293620": {
          "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": ""
          }
        },
        "6f6b64b779664a14bfab0b082847a713": {
          "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
          }
        },
        "08429e11961447189846b97dee04186c": {
          "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": ""
          }
        },
        "821cb81683974ea5ba2488ff41aeb3f0": {
          "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
          }
        },
        "08a01159787440dda7a02041a9c05250": {
          "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": ""
          }
        },
        "d924d290ec2c488f98d29a034c4308fe": {
          "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_8acd136ef48044d884b133b027c07a89",
              "IPY_MODEL_7e76938cec534176ba49da562c332e11",
              "IPY_MODEL_f19dfe296e1f4a93bd19bccb63cc69fc"
            ],
            "layout": "IPY_MODEL_ead5fbeaec0641069ff245bf7828a6db"
          }
        },
        "8acd136ef48044d884b133b027c07a89": {
          "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_910dafa67e0949b0a6371084e808b192",
            "placeholder": "​",
            "style": "IPY_MODEL_0d350638b1214cd5953059a9f8ae0ab4",
            "value": "Downloading pytorch_model.bin: 100%"
          }
        },
        "7e76938cec534176ba49da562c332e11": {
          "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_4630ace2df564055a22aa1e1b4d043c0",
            "max": 73062448,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_85168b8f8b6140a7ba53e595f4575b89",
            "value": 73062448
          }
        },
        "f19dfe296e1f4a93bd19bccb63cc69fc": {
          "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_f8263188caf145e8a20f56459bb4f1e0",
            "placeholder": "​",
            "style": "IPY_MODEL_022f884ecd0e48c3a451d72412eb58ed",
            "value": " 73.1M/73.1M [00:01&lt;00:00, 43.8MB/s]"
          }
        },
        "ead5fbeaec0641069ff245bf7828a6db": {
          "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
          }
        },
        "910dafa67e0949b0a6371084e808b192": {
          "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
          }
        },
        "0d350638b1214cd5953059a9f8ae0ab4": {
          "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": ""
          }
        },
        "4630ace2df564055a22aa1e1b4d043c0": {
          "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
          }
        },
        "85168b8f8b6140a7ba53e595f4575b89": {
          "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": ""
          }
        },
        "f8263188caf145e8a20f56459bb4f1e0": {
          "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
          }
        },
        "022f884ecd0e48c3a451d72412eb58ed": {
          "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": "42hEcykHowQR"
      },
      "source": [
        "# 1.2. Load and Processs Data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "233c030b-ec9b-4736-b29f-974c069fe621",
        "id": "1f1sMZ1XowQS"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting transformers\n",
            "  Downloading transformers-4.30.0-py3-none-any.whl (7.2 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m7.2/7.2 MB\u001b[0m \u001b[31m59.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from transformers) (3.12.0)\n",
            "Collecting huggingface-hub<1.0,>=0.14.1 (from transformers)\n",
            "  Downloading huggingface_hub-0.15.1-py3-none-any.whl (236 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m236.8/236.8 kB\u001b[0m \u001b[31m27.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.10/dist-packages (from transformers) (1.22.4)\n",
            "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.10/dist-packages (from transformers) (23.1)\n",
            "Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.10/dist-packages (from transformers) (6.0)\n",
            "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.10/dist-packages (from transformers) (2022.10.31)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from transformers) (2.27.1)\n",
            "Collecting tokenizers!=0.11.3,<0.14,>=0.11.1 (from transformers)\n",
            "  Downloading tokenizers-0.13.3-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (7.8 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m7.8/7.8 MB\u001b[0m \u001b[31m87.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting safetensors>=0.3.1 (from transformers)\n",
            "  Downloading safetensors-0.3.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.3 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.3/1.3 MB\u001b[0m \u001b[31m79.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.10/dist-packages (from transformers) (4.65.0)\n",
            "Requirement already satisfied: fsspec in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.14.1->transformers) (2023.4.0)\n",
            "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.14.1->transformers) (4.5.0)\n",
            "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (1.26.15)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (2022.12.7)\n",
            "Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (2.0.12)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (3.4)\n",
            "Installing collected packages: tokenizers, safetensors, huggingface-hub, transformers\n",
            "Successfully installed huggingface-hub-0.15.1 safetensors-0.3.1 tokenizers-0.13.3 transformers-4.30.0\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.3/1.3 MB\u001b[0m \u001b[31m27.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hLooking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: sentencepiece in /usr/local/lib/python3.10/dist-packages (0.1.99)\n"
          ]
        }
      ],
      "source": [
        "# Installing required packages\n",
        "!pip install transformers\n",
        "!pip install -q sentencepiece\n",
        "!pip install sentencepiece\n",
        "\n",
        "# Importing necessary libraries\n",
        "import re\n",
        "import json\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "from tqdm import tqdm\n",
        "from tqdm.auto import tqdm\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.utils import shuffle\n",
        "from transformers import BertModel, BertConfig, BertTokenizer, AdamW"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a0758768-ee40-4ec6-d60c-17e903d036db",
        "id": "bT1lsPmsotOM"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Cloning into 'PQuAD'...\n",
            "remote: Enumerating objects: 27, done.\u001b[K\n",
            "remote: Counting objects: 100% (27/27), done.\u001b[K\n",
            "remote: Compressing objects: 100% (22/22), done.\u001b[K\n",
            "remote: Total 27 (delta 9), reused 15 (delta 3), pack-reused 0\u001b[K\n",
            "Unpacking objects: 100% (27/27), 5.71 MiB | 2.88 MiB/s, done.\n"
          ]
        }
      ],
      "source": [
        "# Clone the PQuAD dataset from GitHub\n",
        "!git clone https://github.com/AUT-NLP/PQuAD.git\n",
        "\n",
        "# Load the PQuad dataset\n",
        "train_file_path = '/content/PQuAD/Dataset/Train.json'\n",
        "test_file_path = '/content/PQuAD/Dataset/Test.json'\n",
        "val_file_path = '/content/PQuAD/Dataset/Validation.json'"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "78bfdc71-6052-4bdf-fa9f-d5402c040734",
        "id": "B_QHHJkNotOO"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Dataset Statistics (Train):\n",
            "Total Questions: 8979\n",
            "Total Unanswerable Questions: 15721\n",
            "Mean # of Paragraph Tokens: 919.56\n",
            "Mean # of Question Tokens: 73.69\n",
            "Mean # of Answer Tokens: 28.06\n",
            "\n"
          ]
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Dataset Statistics (Test):\n",
            "Total Questions: 1059\n",
            "Total Unanswerable Questions: 1914\n",
            "Mean # of Paragraph Tokens: 968.15\n",
            "Mean # of Question Tokens: 81.91\n",
            "Mean # of Answer Tokens: 41.80\n",
            "\n"
          ]
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Dataset Statistics (Validation):\n",
            "Total Questions: 1103\n",
            "Total Unanswerable Questions: 1981\n",
            "Mean # of Paragraph Tokens: 905.68\n",
            "Mean # of Question Tokens: 76.47\n",
            "Mean # of Answer Tokens: 42.83\n",
            "\n"
          ]
        }
      ],
      "source": [
        "import json\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "# Path to the dataset files\n",
        "train_file_path = '/content/PQuAD/Dataset/Train.json'\n",
        "test_file_path = '/content/PQuAD/Dataset/Test.json'\n",
        "val_file_path = '/content/PQuAD/Dataset/Validation.json'\n",
        "\n",
        "def load_dataset(file_path):\n",
        "    with open(file_path, 'r', encoding='utf-8') as file:\n",
        "        dataset = json.load(file)\n",
        "    return dataset\n",
        "\n",
        "def plot_answer_distribution(dataset, dataset_name, file_name):\n",
        "    num_answers = []\n",
        "    for data in dataset['data']:\n",
        "        for paragraph in data['paragraphs']:\n",
        "            for qa in paragraph['qas']:\n",
        "                num_answers.append(len(qa['answers']))\n",
        "\n",
        "    bins = np.arange(max(num_answers) + 2)\n",
        "    hist, edges = np.histogram(num_answers, bins=bins)\n",
        "\n",
        "    fig, ax = plt.subplots()\n",
        "    ax.bar(edges[:-1], hist, width=0.8, align='center', color='steelblue')\n",
        "\n",
        "    # Add count labels in the middle and above each bar\n",
        "    for i, count in enumerate(hist):\n",
        "        if count > 0:\n",
        "            ax.text(edges[i], count + 1, str(int(count)), ha='center', va='bottom')\n",
        "\n",
        "    plt.xlabel('Number of Answers')\n",
        "    plt.ylabel('Number of Questions')\n",
        "    plt.title(f'Distribution of Number of Answers for Questions ({dataset_name})')\n",
        "    plt.xticks(np.arange(max(num_answers) + 1), [f'{i}-answer' for i in range(max(num_answers) + 1)], rotation=90)\n",
        "    plt.legend([dataset_name])\n",
        "    plt.savefig(file_name, format='pdf', bbox_inches='tight')\n",
        "    plt.show()\n",
        "\n",
        "def print_dataset_statistics(dataset, dataset_name):\n",
        "    total_questions = 0\n",
        "    total_unanswerable_questions = 0\n",
        "    total_paragraph_tokens = 0\n",
        "    total_question_tokens = 0\n",
        "    total_answer_tokens = 0\n",
        "\n",
        "    for data in dataset['data']:\n",
        "        total_questions += len(data['paragraphs'])\n",
        "        for paragraph in data['paragraphs']:\n",
        "            for qa in paragraph['qas']:\n",
        "                if qa['is_impossible']:\n",
        "                    total_unanswerable_questions += 1\n",
        "                total_paragraph_tokens += len(paragraph['context'].split())\n",
        "                total_question_tokens += len(qa['question'].split())\n",
        "                total_answer_tokens += sum([len(answer['text'].split()) for answer in qa['answers']])\n",
        "\n",
        "    mean_paragraph_tokens = total_paragraph_tokens / total_questions\n",
        "    mean_question_tokens = total_question_tokens / total_questions\n",
        "    mean_answer_tokens = total_answer_tokens / total_questions\n",
        "\n",
        "    print(f\"Dataset Statistics ({dataset_name}):\")\n",
        "    print(f\"Total Questions: {total_questions}\")\n",
        "    print(f\"Total Unanswerable Questions: {total_unanswerable_questions}\")\n",
        "    print(f\"Mean # of Paragraph Tokens: {mean_paragraph_tokens:.2f}\")\n",
        "    print(f\"Mean # of Question Tokens: {mean_question_tokens:.2f}\")\n",
        "    print(f\"Mean # of Answer Tokens: {mean_answer_tokens:.2f}\")\n",
        "    print()\n",
        "\n",
        "# Load the train dataset\n",
        "train_dataset = load_dataset(train_file_path)\n",
        "\n",
        "# Plot the distribution of the number of answers for train questions and save as PDF\n",
        "plot_answer_distribution(train_dataset, 'Train', 'traindistribution.pdf')\n",
        "\n",
        "# Print the statistical information of the train dataset\n",
        "print_dataset_statistics(train_dataset, 'Train')\n",
        "\n",
        "# Load the test dataset\n",
        "test_dataset = load_dataset(test_file_path)\n",
        "\n",
        "# Plot the distribution of the number of answers for test questions and save as PDF\n",
        "plot_answer_distribution(test_dataset, 'Test', 'testdistribution.pdf')\n",
        "\n",
        "# Print the statistical information of the test dataset\n",
        "print_dataset_statistics(test_dataset, 'Test')\n",
        "\n",
        "# Load the validation dataset\n",
        "val_dataset = load_dataset(val_file_path)\n",
        "\n",
        "# Plot the distribution of the number of answers for validation questions and save as PDF\n",
        "plot_answer_distribution(val_dataset, 'Validation', 'validationdistribution.pdf')\n",
        "\n",
        "# Print the statistical information of the validation dataset\n",
        "print_dataset_statistics(val_dataset, 'Validation')\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "mr7BKsIdotOR"
      },
      "outputs": [],
      "source": [
        "# Set the minimum and maximum length for data\n",
        "min_len, max_len = 128, 256\n",
        "\n",
        "# Set the value of N\n",
        "N = 3"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 241
        },
        "outputId": "282f6018-cf37-44b4-e679-fe6863eadfeb",
        "id": "CMInwCXjotOS"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Size of the train DataFrame before concatenation is 63994\n",
            "Size of the train DataFrame after concatenation is 63994\n"
          ]
        },
        {
          "data": {
            "text/html": [
              "\n",
              "  <div id=\"df-0084ce54-0b8d-4f27-8025-7a48c821f11f\">\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>Id</th>\n",
              "      <th>title</th>\n",
              "      <th>context</th>\n",
              "      <th>question</th>\n",
              "      <th>ans_start</th>\n",
              "      <th>text</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>101001</td>\n",
              "      <td>آرسنال</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>موقعیت جغرافی باشگاه فوتبال آرسنال را بگویید؟</td>\n",
              "      <td>86</td>\n",
              "      <td>شمال شهر لندن</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>101002</td>\n",
              "      <td>آرسنال</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>لیگ برتر انگلستان موفق به کسب چند عنوان قهرمان...</td>\n",
              "      <td>173</td>\n",
              "      <td>۱۴</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>101003</td>\n",
              "      <td>آرسنال</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>بیشترین بازی بدون باخت پیاپی متعلق به کدام باش...</td>\n",
              "      <td>61</td>\n",
              "      <td>باشگاه فوتبال انگلیسی</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>101004</td>\n",
              "      <td>آرسنال</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>باشگاه فوتبال آرسنال موفق به کسب چند عنوان قهر...</td>\n",
              "      <td>119</td>\n",
              "      <td>۱۳</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>101005</td>\n",
              "      <td>آرسنال</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>باشگاه فوتبال آرسنال چند عنوان قهرمانی در جام ...</td>\n",
              "      <td>214</td>\n",
              "      <td>۱۶</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-0084ce54-0b8d-4f27-8025-7a48c821f11f')\"\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-0084ce54-0b8d-4f27-8025-7a48c821f11f 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-0084ce54-0b8d-4f27-8025-7a48c821f11f');\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",
              "  "
            ],
            "text/plain": [
              "       Id   title                                            context  \\\n",
              "0  101001  آرسنال  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "1  101002  آرسنال  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "2  101003  آرسنال  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "3  101004  آرسنال  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "4  101005  آرسنال  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "\n",
              "                                            question  ans_start  \\\n",
              "0      موقعیت جغرافی باشگاه فوتبال آرسنال را بگویید؟         86   \n",
              "1  لیگ برتر انگلستان موفق به کسب چند عنوان قهرمان...        173   \n",
              "2  بیشترین بازی بدون باخت پیاپی متعلق به کدام باش...         61   \n",
              "3  باشگاه فوتبال آرسنال موفق به کسب چند عنوان قهر...        119   \n",
              "4  باشگاه فوتبال آرسنال چند عنوان قهرمانی در جام ...        214   \n",
              "\n",
              "                    text  \n",
              "0          شمال شهر لندن  \n",
              "1                     ۱۴  \n",
              "2  باشگاه فوتبال انگلیسی  \n",
              "3                     ۱۳  \n",
              "4                     ۱۶  "
            ]
          },
          "execution_count": 5,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "def json_to_dataframe(file):\n",
        "    # Open the JSON file\n",
        "    f = open(file, \"r\")\n",
        "    data = json.loads(f.read())  # Load the JSON file\n",
        "\n",
        "    # Create empty lists to store values\n",
        "    ids = []\n",
        "    titles = []\n",
        "    contexts = []\n",
        "    questions = []\n",
        "    ans_starts = []\n",
        "    texts = []\n",
        "\n",
        "    # Iterate over the JSON data\n",
        "    for i in range(len(data['data'])):\n",
        "        title = data['data'][i]['title']  # Extract the 'title' value\n",
        "\n",
        "        # Iterate over the paragraphs in the JSON data\n",
        "        for p in range(len(data['data'][i]['paragraphs'])):\n",
        "            context = data['data'][i]['paragraphs'][p]['context']  # Extract the 'context' value\n",
        "\n",
        "            # Iterate over the questions in the JSON data\n",
        "            for q in range(len(data['data'][i]['paragraphs'][p]['qas'])):\n",
        "                question = data['data'][i]['paragraphs'][p]['qas'][q]['question']  # Extract the 'question' value\n",
        "                qid = data['data'][i]['paragraphs'][p]['qas'][q]['id']  # Extract the 'id' value\n",
        "\n",
        "                # Check if the question has answers\n",
        "                if len(data['data'][i]['paragraphs'][p]['qas'][q]['answers']) == 0:\n",
        "                    ans_start = -1\n",
        "                    text = ''\n",
        "\n",
        "                    # Append the values to the lists\n",
        "                    titles.append(title)\n",
        "                    contexts.append(context)\n",
        "                    questions.append(question)\n",
        "                    ids.append(qid)\n",
        "                    ans_starts.append(ans_start)\n",
        "                    texts.append(text)\n",
        "                else:\n",
        "                    # Iterate over the answers in the JSON data\n",
        "                    for a in range(len(data['data'][i]['paragraphs'][p]['qas'][q]['answers'])):\n",
        "                        ans_start = data['data'][i]['paragraphs'][p]['qas'][q]['answers'][a]['answer_start']  # Extract the 'answer_start' value\n",
        "                        text = data['data'][i]['paragraphs'][p]['qas'][q]['answers'][a]['text']  # Extract the 'text' value\n",
        "\n",
        "                        # Append the values to the lists\n",
        "                        titles.append(title)\n",
        "                        contexts.append(context)\n",
        "                        questions.append(question)\n",
        "                        ids.append(qid)\n",
        "                        ans_starts.append(ans_start)\n",
        "                        texts.append(text)\n",
        "\n",
        "    # Create an empty DataFrame\n",
        "    new_df = pd.DataFrame(columns=['Id', 'title', 'context', 'question', 'ans_start', 'text'])\n",
        "\n",
        "    # Set the values of the DataFrame columns\n",
        "    new_df['Id'] = ids\n",
        "    new_df['title'] = titles\n",
        "    new_df['context'] = contexts\n",
        "    new_df['question'] = questions\n",
        "    new_df['ans_start'] = ans_starts\n",
        "    new_df['text'] = texts\n",
        "\n",
        "    # Drop duplicate rows from the DataFrame\n",
        "    final_df = new_df.drop_duplicates(keep='first')\n",
        "\n",
        "    return final_df\n",
        "\n",
        "\n",
        "# Convert the train JSON file to a DataFrame\n",
        "df_train = json_to_dataframe(train_file_path)\n",
        "\n",
        "# Get the number of rows in the train DataFrame\n",
        "train_rows = df_train.shape[0]\n",
        "print('Size of the train DataFrame before concatenation is {}'.format(train_rows))\n",
        "\n",
        "# Convert the test JSON file to a DataFrame\n",
        "df_test = json_to_dataframe(test_file_path)\n",
        "\n",
        "# Convert the validation JSON file to a DataFrame\n",
        "df_validation = json_to_dataframe(val_file_path)\n",
        "\n",
        "# Concatenate the train and validation DataFrames\n",
        "# frames = [df_train, df_validation]\n",
        "# df_train = pd.concat(frames)\n",
        "\n",
        "# Get the number of rows in the concatenated train DataFrame\n",
        "train_rows = df_train.shape[0]\n",
        "print('Size of the train DataFrame after concatenation is {}'.format(train_rows))\n",
        "\n",
        "# Display the first few rows of the train DataFrame\n",
        "df_train.head()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "4LGi4GBjotOW"
      },
      "outputs": [],
      "source": [
        "def add_end_index(answers_text, answers_start, contexts):\n",
        "    new_answers = []\n",
        "\n",
        "    # Loop through each answer-context pair\n",
        "    for answer_text, answer_start, context in tqdm(zip(answers_text, answers_start, contexts)):\n",
        "        start_shift = 0\n",
        "\n",
        "        # Remove start half-spaces from the answer text\n",
        "        text = re.sub(\"^\\u200c\", \"\", answer_text)\n",
        "        \n",
        "        # Check if the length of the text is one less than the length of the original answer\n",
        "        if len(list(text)) == (len(list(answer_text)) - 1):\n",
        "            start_shift += 1\n",
        "        \n",
        "        # Remove end half-spaces from the text\n",
        "        text = re.sub(\"\\u200c$\", \"\", text)\n",
        "        \n",
        "        # Remove leading and trailing whitespaces from the text\n",
        "        text = re.sub(\"^\\s+\", '', text)\n",
        "        text = re.sub(\"\\s+$\", '', text)\n",
        "\n",
        "        # Adjust the answer_start index by the start_shift value\n",
        "        answer_start += start_shift\n",
        "        \n",
        "        # Calculate the end index of the answer\n",
        "        end_idx = answer_start + len(text)\n",
        "\n",
        "        # Check if the answer is correct\n",
        "        if context[answer_start:end_idx] == text:\n",
        "            # If the answer is correct, set the answer_end index to end_idx\n",
        "            answer_end = end_idx\n",
        "        else:\n",
        "            # If the answer is off by 1-2 tokens, adjust the answer_start and answer_end indices\n",
        "            for n in [1, 2]:\n",
        "                if context[answer_start - n:end_idx - n] == text:\n",
        "                    answer_start = answer_start - n\n",
        "                    answer_end = end_idx - n\n",
        "\n",
        "        # Append the modified answer to the new_answers list\n",
        "        new_answers.append({'text': text, 'answer_start': answer_start, 'answer_end': answer_end})\n",
        "\n",
        "    return new_answers\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "sixmjbpwotOX"
      },
      "outputs": [],
      "source": [
        "def prepare_data(dataset):\n",
        "    # Extract necessary columns from the dataset\n",
        "    answer_start = dataset['ans_start'].tolist()\n",
        "    text = dataset['text'].tolist()\n",
        "    questions = dataset['question'].tolist()\n",
        "    contexts = dataset['context'].tolist()\n",
        "\n",
        "    # Call the add_end_index function to process answers\n",
        "    answers = add_end_index(text, answer_start, contexts)\n",
        "\n",
        "    # Return a dictionary with the prepared data\n",
        "    return {\n",
        "        'question': questions,\n",
        "        'context': contexts,\n",
        "        'answers': answers\n",
        "    }\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "26550fed-4b7a-450a-a54d-c476ecc0bfd2",
        "id": "jeyLpvHXotOY"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "63994it [00:00, 148179.51it/s]\n",
            "10379it [00:00, 134221.35it/s]\n",
            "10417it [00:00, 115438.75it/s]\n"
          ]
        },
        {
          "data": {
            "text/plain": [
              "{'text': 'پنج عنوان قهرمانی در چمپیونشیپ و دو قهرمانی در جام حذفی',\n",
              " 'answer_start': 288,\n",
              " 'answer_end': 343}"
            ]
          },
          "execution_count": 8,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# Prepare the training dataset\n",
        "train_dataset = prepare_data(df_train)\n",
        "\n",
        "# Prepare the validation dataset\n",
        "val_dataset = prepare_data(df_validation)\n",
        "\n",
        "# Prepare the test dataset\n",
        "test_dataset = prepare_data(df_test)\n",
        "\n",
        "# Access the second answer in the training dataset\n",
        "answer = train_dataset['answers'][10]\n",
        "train_dataset['answers'][10]"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# 1.3. ParsBERT (Data Processing and Training)"
      ],
      "metadata": {
        "id": "FKVAJUxGo5AN"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 81,
          "referenced_widgets": [
            "523cee3b13c047b9b1b751a66e892aa4",
            "a70762f851db4b5a88522d4ef22a26e5",
            "0e326990c8c04c6e9d9e963a86798be8",
            "4c52af20921e49e69157c663af3a7e5f",
            "2a1d8b642bf24a2fb9a6f64ecab14d0f",
            "568850c91c704a608352d4d901386ffc",
            "72fa645faf3740acb52527988a7c7613",
            "b36ae226730a41f58e63161386fdec8a",
            "9fcb3efc8db647698629c589a946c041",
            "0d60fdcf457a48cfbf9335053a5fd2e4",
            "fb8846b4bc2a45f386deaec19983715e",
            "dcbb2a5bba0545158569d403d59ce4fa",
            "6c538f4dcbfe4269ad5a199f9710d421",
            "0c3193bdd222439d979a5f138d0ac32a",
            "d3ac28cb8c3e44bbaa303fb49146772f",
            "9a855bcaeb6f45b490dad700bd58d61f",
            "fe44cafa9bb24cb4aa9d958e2425fbeb",
            "7382b04b970f41c2a8e9e2b15624147e",
            "2cb920bc986c4ea89b04240ecf1ca12c",
            "7055aa4aea54495abbbbead7a0b3e50c",
            "03fa105a43da4daca42e4fa3a0a5c68d",
            "216f8f27554c4eebaf1c6606158d1b5f"
          ]
        },
        "id": "K5xP-x1hyaWl",
        "outputId": "2f73d4ed-41b1-471b-fc54-5d4a7107ba21"
      },
      "outputs": [
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "523cee3b13c047b9b1b751a66e892aa4",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "Downloading (…)solve/main/vocab.txt:   0%|          | 0.00/1.22M [00:00<?, ?B/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "dcbb2a5bba0545158569d403d59ce4fa",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "Downloading (…)lve/main/config.json:   0%|          | 0.00/434 [00:00<?, ?B/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Specify the model name or path\n",
        "model_name_or_path = 'HooshvareLab/bert-base-parsbert-uncased'\n",
        "\n",
        "# Create a tokenizer instance from the specified model\n",
        "tokenizer = BertTokenizer.from_pretrained(model_name_or_path)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9598db53-1ae6-411b-cd57-7efaf9ff97f4",
        "id": "rH2riRhjotOa"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Total size of data is 63994\n",
            "Number of data without [UNK] and containing 128-256 tokens is 51797\n",
            "Percentage of data without [UNK] and containing 128-256 tokens is 80.94040066256211\n"
          ]
        }
      ],
      "source": [
        "# Create empty lists to store the new tokenized data\n",
        "new_context, new_question, new_answer = [[] for _ in range(3)]\n",
        "\n",
        "# Get the total size of the data\n",
        "len_data = len(train_dataset['answers'])\n",
        "print('Total size of data is {}'.format(len_data))\n",
        "\n",
        "# Iterate through each data instance\n",
        "for i in range(len_data):\n",
        "    # Tokenize the answer, context, and question\n",
        "    tokenized_answer = tokenizer.tokenize(train_dataset['answers'][i]['text'])\n",
        "    tokenized_context = tokenizer.tokenize(train_dataset['context'][i])\n",
        "    tokenized_question = tokenizer.tokenize(train_dataset['question'][i])\n",
        "    \n",
        "    # Calculate the total number of tokens\n",
        "    num = len(tokenized_context) + len(tokenized_question)\n",
        "    \n",
        "    # Check if the total number of tokens is within the desired range\n",
        "    if num > (min_len - 3) and num <= (max_len - 3):  # 3 for three special tokens: 1 for [CLS] and 2 for [SEP]\n",
        "        if '[UNK]' not in tokenized_answer:\n",
        "            # Append the tokenized data to the new lists\n",
        "            new_context.append(train_dataset['context'][i])\n",
        "            new_question.append(train_dataset['question'][i])\n",
        "            new_answer.append({\n",
        "                'text': train_dataset['answers'][i]['text'],\n",
        "                'answer_start': train_dataset['answers'][i]['answer_start'],\n",
        "                'answer_end': train_dataset['answers'][i]['answer_end']\n",
        "            })\n",
        "\n",
        "# Print the number of data without [UNK] and containing 128-256 tokens\n",
        "print('Number of data without [UNK] and containing 128-256 tokens is {}'.format(len(new_context)))\n",
        "\n",
        "# Print the percentage of data without [UNK] and containing 128-256 tokens\n",
        "print('Percentage of data without [UNK] and containing 128-256 tokens is {}'.format(100 * len(new_context) / len_data))\n",
        "\n",
        "# Create a new training dataset with the filtered data\n",
        "new_train_dataset = {\n",
        "    'question': new_question,\n",
        "    'context': new_context,\n",
        "    'answers': new_answer\n",
        "}\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4b30bfa8-e207-4274-f675-f46bc9c75f80",
        "id": "eGS4bb_botOb"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Total size of data is 10379\n",
            "Number of data without [UNK] and containing 128-256 tokens is 8258\n",
            "Percentage of data without [UNK] and containing 128-256 tokens is 79.56450525098757\n"
          ]
        }
      ],
      "source": [
        "# Create empty lists to store the new tokenized data\n",
        "new_context, new_question, new_answer = [[] for _ in range(3)]\n",
        "\n",
        "# Get the total size of the data\n",
        "len_data = len(val_dataset['answers'])\n",
        "print('Total size of data is {}'.format(len_data))\n",
        "\n",
        "# Iterate through each data instance\n",
        "for i in range(len_data):\n",
        "    # Tokenize the answer, context, and question\n",
        "    tokenized_answer = tokenizer.tokenize(val_dataset['answers'][i]['text'])\n",
        "    tokenized_context = tokenizer.tokenize(val_dataset['context'][i])\n",
        "    tokenized_question = tokenizer.tokenize(val_dataset['question'][i])\n",
        "    \n",
        "    # Calculate the total number of tokens\n",
        "    num = len(tokenized_context) + len(tokenized_question)\n",
        "    \n",
        "    # Check if the total number of tokens is within the desired range\n",
        "    if num > (min_len - 3) and num <= (max_len - 3):  # 3 for three special tokens, [CLS] and 2 [SEP]\n",
        "        if '[UNK]' not in tokenized_answer:\n",
        "            # Append the tokenized data to the new lists\n",
        "            new_context.append(val_dataset['context'][i])\n",
        "            new_question.append(val_dataset['question'][i])\n",
        "            new_answer.append({\n",
        "                'text': val_dataset['answers'][i]['text'],\n",
        "                'answer_start': val_dataset['answers'][i]['answer_start'],\n",
        "                'answer_end': val_dataset['answers'][i]['answer_end']\n",
        "            })\n",
        "\n",
        "# Print the number of data without [UNK] and containing 128-256 tokens\n",
        "print('Number of data without [UNK] and containing 128-256 tokens is {}'.format(len(new_context)))\n",
        "\n",
        "# Print the percentage of data without [UNK] and containing 128-256 tokens\n",
        "print('Percentage of data without [UNK] and containing 128-256 tokens is {}'.format(100 * len(new_context) / len_data))\n",
        "\n",
        "# Create a new validation dataset with the filtered data\n",
        "new_val_dataset = {\n",
        "    'question': new_question,\n",
        "    'context': new_context,\n",
        "    'answers': new_answer\n",
        "}\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "636bd979-b883-4db1-b800-4fb4c5a8fd38",
        "id": "kBtdGZd_otOd"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Total size of data is 10417\n",
            "Number of data without [UNK] and containing 128-256 tokens is 8367\n",
            "Percentage of data without [UNK] and containing 128-256 tokens is 80.32062973984833\n"
          ]
        }
      ],
      "source": [
        "# Create empty lists to store the new tokenized data\n",
        "new_context, new_question, new_answer = [[] for _ in range(3)]\n",
        "\n",
        "# Get the total size of the data\n",
        "len_data = len(test_dataset['answers'])\n",
        "print('Total size of data is {}'.format(len_data))\n",
        "\n",
        "# Iterate through each data instance\n",
        "for i in range(len_data):\n",
        "    # Tokenize the answer, context, and question\n",
        "    tokenized_answer = tokenizer.tokenize(test_dataset['answers'][i]['text'])\n",
        "    tokenized_context = tokenizer.tokenize(test_dataset['context'][i])\n",
        "    tokenized_question = tokenizer.tokenize(test_dataset['question'][i])\n",
        "    \n",
        "    # Calculate the total number of tokens\n",
        "    num = len(tokenized_context) + len(tokenized_question)\n",
        "    \n",
        "    # Check if the total number of tokens is within the desired range\n",
        "    if num > (min_len - 3) and num <= (max_len - 3):  # 3 for three special tokens, [CLS] and 2 [SEP]\n",
        "        if '[UNK]' not in tokenized_answer:\n",
        "            # Append the tokenized data to the new lists\n",
        "            new_context.append(test_dataset['context'][i])\n",
        "            new_question.append(test_dataset['question'][i])\n",
        "            new_answer.append({\n",
        "                'text': test_dataset['answers'][i]['text'],\n",
        "                'answer_start': test_dataset['answers'][i]['answer_start'],\n",
        "                'answer_end': test_dataset['answers'][i]['answer_end']\n",
        "            })\n",
        "\n",
        "# Print the number of data without [UNK] and containing 128-256 tokens\n",
        "print('Number of data without [UNK] and containing 128-256 tokens is {}'.format(len(new_context)))\n",
        "\n",
        "# Print the percentage of data without [UNK] and containing 128-256 tokens\n",
        "print('Percentage of data without [UNK] and containing 128-256 tokens is {}'.format(100 * len(new_context) / len_data))\n",
        "\n",
        "# Create a new validation dataset with the filtered data\n",
        "new_test_dataset = {\n",
        "    'question': new_question,\n",
        "    'context': new_context,\n",
        "    'answers': new_answer\n",
        "}\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "XMBCwePeotOe"
      },
      "outputs": [],
      "source": [
        "# Create a DataFrame for the test dataset\n",
        "test_df = pd.DataFrame.from_dict(new_test_dataset)\n",
        "\n",
        "# Create a DataFrame for the validation dataset\n",
        "validation_df = pd.DataFrame.from_dict(new_val_dataset)\n",
        "\n",
        "# Create a DataFrame for the training dataset\n",
        "train_df = pd.DataFrame.from_dict(new_train_dataset)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 241
        },
        "outputId": "8932c0cf-7a1a-45f3-b7d2-6867c8879cea",
        "id": "TEWySJ8iotOf"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Length of the training dataset: 51797\n",
            "Training dataset:\n"
          ]
        },
        {
          "data": {
            "text/html": [
              "\n",
              "  <div id=\"df-45715f08-7cca-464d-8390-9e2e65023943\">\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>question</th>\n",
              "      <th>context</th>\n",
              "      <th>answers</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>لیگ برتر انگلستان موفق به کسب چند عنوان قهرمان...</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>{'text': '۱۴', 'answer_start': 173, 'answer_en...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>بیشترین بازی بدون باخت پیاپی متعلق به کدام باش...</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>{'text': 'باشگاه فوتبال انگلیسی', 'answer_star...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>باشگاه فوتبال آرسنال موفق به کسب چند عنوان قهر...</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>{'text': '۱۳', 'answer_start': 119, 'answer_en...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>باشگاه فوتبال آرسنال چند عنوان قهرمانی در جام ...</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>{'text': '۱۶', 'answer_start': 214, 'answer_en...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>باشگاه فوتبال آرسنال چند قهرمانی در جام اتحادی...</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>{'text': 'دو قهرمانی', 'answer_start': 249, 'a...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-45715f08-7cca-464d-8390-9e2e65023943')\"\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-45715f08-7cca-464d-8390-9e2e65023943 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-45715f08-7cca-464d-8390-9e2e65023943');\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",
              "  "
            ],
            "text/plain": [
              "                                            question  \\\n",
              "0  لیگ برتر انگلستان موفق به کسب چند عنوان قهرمان...   \n",
              "1  بیشترین بازی بدون باخت پیاپی متعلق به کدام باش...   \n",
              "2  باشگاه فوتبال آرسنال موفق به کسب چند عنوان قهر...   \n",
              "3  باشگاه فوتبال آرسنال چند عنوان قهرمانی در جام ...   \n",
              "4  باشگاه فوتبال آرسنال چند قهرمانی در جام اتحادی...   \n",
              "\n",
              "                                             context  \\\n",
              "0  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "1  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "2  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "3  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "4  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "\n",
              "                                             answers  \n",
              "0  {'text': '۱۴', 'answer_start': 173, 'answer_en...  \n",
              "1  {'text': 'باشگاه فوتبال انگلیسی', 'answer_star...  \n",
              "2  {'text': '۱۳', 'answer_start': 119, 'answer_en...  \n",
              "3  {'text': '۱۶', 'answer_start': 214, 'answer_en...  \n",
              "4  {'text': 'دو قهرمانی', 'answer_start': 249, 'a...  "
            ]
          },
          "execution_count": 16,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# Print the length of the training dataset\n",
        "print(\"Length of the training dataset: {}\".format(len(train_df)))\n",
        "\n",
        "# Display the first few rows of the training dataset\n",
        "print(\"Training dataset:\")\n",
        "train_df.head()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 241
        },
        "outputId": "499c4a10-a8e6-4ca4-ec29-bf5298eb0cf6",
        "id": "kap9Fqf0otOg"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Length of the validation dataset: 8258\n",
            "Validation dataset:\n"
          ]
        },
        {
          "data": {
            "text/html": [
              "\n",
              "  <div id=\"df-f02abff1-df9c-41d1-8c44-6fb635e663b2\">\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>question</th>\n",
              "      <th>context</th>\n",
              "      <th>answers</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>آژانس بین‌المللی انرژی اتمی برای حفاظت از بازی...</td>\n",
              "      <td>با توجه به خشونت‌های اخیری که در ریودوژانیرو و...</td>\n",
              "      <td>{'text': 'برای جلوگیری از وقوع حملات تروریستی ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>آژانس بین‌المللی انرژی اتمی برای حفاظت از بازی...</td>\n",
              "      <td>با توجه به خشونت‌های اخیری که در ریودوژانیرو و...</td>\n",
              "      <td>{'text': 'امکانات لازم را در اختیار این کشور ق...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>چرا آژانس بین‌المللی انرژی اتمی امکانات لازم ر...</td>\n",
              "      <td>با توجه به خشونت‌های اخیری که در ریودوژانیرو و...</td>\n",
              "      <td>{'text': 'برای جلوگیری از وقوع حملات تروریستی ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>چرا آژانس بین‌المللی انرژی اتمی امکانات لازم ر...</td>\n",
              "      <td>با توجه به خشونت‌های اخیری که در ریودوژانیرو و...</td>\n",
              "      <td>{'text': 'برای جلوگیری از وقوع حملات تروریستی ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>طبق آمار اوایل ماه دسامبر 2015 چند نفر به تب د...</td>\n",
              "      <td>مشکل دیگر برزیل مقابله با پشه‌ها و برخی بیماری...</td>\n",
              "      <td>{'text': '۱٫۵۸ میلیون نفر', 'answer_start': 10...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-f02abff1-df9c-41d1-8c44-6fb635e663b2')\"\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-f02abff1-df9c-41d1-8c44-6fb635e663b2 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-f02abff1-df9c-41d1-8c44-6fb635e663b2');\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",
              "  "
            ],
            "text/plain": [
              "                                            question  \\\n",
              "0  آژانس بین‌المللی انرژی اتمی برای حفاظت از بازی...   \n",
              "1  آژانس بین‌المللی انرژی اتمی برای حفاظت از بازی...   \n",
              "2  چرا آژانس بین‌المللی انرژی اتمی امکانات لازم ر...   \n",
              "3  چرا آژانس بین‌المللی انرژی اتمی امکانات لازم ر...   \n",
              "4  طبق آمار اوایل ماه دسامبر 2015 چند نفر به تب د...   \n",
              "\n",
              "                                             context  \\\n",
              "0  با توجه به خشونت‌های اخیری که در ریودوژانیرو و...   \n",
              "1  با توجه به خشونت‌های اخیری که در ریودوژانیرو و...   \n",
              "2  با توجه به خشونت‌های اخیری که در ریودوژانیرو و...   \n",
              "3  با توجه به خشونت‌های اخیری که در ریودوژانیرو و...   \n",
              "4  مشکل دیگر برزیل مقابله با پشه‌ها و برخی بیماری...   \n",
              "\n",
              "                                             answers  \n",
              "0  {'text': 'برای جلوگیری از وقوع حملات تروریستی ...  \n",
              "1  {'text': 'امکانات لازم را در اختیار این کشور ق...  \n",
              "2  {'text': 'برای جلوگیری از وقوع حملات تروریستی ...  \n",
              "3  {'text': 'برای جلوگیری از وقوع حملات تروریستی ...  \n",
              "4  {'text': '۱٫۵۸ میلیون نفر', 'answer_start': 10...  "
            ]
          },
          "execution_count": 17,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# Print the length of the validation dataset\n",
        "print(\"Length of the validation dataset: {}\".format(len(validation_df)))\n",
        "\n",
        "# Display the first few rows of the validation dataset\n",
        "print(\"Validation dataset:\")\n",
        "validation_df.head()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 241
        },
        "outputId": "2a120d21-b703-4e20-dd0b-e08808f4adfb",
        "id": "XPahiBR_otOh"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Length of the test dataset: 8367\n",
            "Test dataset:\n"
          ]
        },
        {
          "data": {
            "text/html": [
              "\n",
              "  <div id=\"df-2b09633f-6cf6-480c-a732-772f74217c87\">\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>question</th>\n",
              "      <th>context</th>\n",
              "      <th>answers</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>چند اسم از اسماء قرآن در خود قرآن آمده‌است؟</td>\n",
              "      <td>به نوشته محمد صادقی تهرانی، اسامی قرآن، از جم...</td>\n",
              "      <td>{'text': 'حدود چهل اسم', 'answer_start': 582, ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>چند اسم از اسماء قرآن در خود قرآن آمده‌است؟</td>\n",
              "      <td>به نوشته محمد صادقی تهرانی، اسامی قرآن، از جم...</td>\n",
              "      <td>{'text': 'حدود چهل اسم در قرآن آمده', 'answe...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>چند اسم از اسماء قرآن در خود قرآن آمده‌است؟</td>\n",
              "      <td>به نوشته محمد صادقی تهرانی، اسامی قرآن، از جم...</td>\n",
              "      <td>{'text': 'چهل اسم', 'answer_start': 587, 'answ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>از جمله امتیازات قرآن چیست؟</td>\n",
              "      <td>به نوشته محمد صادقی تهرانی، اسامی قرآن، از جم...</td>\n",
              "      <td>{'text': 'این است که چه اسم‌های خود قرآن و چه...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>از جمله امتیازات قرآن چیست؟</td>\n",
              "      <td>به نوشته محمد صادقی تهرانی، اسامی قرآن، از جم...</td>\n",
              "      <td>{'text': 'چه اسم‌های خود قرآن و چه اسم‌هایی ک...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-2b09633f-6cf6-480c-a732-772f74217c87')\"\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-2b09633f-6cf6-480c-a732-772f74217c87 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-2b09633f-6cf6-480c-a732-772f74217c87');\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",
              "  "
            ],
            "text/plain": [
              "                                       question  \\\n",
              "0  چند اسم از اسماء قرآن در خود قرآن آمده‌است؟   \n",
              "1  چند اسم از اسماء قرآن در خود قرآن آمده‌است؟   \n",
              "2  چند اسم از اسماء قرآن در خود قرآن آمده‌است؟   \n",
              "3                   از جمله امتیازات قرآن چیست؟   \n",
              "4                   از جمله امتیازات قرآن چیست؟   \n",
              "\n",
              "                                             context  \\\n",
              "0  به نوشته محمد صادقی تهرانی، اسامی قرآن، از جم...   \n",
              "1  به نوشته محمد صادقی تهرانی، اسامی قرآن، از جم...   \n",
              "2  به نوشته محمد صادقی تهرانی، اسامی قرآن، از جم...   \n",
              "3  به نوشته محمد صادقی تهرانی، اسامی قرآن، از جم...   \n",
              "4  به نوشته محمد صادقی تهرانی، اسامی قرآن، از جم...   \n",
              "\n",
              "                                             answers  \n",
              "0  {'text': 'حدود چهل اسم', 'answer_start': 582, ...  \n",
              "1  {'text': 'حدود چهل اسم در قرآن آمده', 'answe...  \n",
              "2  {'text': 'چهل اسم', 'answer_start': 587, 'answ...  \n",
              "3  {'text': 'این است که چه اسم‌های خود قرآن و چه...  \n",
              "4  {'text': 'چه اسم‌های خود قرآن و چه اسم‌هایی ک...  "
            ]
          },
          "execution_count": 18,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# Print the length of the test dataset\n",
        "print(\"Length of the test dataset: {}\".format(len(test_df)))\n",
        "\n",
        "# Display the first few rows of the test dataset\n",
        "print(\"Test dataset:\")\n",
        "test_df.head()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "lj0ySqKcotOi"
      },
      "outputs": [],
      "source": [
        "def prepare_data_2(dataset):\n",
        "    # Extract the questions, contexts, and answers from the dataset\n",
        "    questions = dataset['question'].tolist()\n",
        "    contexts = dataset['context'].tolist()\n",
        "    answers = dataset['answers'].tolist()\n",
        "    \n",
        "    # Return a dictionary containing the prepared data\n",
        "    return {\n",
        "        'question': questions,\n",
        "        'context': contexts,\n",
        "        'answers': answers\n",
        "    }\n",
        "\n",
        "# Prepare the training dataset\n",
        "train_dataset = prepare_data_2(train_df)\n",
        "\n",
        "# Prepare the validation dataset\n",
        "val_dataset = prepare_data_2(validation_df)\n",
        "\n",
        "# Prepare the test dataset\n",
        "test_dataset = prepare_data_2(test_df)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "background_save": true
        },
        "id": "ztEJQmSEotOj"
      },
      "outputs": [],
      "source": [
        "# Tokenize the training data using the tokenizer\n",
        "train_data = tokenizer(train_dataset['context'], train_dataset['question'],\n",
        "                       truncation=False, padding='max_length',\n",
        "                       max_length=max_len, return_tensors='pt')\n",
        "\n",
        "# Tokenize the validation data using the tokenizer\n",
        "validation_data = tokenizer(val_dataset['context'], val_dataset['question'],\n",
        "                            truncation=False, padding='max_length',\n",
        "                            max_length=max_len, return_tensors='pt')\n",
        "\n",
        "# Tokenize the test data using the tokenizer\n",
        "test_data = tokenizer(test_dataset['context'], test_dataset['question'],\n",
        "                            truncation=False, padding='max_length',\n",
        "                            max_length=max_len, return_tensors='pt')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b127eeda-5873-48b6-c941-0979dcf883f1",
        "id": "uZl8ZFlIotOj"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 51797/51797 [03:17<00:00, 261.71it/s]\n",
            "100%|██████████| 8258/8258 [00:32<00:00, 256.24it/s]\n",
            "100%|██████████| 8367/8367 [00:35<00:00, 237.71it/s]\n"
          ]
        }
      ],
      "source": [
        "# Function to convert character index to token index\n",
        "def char_idx_to_token_idx(tokenizer, char_idx, sentence):\n",
        "    # Create a list of binary values indicating whether each character is a space or a half-space\n",
        "    char = [0 if sentence[i] == ' ' or sentence[i] == '‌' else 1 for i in range(len(sentence))]\n",
        "    tokens = tokenizer.tokenize(sentence)\n",
        "    index = char_idx\n",
        "\n",
        "    # Adjust the index to consider half-spaces\n",
        "    for i in range(index):\n",
        "        if char[i] != 1:\n",
        "            index -= 1\n",
        "\n",
        "    counter = 0\n",
        "    flag_continue = True\n",
        "    token_index = 0\n",
        "\n",
        "    # Find the token index corresponding to the character index\n",
        "    for i in range(len(tokens)):\n",
        "        if tokens[i].startswith('##'):\n",
        "            tokens[i] = tokens[i][2:]\n",
        "        for j in range(len(tokens[i])):\n",
        "            counter += 1\n",
        "            if counter == index:\n",
        "                flag_continue = False\n",
        "                token_index = i\n",
        "                break\n",
        "        if not flag_continue:\n",
        "            break\n",
        "    \n",
        "    return token_index             \n",
        "\n",
        "# Function to add token positions for answer start and end\n",
        "def add_token_positions(tokenizer, encodings, answers, contexts):\n",
        "    target_start_list = []\n",
        "    target_end_list = []\n",
        "    for i in tqdm(range(len(answers))):\n",
        "        target_start = [0] * max_len\n",
        "        target_end = [0] * max_len\n",
        "        start_idx = answers[i]['answer_start']\n",
        "        end_idx = answers[i]['answer_end']\n",
        "        \n",
        "        if start_idx <= len(contexts[i]) and end_idx <= len(contexts[i]):\n",
        "            # Answerable question\n",
        "            if start_idx != -1 and end_idx != -1:\n",
        "                start_token_idx = char_idx_to_token_idx(tokenizer, start_idx, contexts[i]) + 1\n",
        "                end_token_idx = char_idx_to_token_idx(tokenizer, end_idx, contexts[i]) + 1\n",
        "                target_start[start_token_idx] = 1\n",
        "                target_end[end_token_idx] = 1\n",
        "                target_start_list.append(target_start)\n",
        "                target_end_list.append(target_end)\n",
        "\n",
        "            # Unanswerable question\n",
        "            else:\n",
        "                target_start[0] = 1\n",
        "                target_end[0] = 1\n",
        "                target_start_list.append(target_start)\n",
        "                target_end_list.append(target_end)\n",
        "        else:\n",
        "            continue\n",
        "    \n",
        "    # Update the encodings with the target start and end lists\n",
        "    encodings.update({'targets_start': target_start_list, 'targets_end': target_end_list})\n",
        "\n",
        "# Add token positions to the train_data encodings\n",
        "add_token_positions(tokenizer, train_data, train_dataset['answers'], train_dataset['context'])\n",
        "\n",
        "# Add token positions to the validation_data encodings\n",
        "add_token_positions(tokenizer, validation_data, val_dataset['answers'], val_dataset['context'])\n",
        "\n",
        "# Add token positions to the test_data encodings\n",
        "add_token_positions(tokenizer, test_data, test_dataset['answers'], test_dataset['context'])\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "BpKJ0Hj_otOl"
      },
      "outputs": [],
      "source": [
        "# Custom Dataset class\n",
        "class CustomDataset(torch.utils.data.Dataset):\n",
        "    def __init__(self, data):\n",
        "        \"\"\"\n",
        "        Initialize the CustomDataset.\n",
        "\n",
        "        Args:\n",
        "            data (dict): The data dictionary containing input tensors.\n",
        "        \"\"\"\n",
        "        self.data = data\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        \"\"\"\n",
        "        Get an item from the dataset.\n",
        "\n",
        "        Args:\n",
        "            idx (int): The index of the item.\n",
        "\n",
        "        Returns:\n",
        "            dict: A dictionary containing input tensors.\n",
        "        \"\"\"\n",
        "        return {key: torch.tensor(val[idx]) for key, val in self.data.items()}\n",
        "    \n",
        "    def __len__(self):\n",
        "        \"\"\"\n",
        "        Get the length of the dataset.\n",
        "\n",
        "        Returns:\n",
        "            int: The length of the dataset.\n",
        "        \"\"\"\n",
        "        return len(self.data['input_ids'])\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2b19256b-2f5f-4ddf-f43b-ff75038ff108",
        "id": "eFcPX9wDotOl"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Length of training dataset: 51797\n",
            "Length of validation dataset: 8258\n",
            "Length of test dataset: 8367\n"
          ]
        }
      ],
      "source": [
        "# Create a custom dataset for training\n",
        "train_datas = CustomDataset(train_data)\n",
        "print(\"Length of training dataset: {}\".format(len(train_datas)))\n",
        "\n",
        "# Create a data loader for training\n",
        "train_loader = torch.utils.data.DataLoader(train_datas, batch_size=32, shuffle=True)\n",
        "\n",
        "# Create a custom dataset for validation\n",
        "validation_datas = CustomDataset(validation_data)\n",
        "print(\"Length of validation dataset: {}\".format(len(validation_datas)))\n",
        "\n",
        "# Create a data loader for validation\n",
        "validation_loader = torch.utils.data.DataLoader(validation_datas, batch_size=32, shuffle=False)\n",
        "\n",
        "# Create a custom dataset for test\n",
        "test_datas = CustomDataset(test_data)\n",
        "print(\"Length of test dataset: {}\".format(len(test_datas)))\n",
        "\n",
        "# Create a data loader for test\n",
        "test_loader = torch.utils.data.DataLoader(test_datas, batch_size=32, shuffle=False)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "DrmneAeDotOm"
      },
      "outputs": [],
      "source": [
        "MODEL_NAME_OR_PATH = 'HooshvareLab/bert-base-parsbert-uncased'\n",
        "\n",
        "class QAModel(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(QAModel, self).__init__()\n",
        "\n",
        "        # Initialize the BERT model\n",
        "        self.bert = BertModel.from_pretrained(MODEL_NAME_OR_PATH, return_dict=False)\n",
        "        # Add a linear layer for classification\n",
        "        self.classifier = nn.Linear(768, 2)\n",
        "    \n",
        "    def forward(self, input_ids, attention_mask, token_type_ids):\n",
        "        # Pass the input through the BERT model\n",
        "        sequence_output, pooled_output = self.bert(\n",
        "            input_ids=input_ids, \n",
        "            attention_mask=attention_mask, \n",
        "            token_type_ids=token_type_ids)\n",
        "        \n",
        "        # Apply linear layer to the BERT output\n",
        "        # Shape: (batch_size, num_tokens, 768)\n",
        "        logits = self.classifier(sequence_output)\n",
        "        # Shape: (batch_size, num_tokens, 2)\n",
        "        \n",
        "        # Split the logits into start and end logits\n",
        "        start_logits, end_logits = logits.split(1, dim=-1)\n",
        "        start_logits = start_logits.squeeze(-1)\n",
        "        end_logits = end_logits.squeeze(-1)\n",
        "        # Shape: (batch_size, num_tokens), (batch_size, num_tokens)\n",
        "\n",
        "        return start_logits, end_logits\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "wi9FDruaotOn"
      },
      "outputs": [],
      "source": [
        "def loss_fn(start_logits, end_logits, start_targets, end_targets):\n",
        "    \"\"\"\n",
        "    Compute the loss function given the predicted start and end logits and the target start and end positions.\n",
        "\n",
        "    Args:\n",
        "        start_logits (torch.Tensor): Predicted start logits of shape (batch_size, num_tokens).\n",
        "        end_logits (torch.Tensor): Predicted end logits of shape (batch_size, num_tokens).\n",
        "        start_targets (torch.Tensor): Target start positions of shape (batch_size, num_tokens).\n",
        "        end_targets (torch.Tensor): Target end positions of shape (batch_size, num_tokens).\n",
        "\n",
        "    Returns:\n",
        "        torch.Tensor: Loss value.\n",
        "    \"\"\"\n",
        "    # Compute the binary cross-entropy loss for start and end logits\n",
        "    start_loss = nn.BCEWithLogitsLoss()(start_logits, start_targets.float())\n",
        "    end_loss = nn.BCEWithLogitsLoss()(end_logits, end_targets.float())\n",
        "\n",
        "    # Return the sum of the two losses\n",
        "    return start_loss + end_loss\n",
        "\n",
        "def evaluate_f1(start_pred, start_target, end_pred, end_target):\n",
        "    \"\"\"\n",
        "    Compute the F1 score given the predicted start and end positions and the target start and end positions.\n",
        "\n",
        "    Args:\n",
        "        start_pred (int): Predicted start position.\n",
        "        start_target (int): Target start position.\n",
        "        end_pred (int): Predicted end position.\n",
        "        end_target (int): Target end position.\n",
        "\n",
        "    Returns:\n",
        "        float: F1 score.\n",
        "    \"\"\"\n",
        "    # Generate arrays of tokens for prediction and target spans\n",
        "    pred = np.arange(start_pred, end_pred + 1)\n",
        "    tar = np.arange(start_target, end_target + 1)\n",
        "\n",
        "    # Compute the number of tokens shared between prediction and target\n",
        "    tp_list = list(set.intersection(*map(set, [pred, tar])))\n",
        "\n",
        "    # Compute the number of tokens in prediction not in target\n",
        "    fp_list = list(set(pred).symmetric_difference(set(tp_list)))\n",
        "\n",
        "    # Compute the number of tokens in target not in prediction\n",
        "    fn_list = list(set(tar).symmetric_difference(set(tp_list)))\n",
        "\n",
        "    tp, fp, fn = len(tp_list), len(fp_list), len(fn_list)\n",
        "\n",
        "    # Compute precision, recall, and F1 score\n",
        "    if (tp + fp) != 0:\n",
        "        precision = tp / (tp + fp)\n",
        "    else:\n",
        "        precision = 0\n",
        "\n",
        "    if (tp + fn) != 0:\n",
        "        recall = tp / (tp + fn)\n",
        "    else:\n",
        "        recall = 0\n",
        "\n",
        "    if (precision + recall) != 0:\n",
        "        f1 = (2 * precision * recall) / (precision + recall)\n",
        "    else:\n",
        "        f1 = 0\n",
        "\n",
        "    return f1\n",
        "\n",
        "\n",
        "def generate_indexes(start_logits, end_logits, N, max_index_list):\n",
        "    \"\"\"\n",
        "    Generate the start and end indexes for the predicted spans.\n",
        "\n",
        "    Args:\n",
        "        start_logits (numpy.ndarray): Predicted start logits of shape (batch_size, num_tokens).\n",
        "        end_logits (numpy.ndarray): Predicted end logits of shape (batch_size, num_tokens).\n",
        "        N (int): Number of top start and end indexes to consider.\n",
        "        max_index_list (list): List of maximum indexes for each example.\n",
        "\n",
        "    Returns:\n",
        "        tuple: Final start and end indexes for the predicted spans.\n",
        "    \"\"\"\n",
        "    output_start = start_logits\n",
        "    output_end = end_logits\n",
        "    dimension = output_start.shape[1]\n",
        "\n",
        "    list_start, list_end = [], []\n",
        "    for n in range(output_start.shape[0]):\n",
        "        start_indexes = np.arange(output_start.shape[1])\n",
        "        start_probs = output_start[n]\n",
        "        list_start.append(dict(zip(start_indexes, start_probs)))\n",
        "\n",
        "        end_indexes = np.arange(output_start.shape[1])\n",
        "        end_probs = output_end[n]\n",
        "        list_end.append(dict(zip(end_indexes, end_probs)))\n",
        "\n",
        "    sorted_start_list, sorted_end_list = [], []\n",
        "    for j in range(len(list_start)):\n",
        "        sort_start_probs = sorted(list_start[j].items(), key=lambda x: x[1], reverse=True)\n",
        "        sort_end_probs = sorted(list_end[j].items(), key=lambda x: x[1], reverse=True)\n",
        "        sorted_start_list.append(sort_start_probs)\n",
        "        sorted_end_list.append(sort_end_probs)\n",
        "\n",
        "    final_start_idx, final_end_idx = [], []\n",
        "\n",
        "    for c in range(len(list_start)):\n",
        "        start_idx, end_idx, prob = 0, 0, 0\n",
        "        for a in range(N):\n",
        "            for b in range(N):\n",
        "                if (sorted_start_list[c][a][1] + sorted_end_list[c][b][1]) > prob:\n",
        "                    if (sorted_start_list[c][a][0] <= sorted_end_list[c][b][0]) and (\n",
        "                            sorted_end_list[c][b][0] < max_index_list[c]):\n",
        "                        prob = sorted_start_list[c][a][1] + sorted_end_list[c][b][1]\n",
        "                        start_idx = sorted_start_list[c][a][0]\n",
        "                        end_idx = sorted_end_list[c][b][0]\n",
        "        final_start_idx.append(start_idx)\n",
        "        final_end_idx.append(end_idx)\n",
        "\n",
        "    return final_start_idx, final_end_idx\n",
        "\n",
        "\n",
        "def evaluate_model(start_logits, end_logits, N, max_index_list, target_start, target_end):\n",
        "    \"\"\"\n",
        "    Evaluate the model by computing the F1 score.\n",
        "\n",
        "    Args:\n",
        "        start_logits (numpy.ndarray): Predicted start logits of shape (batch_size, num_tokens).\n",
        "        end_logits (numpy.ndarray): Predicted end logits of shape (batch_size, num_tokens).\n",
        "        N (int): Number of top start and end indexes to consider.\n",
        "        max_index_list (list): List of maximum indexes for each example.\n",
        "        target_start (list): List of target start positions.\n",
        "        target_end (list): List of target end positions.\n",
        "\n",
        "    Returns:\n",
        "        float: Mean F1 score.\n",
        "    \"\"\"\n",
        "    final_start_idx, final_end_idx = generate_indexes(start_logits, end_logits, N, max_index_list)\n",
        "    f1 = []\n",
        "    for i in range(len(final_start_idx)):\n",
        "        f1.append(evaluate_f1(final_start_idx[i], target_start[i], final_end_idx[i], target_end[i]))\n",
        "\n",
        "    return np.mean(f1)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 138,
          "referenced_widgets": [
            "482d7b981c514225a5a1d6779c862a87",
            "4b5fb58991b64754a3e80ec1c9c2add2",
            "516ae2e29203486e942e26c8e60eb61b",
            "6089d13094cc41d2bcb0fee3095859e3",
            "2cf6903d18c64aefbc72b711fa12fb6e",
            "b0b08c791b3a49c49aa06d582329ad79",
            "a82f9ed43e50457dad00cdca93f525ae",
            "426220c923f14e4b99fc11d3d8bee863",
            "450976558bb24f08bfcea7a69bf2f157",
            "2244f26f4e804eddaaf269a52a9be8d9",
            "7ab236f0d24449bab67e3c1daaae53bc"
          ]
        },
        "outputId": "6dfa04c5-4e25-4367-8c98-26fadc25dcc0",
        "id": "MCJPKNn-otOp"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Device: cuda:0\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading pytorch_model.bin:   0%|          | 0.00/654M [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "482d7b981c514225a5a1d6779c862a87"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Some weights of the model checkpoint at HooshvareLab/bert-base-parsbert-uncased were not used when initializing BertModel: ['cls.predictions.transform.dense.bias', 'cls.predictions.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.weight', 'cls.predictions.decoder.bias', 'cls.seq_relationship.bias', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.dense.weight']\n",
            "- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
            "- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
          ]
        }
      ],
      "source": [
        "import warnings\n",
        "warnings.filterwarnings('ignore')\n",
        "\n",
        "# Clear GPU cache\n",
        "torch.cuda.empty_cache()\n",
        "\n",
        "# Check if CUDA is available and set the device accordingly\n",
        "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
        "print(\"Device:\", device)\n",
        "\n",
        "# Create an instance of the QAModel\n",
        "qa_model = QAModel()\n",
        "qa_model.to(device)  # Move the model to the appropriate device\n",
        "qa_model.train()  # Set the model in training mode\n",
        "\n",
        "# Create an instance of the AdamW optimizer with a learning rate of 5e-5\n",
        "optimizer = AdamW(qa_model.parameters(), lr=5e-5)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "iE3D59ICotOq"
      },
      "outputs": [],
      "source": [
        "MODEL_NAME_OR_PATH = 'HooshvareLab/bert-base-parsbert-uncased'\n",
        "\n",
        "from transformers import PreTrainedModel\n",
        "\n",
        "class QAModel2(PreTrainedModel):\n",
        "    \"\"\"\n",
        "    QA Model based on BERT for sequence classification.\n",
        "    \"\"\"\n",
        "\n",
        "    def __init__(self, config):\n",
        "        super(QAModel2, self).__init__(config)\n",
        "        self.bert = BertModel.from_pretrained(MODEL_NAME_OR_PATH, return_dict=False)\n",
        "        # self.dropout = nn.Dropout(p=dropout_rate)\n",
        "        self.classifier = nn.Linear(config.hidden_size, config.num_labels)\n",
        "\n",
        "    def forward(self, input_ids, attention_mask, token_type_ids):\n",
        "        \"\"\"\n",
        "        Forward pass of the QA model.\n",
        "        \n",
        "        Args:\n",
        "            input_ids (tensor): Input tensor of shape (batch_size, sequence_length).\n",
        "            attention_mask (tensor): Attention mask tensor of shape (batch_size, sequence_length).\n",
        "            token_type_ids (tensor): Token type IDs tensor of shape (batch_size, sequence_length).\n",
        "        \n",
        "        Returns:\n",
        "            start_logits (tensor): Start logits tensor of shape (batch_size, sequence_length).\n",
        "            end_logits (tensor): End logits tensor of shape (batch_size, sequence_length).\n",
        "        \"\"\"\n",
        "        sequence_output, pooled_output = self.bert(\n",
        "            input_ids=input_ids,\n",
        "            attention_mask=attention_mask,\n",
        "            token_type_ids=token_type_ids\n",
        "        )\n",
        "        \n",
        "        # Apply dropout to the sequence output\n",
        "        # sequence_output = self.dropout(sequence_output)\n",
        "\n",
        "        # (batch_size, num_tokens, 768)\n",
        "        logits = self.classifier(sequence_output)\n",
        "        # (batch_size, num_tokens, 2)\n",
        "        \n",
        "        # Split the logits into start and end logits\n",
        "        start_logits, end_logits = logits.split(1, dim=-1)\n",
        "        start_logits = start_logits.squeeze(-1)\n",
        "        end_logits = end_logits.squeeze(-1)\n",
        "        # (batch_size, num_tokens), (batch_size, num_tokens)\n",
        "\n",
        "        return start_logits, end_logits\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "34a47723-390f-47c3-fcd1-3281221c0435",
        "id": "2HX-gpJ1otOr"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Some weights of the model checkpoint at HooshvareLab/bert-base-parsbert-uncased were not used when initializing BertModel: ['cls.predictions.transform.dense.bias', 'cls.predictions.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.weight', 'cls.predictions.decoder.bias', 'cls.seq_relationship.bias', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.dense.weight']\n",
            "- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
            "- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
          ]
        }
      ],
      "source": [
        "from transformers import AutoConfig, AutoTokenizer\n",
        "\n",
        "# Clear GPU cache\n",
        "torch.cuda.empty_cache()\n",
        "\n",
        "# Determine the device (GPU or CPU)\n",
        "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
        "\n",
        "# Load the configuration for the BERT model\n",
        "config = AutoConfig.from_pretrained(\"HooshvareLab/bert-base-parsbert-uncased\")\n",
        "\n",
        "# Create an instance of the QAModel2 using the loaded configuration\n",
        "model = QAModel2(config)\n",
        "\n",
        "# Move the model to the appropriate device (GPU or CPU)\n",
        "model.to(device)\n",
        "\n",
        "# Set the model to training mode\n",
        "model.train()\n",
        "\n",
        "# Define the optimizer\n",
        "optim = AdamW(model.parameters(), lr=5e-5)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ce9a22c8-e167-4db3-fc3b-bd225ec81a82",
        "id": "MmwW5PfUotOr"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 1 - Training: 100%|██████████| 1619/1619 [35:16<00:00,  1.31s/it]\n",
            "Epoch 1 - Validation: 100%|██████████| 259/259 [10:53<00:00,  2.52s/it]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model Saved in Google Drive!\n",
            "Epoch 1: Train Loss: 0.01552, Valid Loss: 0.01891, Exact Match: 32.72%, F1: 70.76%\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 2 - Training: 100%|██████████| 1619/1619 [35:16<00:00,  1.31s/it]\n",
            "Epoch 2 - Validation: 100%|██████████| 259/259 [10:41<00:00,  2.48s/it]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 2: Train Loss: 0.01209, Valid Loss: 0.01946, Exact Match: 34.74%, F1: 72.29%\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 3 - Training: 100%|██████████| 1619/1619 [35:17<00:00,  1.31s/it]\n",
            "Epoch 3 - Validation: 100%|██████████| 259/259 [10:46<00:00,  2.50s/it]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 3: Train Loss: 0.00930, Valid Loss: 0.02048, Exact Match: 34.31%, F1: 71.04%\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 4 - Training: 100%|██████████| 1619/1619 [35:17<00:00,  1.31s/it]\n",
            "Epoch 4 - Validation: 100%|██████████| 259/259 [10:56<00:00,  2.53s/it]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 4: Train Loss: 0.00718, Valid Loss: 0.02417, Exact Match: 34.58%, F1: 72.34%\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 5 - Training: 100%|██████████| 1619/1619 [35:17<00:00,  1.31s/it]\n",
            "Epoch 5 - Validation: 100%|██████████| 259/259 [10:56<00:00,  2.53s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 5: Train Loss: 0.00576, Valid Loss: 0.02588, Exact Match: 33.40%, F1: 71.34%\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\n"
          ]
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Set the number of epochs\n",
        "n_epochs = 5\n",
        "\n",
        "# Get the number of batches in the training and validation sets\n",
        "n_train_batches = len(train_loader)\n",
        "n_validation_batches = len(validation_loader)\n",
        "\n",
        "# Initialize softmax function\n",
        "softmax = torch.nn.Softmax(dim=1)\n",
        "\n",
        "# Initialize best validation loss\n",
        "best_valid_loss = float('inf')\n",
        "\n",
        "# Initialize lists to store loss, EM, and F1 scores for each epoch\n",
        "train_losses, valid_losses = [], []\n",
        "exact_match_scores, f1_scores = [], []\n",
        "\n",
        "# Main training loop\n",
        "for epoch in range(n_epochs):\n",
        "  \n",
        "    loop = tqdm(train_loader)\n",
        "    train_running_loss, validation_running_loss = 0.0, 0.0\n",
        "    exact_match = []\n",
        "    f1 = []\n",
        "    \n",
        "    # Training phase\n",
        "    for batch in loop:\n",
        "       \n",
        "        # Zero the gradients\n",
        "        optim.zero_grad()\n",
        "        \n",
        "        # Move data to device\n",
        "        input_ids = batch['input_ids'].to(device)\n",
        "        attention_mask = batch['attention_mask'].to(device)\n",
        "        token_type_ids = batch['token_type_ids'].to(device)\n",
        "        y_start = batch['targets_start'].to(device)\n",
        "        y_end = batch['targets_end'].to(device)\n",
        "        \n",
        "        # Forward pass\n",
        "        out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)  \n",
        "        \n",
        "        # Compute loss\n",
        "        loss = loss_fn(out_start, out_end, y_start, y_end)\n",
        "        train_running_loss += loss\n",
        "        \n",
        "        # Backward pass\n",
        "        loss.backward()\n",
        "        \n",
        "        # Update weights\n",
        "        optim.step()\n",
        "        \n",
        "        loop.set_description(f'Epoch {epoch+1} - Training')\n",
        "    \n",
        "    # Validation phase\n",
        "    with torch.no_grad():  \n",
        "        loop2 = tqdm(validation_loader)\n",
        "        for content in loop2:\n",
        "            \n",
        "            # Move data to device\n",
        "            input_ids = content['input_ids'].to(device)\n",
        "            temp_ids = input_ids.cpu().data.numpy().tolist()\n",
        "            max_ind = [temp_ids[i].index(4) for i in range(0, len(temp_ids))]  # index of first [sep]\n",
        "            attention_mask = content['attention_mask'].to(device)\n",
        "            token_type_ids = content['token_type_ids'].to(device)\n",
        "            y_start = content['targets_start'].to(device)\n",
        "            y_end = content['targets_end'].to(device)\n",
        "            \n",
        "            # Forward pass\n",
        "            out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)\n",
        "            start_pred = softmax(out_start)\n",
        "            end_pred = softmax(out_end)\n",
        "            \n",
        "            # Compute loss\n",
        "            loss2 = loss_fn(out_start, out_end, y_start, y_end)\n",
        "            validation_running_loss += loss2\n",
        "            \n",
        "            # Compute EM and F1 scores\n",
        "            start_label = np.nonzero(y_start).cpu().data.numpy()\n",
        "            end_label = np.nonzero(y_end).cpu().data.numpy()\n",
        "            tensor_label = torch.stack((torch.tensor(start_label[:, 1]), torch.tensor(end_label[:, 1])), -1)\n",
        "            start_log, end_log = generate_indexes(start_pred, end_pred, N, max_ind)\n",
        "            start_log = np.array(start_log)\n",
        "            end_log = np.array(end_log)\n",
        "            tensor_pred = torch.stack((torch.tensor(start_log), torch.tensor(end_log)), -1)\n",
        "            ex_ma = sum([1 if (tensor_pred[i][0] == tensor_label[i][0] and tensor_pred[i][1] == tensor_label[i][1]) else 0 for i in range(0, len(tensor_pred))]) / len(tensor_pred)\n",
        "            exact_match.append(ex_ma)\n",
        "            f1.append(evaluate_model(start_pred.cpu().data.numpy(), end_pred.cpu().data.numpy(), N, max_ind, start_label[:, 1], end_label[:, 1]))\n",
        "            loop2.set_description(f'Epoch {epoch+1} - Validation')\n",
        "\n",
        "    # Compute average losses and scores\n",
        "    train_loss = train_running_loss / n_train_batches\n",
        "    valid_loss = validation_running_loss / n_validation_batches\n",
        "    \n",
        "    # Append losses to the lists\n",
        "    train_losses.append(train_loss)\n",
        "    valid_losses.append(valid_loss)\n",
        "    \n",
        "    # Update best validation loss and save model weights\n",
        "    if valid_loss < best_valid_loss:\n",
        "        best_valid_loss = valid_loss\n",
        "        torch.save(model.state_dict(), '/content/drive/MyDrive/HW5/Q2/Models/Model1.pt')\n",
        "        print('Model Saved in Google Drive!')\n",
        "    \n",
        "    # Compute and append average EM and F1 scores\n",
        "    exact_match_score = 100 * np.mean(exact_match)\n",
        "    f1_score = 100 * np.mean(f1)\n",
        "    exact_match_scores.append(exact_match_score)\n",
        "    f1_scores.append(f1_score)\n",
        "    \n",
        "    # Print epoch statistics\n",
        "    print('Epoch {}: Train Loss: {:.5f}, Valid Loss: {:.5f}, Exact Match: {:.2f}%, F1: {:.2f}%'.format(epoch + 1,\n",
        "                                                                                                     train_loss,\n",
        "                                                                                                     valid_loss,\n",
        "                                                                                                     exact_match_score,\n",
        "                                                                                                     f1_score))\n",
        "\n",
        "# # Plotting the results\n",
        "# epochs = range(1, n_epochs + 1)\n",
        "\n",
        "# # Plotting training and validation losses\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, train_losses, label='Train')\n",
        "# plt.plot(epochs, valid_losses, label='Validation')\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('Loss')\n",
        "# plt.title('Training and Validation Loss')\n",
        "# plt.legend()\n",
        "# plt.savefig('loss_plot.pdf')\n",
        "# plt.show()\n",
        "\n",
        "# # Plotting Exact Match scores\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, exact_match_scores)\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('Exact Match (%)')\n",
        "# plt.title('Exact Match Score')\n",
        "# plt.savefig('exact_match_plot.pdf')\n",
        "# plt.show()\n",
        "\n",
        "# # Plotting F1 scores\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, f1_scores)\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('F1 Score (%)')\n",
        "# plt.title('F1 Score')\n",
        "# plt.savefig('f1_score_plot.pdf')\n",
        "# plt.show()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# 1.4. ParsBERT (Evaluating)"
      ],
      "metadata": {
        "id": "Q7ICrFQIprSr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Softmax function for probability calculation\n",
        "softmax = torch.nn.Softmax(dim=1)\n",
        "\n",
        "# Set the model in evaluation mode\n",
        "model.eval()\n",
        "\n",
        "# Initialize variables for loss and evaluation metrics\n",
        "n_test_batches = len(test_loader)\n",
        "loss = 0\n",
        "exact_match = []\n",
        "f1_scores = []\n",
        "\n",
        "# Iterate over the test data\n",
        "with torch.no_grad():\n",
        "    loop = tqdm(test_loader)\n",
        "    for content in loop:\n",
        "        # Move input tensors to the appropriate device\n",
        "        input_ids = content['input_ids'].to(device)\n",
        "        temp_ids = input_ids.cpu().data.numpy().tolist()\n",
        "        max_ind = [temp_ids[i].index(4) for i in range(0, len(temp_ids))]  # index of first [SEP]\n",
        "        attention_mask = content['attention_mask'].to(device)\n",
        "        token_type_ids = content['token_type_ids'].to(device)\n",
        "        y_start = content['targets_start'].to(device)\n",
        "        y_end = content['targets_end'].to(device)\n",
        "\n",
        "        # Forward pass\n",
        "        out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)\n",
        "        start_pred = softmax(out_start)\n",
        "        end_pred = softmax(out_end)\n",
        "\n",
        "        # Calculate the loss\n",
        "        loss_batch = loss_fn(out_start, out_end, y_start, y_end)\n",
        "        loss += loss_batch.item()\n",
        "\n",
        "        # Convert labels and predictions to numpy arrays\n",
        "        start_label = np.nonzero(y_start).cpu().data.numpy()\n",
        "        end_label = np.nonzero(y_end).cpu().data.numpy()\n",
        "        tensor_label = torch.stack((torch.tensor(start_label[:, 1]), torch.tensor(end_label[:, 1])), -1)\n",
        "        start_log, end_log = generate_indexes(start_pred, end_pred, N, max_ind)\n",
        "        start_log = np.array(start_log)\n",
        "        end_log = np.array(end_log)\n",
        "        tensor_pred = torch.stack((torch.tensor(start_log), torch.tensor(end_log)), -1)\n",
        "\n",
        "        # Calculate exact match and F1 scores\n",
        "        ex_ma = sum([1 if (tensor_pred[i][0] == tensor_label[i][0] and tensor_pred[i][1] == tensor_label[i][1]) else 0\n",
        "                     for i in range(0, len(tensor_pred))]) / len(tensor_pred)\n",
        "        exact_match.append(ex_ma)\n",
        "\n",
        "        f1_score = evaluate_model(start_pred.cpu().data.numpy(), end_pred.cpu().data.numpy(),\n",
        "                                  N, max_ind, start_label[:, 1], end_label[:, 1])\n",
        "        f1_scores.append(f1_score)\n",
        "\n",
        "    # Calculate average loss and metrics\n",
        "    valid_loss = loss / n_test_batches\n",
        "    em_accuracy = 100 * np.mean(exact_match)\n",
        "    f1_mean = 100 * np.mean(f1_scores)\n",
        "    print('Test Loss: {:.4f}, Exact Match: {:.2f}%, F1 Score: {:.2f}%'.format(valid_loss, em_accuracy, f1_mean))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "dcc71c0f-95f2-4d14-f3ff-1eced39707a8",
        "id": "hBoJV_sgotOs"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 262/262 [10:45<00:00,  2.46s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Test Loss: 0.0265, Exact Match: 34.08%, F1 Score: 71.68%\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Drafts!"
      ],
      "metadata": {
        "id": "SmakCZrjpLxA"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e293d12b-eba7-489e-f40d-8d86c48e118e",
        "id": "1pyEgxWs7Ino"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Some weights of the model checkpoint at HooshvareLab/bert-base-parsbert-uncased were not used when initializing BertModel: ['cls.predictions.transform.dense.bias', 'cls.predictions.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.weight', 'cls.predictions.decoder.bias', 'cls.seq_relationship.bias', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.dense.weight']\n",
            "- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
            "- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
          ]
        }
      ],
      "source": [
        "from transformers import AutoConfig, AutoTokenizer\n",
        "\n",
        "# Clear GPU cache\n",
        "torch.cuda.empty_cache()\n",
        "\n",
        "# Determine the device (GPU or CPU)\n",
        "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
        "\n",
        "# Load the configuration for the BERT model\n",
        "config = AutoConfig.from_pretrained(\"HooshvareLab/bert-base-parsbert-uncased\")\n",
        "\n",
        "# Create an instance of the QAModel2 using the loaded configuration\n",
        "model = QAModel2(config)\n",
        "\n",
        "# Move the model to the appropriate device (GPU or CPU)\n",
        "model.to(device)\n",
        "\n",
        "# Set the model to training mode\n",
        "model.train()\n",
        "\n",
        "# Define the optimizer\n",
        "optim = AdamW(model.parameters(), lr=5e-6)\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Set the number of epochs\n",
        "n_epochs = 5\n",
        "\n",
        "# Get the number of batches in the training and validation sets\n",
        "n_train_batches = len(train_loader)\n",
        "n_validation_batches = len(validation_loader)\n",
        "\n",
        "# Initialize softmax function\n",
        "softmax = torch.nn.Softmax(dim=1)\n",
        "\n",
        "# Initialize best validation loss\n",
        "best_valid_loss = float('inf')\n",
        "\n",
        "# Initialize lists to store loss, EM, and F1 scores for each epoch\n",
        "train_losses, valid_losses = [], []\n",
        "exact_match_scores, f1_scores = [], []\n",
        "\n",
        "# Main training loop\n",
        "for epoch in range(n_epochs):\n",
        "  \n",
        "    loop = tqdm(train_loader)\n",
        "    train_running_loss, validation_running_loss = 0.0, 0.0\n",
        "    exact_match = []\n",
        "    f1 = []\n",
        "    \n",
        "    # Training phase\n",
        "    for batch in loop:\n",
        "       \n",
        "        # Zero the gradients\n",
        "        optim.zero_grad()\n",
        "        \n",
        "        # Move data to device\n",
        "        input_ids = batch['input_ids'].to(device)\n",
        "        attention_mask = batch['attention_mask'].to(device)\n",
        "        token_type_ids = batch['token_type_ids'].to(device)\n",
        "        y_start = batch['targets_start'].to(device)\n",
        "        y_end = batch['targets_end'].to(device)\n",
        "        \n",
        "        # Forward pass\n",
        "        out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)  \n",
        "        \n",
        "        # Compute loss\n",
        "        loss = loss_fn(out_start, out_end, y_start, y_end)\n",
        "        train_running_loss += loss\n",
        "        \n",
        "        # Backward pass\n",
        "        loss.backward()\n",
        "        \n",
        "        # Update weights\n",
        "        optim.step()\n",
        "        \n",
        "        loop.set_description(f'Epoch {epoch+1} - Training')\n",
        "    \n",
        "    # Validation phase\n",
        "    with torch.no_grad():  \n",
        "        loop2 = tqdm(validation_loader)\n",
        "        for content in loop2:\n",
        "            \n",
        "            # Move data to device\n",
        "            input_ids = content['input_ids'].to(device)\n",
        "            temp_ids = input_ids.cpu().data.numpy().tolist()\n",
        "            max_ind = [temp_ids[i].index(4) for i in range(0, len(temp_ids))]  # index of first [sep]\n",
        "            attention_mask = content['attention_mask'].to(device)\n",
        "            token_type_ids = content['token_type_ids'].to(device)\n",
        "            y_start = content['targets_start'].to(device)\n",
        "            y_end = content['targets_end'].to(device)\n",
        "            \n",
        "            # Forward pass\n",
        "            out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)\n",
        "            start_pred = softmax(out_start)\n",
        "            end_pred = softmax(out_end)\n",
        "            \n",
        "            # Compute loss\n",
        "            loss2 = loss_fn(out_start, out_end, y_start, y_end)\n",
        "            validation_running_loss += loss2\n",
        "            \n",
        "            # Compute EM and F1 scores\n",
        "            start_label = np.nonzero(y_start).cpu().data.numpy()\n",
        "            end_label = np.nonzero(y_end).cpu().data.numpy()\n",
        "            tensor_label = torch.stack((torch.tensor(start_label[:, 1]), torch.tensor(end_label[:, 1])), -1)\n",
        "            start_log, end_log = generate_indexes(start_pred, end_pred, N, max_ind)\n",
        "            start_log = np.array(start_log)\n",
        "            end_log = np.array(end_log)\n",
        "            tensor_pred = torch.stack((torch.tensor(start_log), torch.tensor(end_log)), -1)\n",
        "            ex_ma = sum([1 if (tensor_pred[i][0] == tensor_label[i][0] and tensor_pred[i][1] == tensor_label[i][1]) else 0 for i in range(0, len(tensor_pred))]) / len(tensor_pred)\n",
        "            exact_match.append(ex_ma)\n",
        "            f1.append(evaluate_model(start_pred.cpu().data.numpy(), end_pred.cpu().data.numpy(), N, max_ind, start_label[:, 1], end_label[:, 1]))\n",
        "            loop2.set_description(f'Epoch {epoch+1} - Validation')\n",
        "\n",
        "    # Compute average losses and scores\n",
        "    train_loss = train_running_loss / n_train_batches\n",
        "    valid_loss = validation_running_loss / n_validation_batches\n",
        "    \n",
        "    # Append losses to the lists\n",
        "    train_losses.append(train_loss)\n",
        "    valid_losses.append(valid_loss)\n",
        "    \n",
        "    # Update best validation loss and save model weights\n",
        "    if valid_loss < best_valid_loss:\n",
        "        best_valid_loss = valid_loss\n",
        "        torch.save(model.state_dict(), '/content/drive/MyDrive/HW5/Q2/Models/Model2.pt')\n",
        "        print('Model Saved in Google Drive!')\n",
        "    \n",
        "    # Compute and append average EM and F1 scores\n",
        "    exact_match_score = 100 * np.mean(exact_match)\n",
        "    f1_score = 100 * np.mean(f1)\n",
        "    exact_match_scores.append(exact_match_score)\n",
        "    f1_scores.append(f1_score)\n",
        "    \n",
        "    # Print epoch statistics\n",
        "    print('Epoch {}: Train Loss: {:.5f}, Valid Loss: {:.5f}, Exact Match: {:.2f}%, F1: {:.2f}%'.format(epoch + 1,\n",
        "                                                                                                     train_loss,\n",
        "                                                                                                     valid_loss,\n",
        "                                                                                                     exact_match_score,\n",
        "                                                                                                     f1_score))\n",
        "\n",
        "# # Plotting the results\n",
        "# epochs = range(1, n_epochs + 1)\n",
        "\n",
        "# # Plotting training and validation losses\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, train_losses, label='Train')\n",
        "# plt.plot(epochs, valid_losses, label='Validation')\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('Loss')\n",
        "# plt.title('Training and Validation Loss')\n",
        "# plt.legend()\n",
        "# plt.savefig('loss_plot.pdf')\n",
        "# plt.show()\n",
        "\n",
        "# # Plotting Exact Match scores\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, exact_match_scores)\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('Exact Match (%)')\n",
        "# plt.title('Exact Match Score')\n",
        "# plt.savefig('exact_match_plot.pdf')\n",
        "# plt.show()\n",
        "\n",
        "# # Plotting F1 scores\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, f1_scores)\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('F1 Score (%)')\n",
        "# plt.title('F1 Score')\n",
        "# plt.savefig('f1_score_plot.pdf')\n",
        "# plt.show()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uUIBfQniUsZR",
        "outputId": "b9f0da6a-e786-4fc9-af54-e059c5857e60"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 1 - Training: 100%|██████████| 1619/1619 [35:17<00:00,  1.31s/it]\n",
            "Epoch 1 - Validation:  73%|███████▎  | 188/259 [07:40<02:52,  2.43s/it]"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 519
        },
        "outputId": "74c9d87d-5b1d-48f6-eefc-9d0e94917f85",
        "id": "7D77eStI7Inp"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 1 - Training: 100%|██████████| 1619/1619 [35:08<00:00,  1.30s/it]\n",
            "Epoch 1 - Validation: 100%|██████████| 259/259 [11:17<00:00,  2.62s/it]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model Saved in Google Drive!\n",
            "Epoch 1: Train Loss: 0.06850, Valid Loss: 0.05115, Exact Match: 6.01%, F1: 9.31%\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 2 - Training: 100%|██████████| 1619/1619 [35:08<00:00,  1.30s/it]\n",
            "Epoch 2 - Validation: 100%|██████████| 259/259 [11:21<00:00,  2.63s/it]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 2: Train Loss: 0.05115, Valid Loss: 0.05115, Exact Match: 6.01%, F1: 9.30%\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 3 - Training:  42%|████▏     | 676/1619 [14:42<20:30,  1.31s/it]\n"
          ]
        },
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-36-cea36f497118>\u001b[0m in \u001b[0;36m<cell line: 21>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     47\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     48\u001b[0m         \u001b[0;31m# Backward pass\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 49\u001b[0;31m         \u001b[0mloss\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     50\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     51\u001b[0m         \u001b[0;31m# Update weights\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/torch/_tensor.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(self, gradient, retain_graph, create_graph, inputs)\u001b[0m\n\u001b[1;32m    485\u001b[0m                 \u001b[0minputs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    486\u001b[0m             )\n\u001b[0;32m--> 487\u001b[0;31m         torch.autograd.backward(\n\u001b[0m\u001b[1;32m    488\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgradient\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    489\u001b[0m         )\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/torch/autograd/__init__.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(tensors, grad_tensors, retain_graph, create_graph, grad_variables, inputs)\u001b[0m\n\u001b[1;32m    198\u001b[0m     \u001b[0;31m# some Python versions print out the first line of a multi-line function\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    199\u001b[0m     \u001b[0;31m# calls in the traceback and some print out the last line\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 200\u001b[0;31m     Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass\n\u001b[0m\u001b[1;32m    201\u001b[0m         \u001b[0mtensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad_tensors_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    202\u001b[0m         allow_unreachable=True, accumulate_grad=True)  # Calls into the C++ engine to run the backward pass\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "# lr = 0.0005\n",
        "# Set the number of epochs\n",
        "n_epochs = 5\n",
        "\n",
        "# Get the number of batches in the training and validation sets\n",
        "n_train_batches = len(train_loader)\n",
        "n_validation_batches = len(validation_loader)\n",
        "\n",
        "# Initialize softmax function\n",
        "softmax = torch.nn.Softmax(dim=1)\n",
        "\n",
        "# Initialize best validation loss\n",
        "best_valid_loss = float('inf')\n",
        "\n",
        "# Initialize lists to store loss, EM, and F1 scores for each epoch\n",
        "train_losses, valid_losses = [], []\n",
        "exact_match_scores, f1_scores = [], []\n",
        "\n",
        "# Main training loop\n",
        "for epoch in range(n_epochs):\n",
        "  \n",
        "    loop = tqdm(train_loader)\n",
        "    train_running_loss, validation_running_loss = 0.0, 0.0\n",
        "    exact_match = []\n",
        "    f1 = []\n",
        "    \n",
        "    # Training phase\n",
        "    for batch in loop:\n",
        "       \n",
        "        # Zero the gradients\n",
        "        optim.zero_grad()\n",
        "        \n",
        "        # Move data to device\n",
        "        input_ids = batch['input_ids'].to(device)\n",
        "        attention_mask = batch['attention_mask'].to(device)\n",
        "        token_type_ids = batch['token_type_ids'].to(device)\n",
        "        y_start = batch['targets_start'].to(device)\n",
        "        y_end = batch['targets_end'].to(device)\n",
        "        \n",
        "        # Forward pass\n",
        "        out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)  \n",
        "        \n",
        "        # Compute loss\n",
        "        loss = loss_fn(out_start, out_end, y_start, y_end)\n",
        "        train_running_loss += loss\n",
        "        \n",
        "        # Backward pass\n",
        "        loss.backward()\n",
        "        \n",
        "        # Update weights\n",
        "        optim.step()\n",
        "        \n",
        "        loop.set_description(f'Epoch {epoch+1} - Training')\n",
        "    \n",
        "    # Validation phase\n",
        "    with torch.no_grad():  \n",
        "        loop2 = tqdm(validation_loader)\n",
        "        for content in loop2:\n",
        "            \n",
        "            # Move data to device\n",
        "            input_ids = content['input_ids'].to(device)\n",
        "            temp_ids = input_ids.cpu().data.numpy().tolist()\n",
        "            max_ind = [temp_ids[i].index(4) for i in range(0, len(temp_ids))]  # index of first [sep]\n",
        "            attention_mask = content['attention_mask'].to(device)\n",
        "            token_type_ids = content['token_type_ids'].to(device)\n",
        "            y_start = content['targets_start'].to(device)\n",
        "            y_end = content['targets_end'].to(device)\n",
        "            \n",
        "            # Forward pass\n",
        "            out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)\n",
        "            start_pred = softmax(out_start)\n",
        "            end_pred = softmax(out_end)\n",
        "            \n",
        "            # Compute loss\n",
        "            loss2 = loss_fn(out_start, out_end, y_start, y_end)\n",
        "            validation_running_loss += loss2\n",
        "            \n",
        "            # Compute EM and F1 scores\n",
        "            start_label = np.nonzero(y_start).cpu().data.numpy()\n",
        "            end_label = np.nonzero(y_end).cpu().data.numpy()\n",
        "            tensor_label = torch.stack((torch.tensor(start_label[:, 1]), torch.tensor(end_label[:, 1])), -1)\n",
        "            start_log, end_log = generate_indexes(start_pred, end_pred, N, max_ind)\n",
        "            start_log = np.array(start_log)\n",
        "            end_log = np.array(end_log)\n",
        "            tensor_pred = torch.stack((torch.tensor(start_log), torch.tensor(end_log)), -1)\n",
        "            ex_ma = sum([1 if (tensor_pred[i][0] == tensor_label[i][0] and tensor_pred[i][1] == tensor_label[i][1]) else 0 for i in range(0, len(tensor_pred))]) / len(tensor_pred)\n",
        "            exact_match.append(ex_ma)\n",
        "            f1.append(evaluate_model(start_pred.cpu().data.numpy(), end_pred.cpu().data.numpy(), N, max_ind, start_label[:, 1], end_label[:, 1]))\n",
        "            loop2.set_description(f'Epoch {epoch+1} - Validation')\n",
        "\n",
        "    # Compute average losses and scores\n",
        "    train_loss = train_running_loss / n_train_batches\n",
        "    valid_loss = validation_running_loss / n_validation_batches\n",
        "    \n",
        "    # Append losses to the lists\n",
        "    train_losses.append(train_loss)\n",
        "    valid_losses.append(valid_loss)\n",
        "    \n",
        "    # Update best validation loss and save model weights\n",
        "    if valid_loss < best_valid_loss:\n",
        "        best_valid_loss = valid_loss\n",
        "        torch.save(model.state_dict(), '/content/drive/MyDrive/HW5/Q2/Models/Model2.pt')\n",
        "        print('Model Saved in Google Drive!')\n",
        "    \n",
        "    # Compute and append average EM and F1 scores\n",
        "    exact_match_score = 100 * np.mean(exact_match)\n",
        "    f1_score = 100 * np.mean(f1)\n",
        "    exact_match_scores.append(exact_match_score)\n",
        "    f1_scores.append(f1_score)\n",
        "    \n",
        "    # Print epoch statistics\n",
        "    print('Epoch {}: Train Loss: {:.5f}, Valid Loss: {:.5f}, Exact Match: {:.2f}%, F1: {:.2f}%'.format(epoch + 1,\n",
        "                                                                                                     train_loss,\n",
        "                                                                                                     valid_loss,\n",
        "                                                                                                     exact_match_score,\n",
        "                                                                                                     f1_score))\n",
        "\n",
        "# # Plotting the results\n",
        "# epochs = range(1, n_epochs + 1)\n",
        "\n",
        "# # Plotting training and validation losses\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, train_losses, label='Train')\n",
        "# plt.plot(epochs, valid_losses, label='Validation')\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('Loss')\n",
        "# plt.title('Training and Validation Loss')\n",
        "# plt.legend()\n",
        "# plt.savefig('loss_plot.pdf')\n",
        "# plt.show()\n",
        "\n",
        "# # Plotting Exact Match scores\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, exact_match_scores)\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('Exact Match (%)')\n",
        "# plt.title('Exact Match Score')\n",
        "# plt.savefig('exact_match_plot.pdf')\n",
        "# plt.show()\n",
        "\n",
        "# # Plotting F1 scores\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, f1_scores)\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('F1 Score (%)')\n",
        "# plt.title('F1 Score')\n",
        "# plt.savefig('f1_score_plot.pdf')\n",
        "# plt.show()\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Softmax function for probability calculation\n",
        "softmax = torch.nn.Softmax(dim=1)\n",
        "\n",
        "# Set the model in evaluation mode\n",
        "model.eval()\n",
        "\n",
        "# Initialize variables for loss and evaluation metrics\n",
        "n_test_batches = len(test_loader)\n",
        "loss = 0\n",
        "exact_match = []\n",
        "f1_scores = []\n",
        "\n",
        "# Iterate over the test data\n",
        "with torch.no_grad():\n",
        "    loop = tqdm(test_loader)\n",
        "    for content in loop:\n",
        "        # Move input tensors to the appropriate device\n",
        "        input_ids = content['input_ids'].to(device)\n",
        "        temp_ids = input_ids.cpu().data.numpy().tolist()\n",
        "        max_ind = [temp_ids[i].index(4) for i in range(0, len(temp_ids))]  # index of first [SEP]\n",
        "        attention_mask = content['attention_mask'].to(device)\n",
        "        token_type_ids = content['token_type_ids'].to(device)\n",
        "        y_start = content['targets_start'].to(device)\n",
        "        y_end = content['targets_end'].to(device)\n",
        "\n",
        "        # Forward pass\n",
        "        out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)\n",
        "        start_pred = softmax(out_start)\n",
        "        end_pred = softmax(out_end)\n",
        "\n",
        "        # Calculate the loss\n",
        "        loss_batch = loss_fn(out_start, out_end, y_start, y_end)\n",
        "        loss += loss_batch.item()\n",
        "\n",
        "        # Convert labels and predictions to numpy arrays\n",
        "        start_label = np.nonzero(y_start).cpu().data.numpy()\n",
        "        end_label = np.nonzero(y_end).cpu().data.numpy()\n",
        "        tensor_label = torch.stack((torch.tensor(start_label[:, 1]), torch.tensor(end_label[:, 1])), -1)\n",
        "        start_log, end_log = generate_indexes(start_pred, end_pred, N, max_ind)\n",
        "        start_log = np.array(start_log)\n",
        "        end_log = np.array(end_log)\n",
        "        tensor_pred = torch.stack((torch.tensor(start_log), torch.tensor(end_log)), -1)\n",
        "\n",
        "        # Calculate exact match and F1 scores\n",
        "        ex_ma = sum([1 if (tensor_pred[i][0] == tensor_label[i][0] and tensor_pred[i][1] == tensor_label[i][1]) else 0\n",
        "                     for i in range(0, len(tensor_pred))]) / len(tensor_pred)\n",
        "        exact_match.append(ex_ma)\n",
        "\n",
        "        f1_score = evaluate_model(start_pred.cpu().data.numpy(), end_pred.cpu().data.numpy(),\n",
        "                                  N, max_ind, start_label[:, 1], end_label[:, 1])\n",
        "        f1_scores.append(f1_score)\n",
        "\n",
        "    # Calculate average loss and metrics\n",
        "    valid_loss = loss / n_test_batches\n",
        "    em_accuracy = 100 * np.mean(exact_match)\n",
        "    f1_mean = 100 * np.mean(f1_scores)\n",
        "    print('Test Loss: {:.4f}, Exact Match: {:.2f}%, F1 Score: {:.2f}%'.format(valid_loss, em_accuracy, f1_mean))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "dcc71c0f-95f2-4d14-f3ff-1eced39707a8",
        "id": "YavKWTV37Inp"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            " 71%|███████   | 186/262 [07:37<03:13,  2.55s/it]"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 980
        },
        "id": "j9eOUTjDomEe",
        "outputId": "d5e2b868-cbd6-44d7-fb15-b58fd75bcaa7"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 1 - Training: 100%|██████████| 1889/1889 [41:20<00:00,  1.31s/it]\n",
            "Epoch 1 - Validation: 100%|██████████| 265/265 [11:03<00:00,  2.50s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "New Model Weights Saved in Drive!\n",
            "Epoch 1: Train Loss: 0.02233, Valid Loss: 0.01914, Exact Match: 32.71%, F1: 70.84%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 2 - Training: 100%|██████████| 1889/1889 [41:22<00:00,  1.31s/it]\n",
            "Epoch 2 - Validation: 100%|██████████| 265/265 [11:00<00:00,  2.49s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "New Model Weights Saved in Drive!\n",
            "Epoch 2: Train Loss: 0.01456, Valid Loss: 0.01832, Exact Match: 35.02%, F1: 72.59%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 3 - Training: 100%|██████████| 1889/1889 [41:22<00:00,  1.31s/it]\n",
            "Epoch 3 - Validation: 100%|██████████| 265/265 [11:06<00:00,  2.52s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 3: Train Loss: 0.01107, Valid Loss: 0.01980, Exact Match: 35.06%, F1: 71.86%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 4 - Training: 100%|██████████| 1889/1889 [41:22<00:00,  1.31s/it]\n",
            "Epoch 4 - Validation: 100%|██████████| 265/265 [11:12<00:00,  2.54s/it]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 4: Train Loss: 0.00854, Valid Loss: 0.02088, Exact Match: 35.49%, F1: 72.87%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        },
        {
          "ename": "TypeError",
          "evalue": "ignored",
          "output_type": "error",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-34-38cd632ca658>\u001b[0m in \u001b[0;36m<cell line: 122>\u001b[0;34m()\u001b[0m\n\u001b[1;32m    120\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfigure\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfigsize\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m12\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m4\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    121\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msubplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m3\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 122\u001b[0;31m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mn_epochs\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtrain_losses\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlabel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'Train Loss'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    123\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mn_epochs\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvalid_losses\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlabel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'Valid Loss'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    124\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mxlabel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Epoch'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/matplotlib/pyplot.py\u001b[0m in \u001b[0;36mplot\u001b[0;34m(scalex, scaley, data, *args, **kwargs)\u001b[0m\n\u001b[1;32m   2810\u001b[0m \u001b[0;34m@\u001b[0m\u001b[0m_copy_docstring_and_deprecators\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mAxes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2811\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mscalex\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mscaley\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdata\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2812\u001b[0;31m     return gca().plot(\n\u001b[0m\u001b[1;32m   2813\u001b[0m         \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mscalex\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mscalex\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mscaley\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mscaley\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2814\u001b[0m         **({\"data\": data} if data is not None else {}), **kwargs)\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/matplotlib/axes/_axes.py\u001b[0m in \u001b[0;36mplot\u001b[0;34m(self, scalex, scaley, data, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1686\u001b[0m         \"\"\"\n\u001b[1;32m   1687\u001b[0m         \u001b[0mkwargs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcbook\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnormalize_kwargs\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmlines\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mLine2D\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1688\u001b[0;31m         \u001b[0mlines\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_get_lines\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdata\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1689\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mline\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mlines\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1690\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd_line\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mline\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/matplotlib/axes/_base.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, data, *args, **kwargs)\u001b[0m\n\u001b[1;32m    309\u001b[0m                 \u001b[0mthis\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    310\u001b[0m                 \u001b[0margs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 311\u001b[0;31m             yield from self._plot_args(\n\u001b[0m\u001b[1;32m    312\u001b[0m                 this, kwargs, ambiguous_fmt_datakey=ambiguous_fmt_datakey)\n\u001b[1;32m    313\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/matplotlib/axes/_base.py\u001b[0m in \u001b[0;36m_plot_args\u001b[0;34m(self, tup, kwargs, return_kwargs, ambiguous_fmt_datakey)\u001b[0m\n\u001b[1;32m    492\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mxy\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m2\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    493\u001b[0m             \u001b[0mx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_check_1d\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mxy\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 494\u001b[0;31m             \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_check_1d\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mxy\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    495\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    496\u001b[0m             \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mindex_of\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mxy\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/matplotlib/cbook/__init__.py\u001b[0m in \u001b[0;36m_check_1d\u001b[0;34m(x)\u001b[0m\n\u001b[1;32m   1346\u001b[0m             \u001b[0;32mnot\u001b[0m \u001b[0mhasattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'ndim'\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mor\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1347\u001b[0m             len(x.shape) < 1):\n\u001b[0;32m-> 1348\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0matleast_1d\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1349\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1350\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/numpy/core/overrides.py\u001b[0m in \u001b[0;36matleast_1d\u001b[0;34m(*args, **kwargs)\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/numpy/core/shape_base.py\u001b[0m in \u001b[0;36matleast_1d\u001b[0;34m(*arys)\u001b[0m\n\u001b[1;32m     63\u001b[0m     \u001b[0mres\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     64\u001b[0m     \u001b[0;32mfor\u001b[0m \u001b[0mary\u001b[0m \u001b[0;32min\u001b[0m \u001b[0marys\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 65\u001b[0;31m         \u001b[0mary\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0masanyarray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mary\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     66\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mary\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mndim\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     67\u001b[0m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mary\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreshape\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/torch/_tensor.py\u001b[0m in \u001b[0;36m__array__\u001b[0;34m(self, dtype)\u001b[0m\n\u001b[1;32m    968\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mhandle_torch_function\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mTensor\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__array__\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdtype\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mdtype\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    969\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mdtype\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 970\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnumpy\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    971\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    972\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnumpy\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mastype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdtype\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcopy\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mTypeError\u001b[0m: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first."
          ]
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1200x400 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Set the number of epochs\n",
        "n_epochs = 4\n",
        "\n",
        "# Get the number of batches in the training and validation sets\n",
        "n_train_batches = len(train_loader)\n",
        "n_validation_batches = len(validation_loader)\n",
        "\n",
        "# Initialize softmax function\n",
        "softmax = torch.nn.Softmax(dim=1)\n",
        "\n",
        "# Initialize best validation loss\n",
        "best_valid_loss = float('inf')\n",
        "\n",
        "# Initialize lists to store loss, EM, and F1 scores for each epoch\n",
        "train_losses, valid_losses = [], []\n",
        "exact_match_scores, f1_scores = [], []\n",
        "\n",
        "# Main training loop\n",
        "for epoch in range(n_epochs):\n",
        "  \n",
        "    loop = tqdm(train_loader)\n",
        "    train_running_loss, validation_running_loss = 0.0, 0.0\n",
        "    exact_match = []\n",
        "    f1 = []\n",
        "    \n",
        "    # Training phase\n",
        "    for batch in loop:\n",
        "       \n",
        "        # Zero the gradients\n",
        "        optim.zero_grad()\n",
        "        \n",
        "        # Move data to device\n",
        "        input_ids = batch['input_ids'].to(device)\n",
        "        attention_mask = batch['attention_mask'].to(device)\n",
        "        token_type_ids = batch['token_type_ids'].to(device)\n",
        "        y_start = batch['targets_start'].to(device)\n",
        "        y_end = batch['targets_end'].to(device)\n",
        "        \n",
        "        # Forward pass\n",
        "        out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)  \n",
        "        \n",
        "        # Compute loss\n",
        "        loss = loss_fn(out_start, out_end, y_start, y_end)\n",
        "        train_running_loss += loss\n",
        "        \n",
        "        # Backward pass\n",
        "        loss.backward()\n",
        "        \n",
        "        # Update weights\n",
        "        optim.step()\n",
        "        \n",
        "        loop.set_description(f'Epoch {epoch+1} - Training')\n",
        "    \n",
        "    # Validation phase\n",
        "    with torch.no_grad():  \n",
        "        loop2 = tqdm(validation_loader)\n",
        "        for content in loop2:\n",
        "            \n",
        "            # Move data to device\n",
        "            input_ids = content['input_ids'].to(device)\n",
        "            temp_ids = input_ids.cpu().data.numpy().tolist()\n",
        "            max_ind = [temp_ids[i].index(4) for i in range(0, len(temp_ids))]  # index of first [sep]\n",
        "            attention_mask = content['attention_mask'].to(device)\n",
        "            token_type_ids = content['token_type_ids'].to(device)\n",
        "            y_start = content['targets_start'].to(device)\n",
        "            y_end = content['targets_end'].to(device)\n",
        "            \n",
        "            # Forward pass\n",
        "            out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)\n",
        "            start_pred = softmax(out_start)\n",
        "            end_pred = softmax(out_end)\n",
        "            \n",
        "            # Compute loss\n",
        "            loss2 = loss_fn(out_start, out_end, y_start, y_end)\n",
        "            validation_running_loss += loss2\n",
        "            \n",
        "            # Compute EM and F1 scores\n",
        "            start_label = np.nonzero(y_start).cpu().data.numpy()\n",
        "            end_label = np.nonzero(y_end).cpu().data.numpy()\n",
        "            tensor_label = torch.stack((torch.tensor(start_label[:, 1]), torch.tensor(end_label[:, 1])), -1)\n",
        "            start_log, end_log = generate_indexes(start_pred, end_pred, N, max_ind)\n",
        "            start_log = np.array(start_log)\n",
        "            end_log = np.array(end_log)\n",
        "            tensor_pred = torch.stack((torch.tensor(start_log), torch.tensor(end_log)), -1)\n",
        "            ex_ma = sum([1 if (tensor_pred[i][0] == tensor_label[i][0] and tensor_pred[i][1] == tensor_label[i][1]) else 0 for i in range(0, len(tensor_pred))]) / len(tensor_pred)\n",
        "            exact_match.append(ex_ma)\n",
        "            f1.append(evaluate_model(start_pred.cpu().data.numpy(), end_pred.cpu().data.numpy(), N, max_ind, start_label[:, 1], end_label[:, 1]))\n",
        "            loop2.set_description(f'Epoch {epoch+1} - Validation')\n",
        "\n",
        "    # Compute average losses and scores\n",
        "    train_loss = train_running_loss / n_train_batches\n",
        "    valid_loss = validation_running_loss / n_validation_batches\n",
        "    \n",
        "    # Append losses to the lists\n",
        "    train_losses.append(train_loss)\n",
        "    valid_losses.append(valid_loss)\n",
        "    \n",
        "    # Update best validation loss and save model weights\n",
        "    if valid_loss < best_valid_loss:\n",
        "        best_valid_loss = valid_loss\n",
        "        torch.save(model.state_dict(), '/content/drive/MyDrive/HW5/Q2/Models/saved_weights4epoch_dropout_0.05.pt')\n",
        "        print('New Model Weights Saved in Drive!')\n",
        "    \n",
        "    # Compute and append average EM and F1 scores\n",
        "    exact_match_score = 100 * np.mean(exact_match)\n",
        "    f1_score = 100 * np.mean(f1)\n",
        "    exact_match_scores.append(exact_match_score)\n",
        "    f1_scores.append(f1_score)\n",
        "    \n",
        "    # Print epoch statistics\n",
        "    print('Epoch {}: Train Loss: {:.5f}, Valid Loss: {:.5f}, Exact Match: {:.2f}%, F1: {:.2f}%'.format(epoch + 1,\n",
        "                                                                                                     train_loss,\n",
        "                                                                                                     valid_loss,\n",
        "                                                                                                     exact_match_score,\n",
        "                                                                                                     f1_score))\n",
        "\n",
        "# Plot the loss, EM, and F1 scores\n",
        "plt.figure(figsize=(12, 4))\n",
        "plt.subplot(1, 3, 1)\n",
        "plt.plot(range(1, n_epochs + 1), train_losses, label='Train Loss')\n",
        "plt.plot(range(1, n_epochs + 1), valid_losses, label='Valid Loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Loss')\n",
        "plt.title('Training and Validation Loss')\n",
        "plt.legend()\n",
        "\n",
        "plt.subplot(1, 3, 2)\n",
        "plt.plot(range(1, n_epochs + 1), exact_match_scores, label='Exact Match')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Exact Match (%)')\n",
        "plt.title('Exact Match Score')\n",
        "plt.legend()\n",
        "\n",
        "plt.subplot(1, 3, 3)\n",
        "plt.plot(range(1, n_epochs + 1), f1_scores, label='F1 Score')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('F1 Score (%)')\n",
        "plt.title('F1 Score')\n",
        "plt.legend()\n",
        "\n",
        "plt.tight_layout()\n",
        "\n",
        "# Save the plot as a PDF\n",
        "plt.savefig('loss_em_f1_scores.pdf', format='pdf')\n",
        "plt.show()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "byvaLHKbFk0_",
        "outputId": "14ee8a09-b333-4ccc-97d7-1f5512896e11"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 1 - Training: 100%|██████████| 1889/1889 [43:38<00:00,  1.39s/it]\n",
            "Epoch 1 - Validation: 100%|██████████| 265/265 [12:27<00:00,  2.82s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "New Model Weights Saved in Drive!\n",
            "Epoch 1: Train Loss: 0.01479, Valid Loss: 0.01887, Exact Match: 34.88%, F1: 72.39%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 2 - Training: 100%|██████████| 1889/1889 [43:31<00:00,  1.38s/it]\n",
            "Epoch 2 - Validation: 100%|██████████| 265/265 [12:13<00:00,  2.77s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 2: Train Loss: 0.01128, Valid Loss: 0.01963, Exact Match: 35.76%, F1: 72.16%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 3 - Training: 100%|██████████| 1889/1889 [43:34<00:00,  1.38s/it]\n",
            "Epoch 3 - Validation: 100%|██████████| 265/265 [12:17<00:00,  2.78s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 3: Train Loss: 0.00879, Valid Loss: 0.02106, Exact Match: 35.81%, F1: 71.99%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 4 - Training: 100%|██████████| 1889/1889 [43:38<00:00,  1.39s/it]\n",
            "Epoch 4 - Validation: 100%|██████████| 265/265 [12:14<00:00,  2.77s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 4: Train Loss: 0.00692, Valid Loss: 0.02193, Exact Match: 34.21%, F1: 72.08%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 5 - Training: 100%|██████████| 1889/1889 [43:38<00:00,  1.39s/it]\n",
            "Epoch 5 - Validation: 100%|██████████| 265/265 [12:20<00:00,  2.80s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 5: Train Loss: 0.00566, Valid Loss: 0.02401, Exact Match: 35.33%, F1: 71.82%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 6 - Training: 100%|██████████| 1889/1889 [43:37<00:00,  1.39s/it]\n",
            "Epoch 6 - Validation: 100%|██████████| 265/265 [12:29<00:00,  2.83s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 6: Train Loss: 0.00480, Valid Loss: 0.02503, Exact Match: 34.44%, F1: 71.28%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 7 - Training: 100%|██████████| 1889/1889 [43:28<00:00,  1.38s/it]\n",
            "Epoch 7 - Validation: 100%|██████████| 265/265 [12:32<00:00,  2.84s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 7: Train Loss: 0.00405, Valid Loss: 0.02792, Exact Match: 34.52%, F1: 72.00%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 8 - Training: 100%|██████████| 1889/1889 [2:22:53<00:00,  4.54s/it]\n",
            "Epoch 8 - Validation: 100%|██████████| 265/265 [12:22<00:00,  2.80s/it]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 8: Train Loss: 0.00361, Valid Loss: 0.02808, Exact Match: 34.05%, F1: 71.69%\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Epoch 9 - Training:  44%|████▍     | 838/1889 [4:44:39<419:09:22, 1435.74s/it]"
          ]
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Set the number of epochs\n",
        "n_epochs = 10\n",
        "\n",
        "# Get the number of batches in the training and validation sets\n",
        "n_train_batches = len(train_loader)\n",
        "n_validation_batches = len(validation_loader)\n",
        "\n",
        "# Initialize softmax function\n",
        "softmax = torch.nn.Softmax(dim=1)\n",
        "\n",
        "# Initialize best validation loss\n",
        "best_valid_loss = float('inf')\n",
        "\n",
        "# Initialize lists to store loss, EM, and F1 scores for each epoch\n",
        "train_losses, valid_losses = [], []\n",
        "exact_match_scores, f1_scores = [], []\n",
        "\n",
        "# Main training loop\n",
        "for epoch in range(n_epochs):\n",
        "  \n",
        "    loop = tqdm(train_loader)\n",
        "    train_running_loss, validation_running_loss = 0.0, 0.0\n",
        "    exact_match = []\n",
        "    f1 = []\n",
        "    \n",
        "    # Training phase\n",
        "    for batch in loop:\n",
        "       \n",
        "        # Zero the gradients\n",
        "        optim.zero_grad()\n",
        "        \n",
        "        # Move data to device\n",
        "        input_ids = batch['input_ids'].to(device)\n",
        "        attention_mask = batch['attention_mask'].to(device)\n",
        "        token_type_ids = batch['token_type_ids'].to(device)\n",
        "        y_start = batch['targets_start'].to(device)\n",
        "        y_end = batch['targets_end'].to(device)\n",
        "        \n",
        "        # Forward pass\n",
        "        out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)  \n",
        "        \n",
        "        # Compute loss\n",
        "        loss = loss_fn(out_start, out_end, y_start, y_end)\n",
        "        train_running_loss += loss\n",
        "        \n",
        "        # Backward pass\n",
        "        loss.backward()\n",
        "        \n",
        "        # Update weights\n",
        "        optim.step()\n",
        "        \n",
        "        loop.set_description(f'Epoch {epoch+1} - Training')\n",
        "    \n",
        "    # Validation phase\n",
        "    with torch.no_grad():  \n",
        "        loop2 = tqdm(validation_loader)\n",
        "        for content in loop2:\n",
        "            \n",
        "            # Move data to device\n",
        "            input_ids = content['input_ids'].to(device)\n",
        "            temp_ids = input_ids.cpu().data.numpy().tolist()\n",
        "            max_ind = [temp_ids[i].index(4) for i in range(0, len(temp_ids))]  # index of first [sep]\n",
        "            attention_mask = content['attention_mask'].to(device)\n",
        "            token_type_ids = content['token_type_ids'].to(device)\n",
        "            y_start = content['targets_start'].to(device)\n",
        "            y_end = content['targets_end'].to(device)\n",
        "            \n",
        "            # Forward pass\n",
        "            out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)\n",
        "            start_pred = softmax(out_start)\n",
        "            end_pred = softmax(out_end)\n",
        "            \n",
        "            # Compute loss\n",
        "            loss2 = loss_fn(out_start, out_end, y_start, y_end)\n",
        "            validation_running_loss += loss2\n",
        "            \n",
        "            # Compute EM and F1 scores\n",
        "            start_label = np.nonzero(y_start).cpu().data.numpy()\n",
        "            end_label = np.nonzero(y_end).cpu().data.numpy()\n",
        "            tensor_label = torch.stack((torch.tensor(start_label[:, 1]), torch.tensor(end_label[:, 1])), -1)\n",
        "            start_log, end_log = generate_indexes(start_pred, end_pred, N, max_ind)\n",
        "            start_log = np.array(start_log)\n",
        "            end_log = np.array(end_log)\n",
        "            tensor_pred = torch.stack((torch.tensor(start_log), torch.tensor(end_log)), -1)\n",
        "            ex_ma = sum([1 if (tensor_pred[i][0] == tensor_label[i][0] and tensor_pred[i][1] == tensor_label[i][1]) else 0 for i in range(0, len(tensor_pred))]) / len(tensor_pred)\n",
        "            exact_match.append(ex_ma)\n",
        "            f1.append(evaluate_model(start_pred.cpu().data.numpy(), end_pred.cpu().data.numpy(), N, max_ind, start_label[:, 1], end_label[:, 1]))\n",
        "            loop2.set_description(f'Epoch {epoch+1} - Validation')\n",
        "\n",
        "    # Compute average losses and scores\n",
        "    train_loss = train_running_loss / n_train_batches\n",
        "    valid_loss = validation_running_loss / n_validation_batches\n",
        "    \n",
        "    # Append losses to the lists\n",
        "    train_losses.append(train_loss)\n",
        "    valid_losses.append(valid_loss)\n",
        "    \n",
        "    # Update best validation loss and save model weights\n",
        "    if valid_loss < best_valid_loss:\n",
        "        best_valid_loss = valid_loss\n",
        "        torch.save(model.state_dict(), '/content/drive/MyDrive/HW5/Q2/Models/saved_weights_dropout_0.05.pt')\n",
        "        print('New Model Weights Saved in Drive!')\n",
        "    \n",
        "    # Compute and append average EM and F1 scores\n",
        "    exact_match_score = 100 * np.mean(exact_match)\n",
        "    f1_score = 100 * np.mean(f1)\n",
        "    exact_match_scores.append(exact_match_score)\n",
        "    f1_scores.append(f1_score)\n",
        "    \n",
        "    # Print epoch statistics\n",
        "    print('Epoch {}: Train Loss: {:.5f}, Valid Loss: {:.5f}, Exact Match: {:.2f}%, F1: {:.2f}%'.format(epoch + 1,\n",
        "                                                                                                     train_loss,\n",
        "                                                                                                     valid_loss,\n",
        "                                                                                                     exact_match_score,\n",
        "                                                                                                     f1_score))\n",
        "\n",
        "# Plot the loss, EM, and F1 scores\n",
        "plt.figure(figsize=(12, 4))\n",
        "plt.subplot(1, 3, 1)\n",
        "plt.plot(range(1, n_epochs + 1), train_losses, label='Train Loss')\n",
        "plt.plot(range(1, n_epochs + 1), valid_losses, label='Valid Loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Loss')\n",
        "plt.title('Training and Validation Loss')\n",
        "plt.legend()\n",
        "\n",
        "plt.subplot(1, 3, 2)\n",
        "plt.plot(range(1, n_epochs + 1), exact_match_scores, label='Exact Match')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Exact Match (%)')\n",
        "plt.title('Exact Match Score')\n",
        "plt.legend()\n",
        "\n",
        "plt.subplot(1, 3, 3)\n",
        "plt.plot(range(1, n_epochs + 1), f1_scores, label='F1 Score')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('F1 Score (%)')\n",
        "plt.title('F1 Score')\n",
        "plt.legend()\n",
        "\n",
        "plt.tight_layout()\n",
        "\n",
        "# Save the plot as a PDF\n",
        "plt.savefig('loss_em_f1_scores.pdf', format='pdf')\n",
        "plt.show()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XjAxxAfuKIpI"
      },
      "source": [
        "# 1.2. Load and Processs Data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0EmedyffDtFc",
        "outputId": "233c030b-ec9b-4736-b29f-974c069fe621"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting transformers\n",
            "  Downloading transformers-4.30.0-py3-none-any.whl (7.2 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m7.2/7.2 MB\u001b[0m \u001b[31m59.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from transformers) (3.12.0)\n",
            "Collecting huggingface-hub<1.0,>=0.14.1 (from transformers)\n",
            "  Downloading huggingface_hub-0.15.1-py3-none-any.whl (236 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m236.8/236.8 kB\u001b[0m \u001b[31m27.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.10/dist-packages (from transformers) (1.22.4)\n",
            "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.10/dist-packages (from transformers) (23.1)\n",
            "Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.10/dist-packages (from transformers) (6.0)\n",
            "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.10/dist-packages (from transformers) (2022.10.31)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from transformers) (2.27.1)\n",
            "Collecting tokenizers!=0.11.3,<0.14,>=0.11.1 (from transformers)\n",
            "  Downloading tokenizers-0.13.3-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (7.8 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m7.8/7.8 MB\u001b[0m \u001b[31m87.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting safetensors>=0.3.1 (from transformers)\n",
            "  Downloading safetensors-0.3.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.3 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.3/1.3 MB\u001b[0m \u001b[31m79.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.10/dist-packages (from transformers) (4.65.0)\n",
            "Requirement already satisfied: fsspec in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.14.1->transformers) (2023.4.0)\n",
            "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.14.1->transformers) (4.5.0)\n",
            "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (1.26.15)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (2022.12.7)\n",
            "Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (2.0.12)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (3.4)\n",
            "Installing collected packages: tokenizers, safetensors, huggingface-hub, transformers\n",
            "Successfully installed huggingface-hub-0.15.1 safetensors-0.3.1 tokenizers-0.13.3 transformers-4.30.0\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.3/1.3 MB\u001b[0m \u001b[31m27.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hLooking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: sentencepiece in /usr/local/lib/python3.10/dist-packages (0.1.99)\n"
          ]
        }
      ],
      "source": [
        "# Installing required packages\n",
        "!pip install transformers\n",
        "!pip install -q sentencepiece\n",
        "!pip install sentencepiece\n",
        "\n",
        "# Importing necessary libraries\n",
        "import re\n",
        "import json\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "from tqdm import tqdm\n",
        "from tqdm.auto import tqdm\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.utils import shuffle\n",
        "from transformers import BertModel, BertConfig, BertTokenizer, AdamW"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Rzx9QgRWdEgR",
        "outputId": "896cd7be-bd48-413a-f2d6-b586b2b0ab2e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Cloning into 'PQuAD'...\n",
            "remote: Enumerating objects: 27, done.\u001b[K\n",
            "remote: Counting objects: 100% (27/27), done.\u001b[K\n",
            "remote: Compressing objects: 100% (22/22), done.\u001b[K\n",
            "remote: Total 27 (delta 9), reused 15 (delta 3), pack-reused 0\u001b[K\n",
            "Unpacking objects: 100% (27/27), 5.71 MiB | 3.59 MiB/s, done.\n"
          ]
        }
      ],
      "source": [
        "# Clone the PQuAD dataset from GitHub\n",
        "!git clone https://github.com/AUT-NLP/PQuAD.git\n",
        "\n",
        "# Load the PQuad dataset\n",
        "train_file_path = '/content/PQuAD/Dataset/Train.json'\n",
        "test_file_path = '/content/PQuAD/Dataset/Test.json'\n",
        "val_file_path = '/content/PQuAD/Dataset/Validation.json'"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "q2MYpVStprcq",
        "outputId": "0853dce2-4df7-4767-da74-27ef5e2a8307"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Dataset Statistics (Train):\n",
            "Total Questions: 8979\n",
            "Total Unanswerable Questions: 15721\n",
            "Mean # of Paragraph Tokens: 919.56\n",
            "Mean # of Question Tokens: 73.69\n",
            "Mean # of Answer Tokens: 28.06\n",
            "\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Dataset Statistics (Test):\n",
            "Total Questions: 1059\n",
            "Total Unanswerable Questions: 1914\n",
            "Mean # of Paragraph Tokens: 968.15\n",
            "Mean # of Question Tokens: 81.91\n",
            "Mean # of Answer Tokens: 41.80\n",
            "\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Dataset Statistics (Validation):\n",
            "Total Questions: 1103\n",
            "Total Unanswerable Questions: 1981\n",
            "Mean # of Paragraph Tokens: 905.68\n",
            "Mean # of Question Tokens: 76.47\n",
            "Mean # of Answer Tokens: 42.83\n",
            "\n"
          ]
        }
      ],
      "source": [
        "import json\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "# Path to the dataset files\n",
        "train_file_path = '/content/PQuAD/Dataset/Train.json'\n",
        "test_file_path = '/content/PQuAD/Dataset/Test.json'\n",
        "val_file_path = '/content/PQuAD/Dataset/Validation.json'\n",
        "\n",
        "def load_dataset(file_path):\n",
        "    with open(file_path, 'r', encoding='utf-8') as file:\n",
        "        dataset = json.load(file)\n",
        "    return dataset\n",
        "\n",
        "def plot_answer_distribution(dataset, dataset_name, file_name):\n",
        "    num_answers = []\n",
        "    for data in dataset['data']:\n",
        "        for paragraph in data['paragraphs']:\n",
        "            for qa in paragraph['qas']:\n",
        "                num_answers.append(len(qa['answers']))\n",
        "\n",
        "    bins = np.arange(max(num_answers) + 2)\n",
        "    hist, edges = np.histogram(num_answers, bins=bins)\n",
        "\n",
        "    fig, ax = plt.subplots()\n",
        "    ax.bar(edges[:-1], hist, width=0.8, align='center', color='steelblue')\n",
        "\n",
        "    # Add count labels in the middle and above each bar\n",
        "    for i, count in enumerate(hist):\n",
        "        if count > 0:\n",
        "            ax.text(edges[i], count + 1, str(int(count)), ha='center', va='bottom')\n",
        "\n",
        "    plt.xlabel('Number of Answers')\n",
        "    plt.ylabel('Number of Questions')\n",
        "    plt.title(f'Distribution of Number of Answers for Questions ({dataset_name})')\n",
        "    plt.xticks(np.arange(max(num_answers) + 1), [f'{i}-answer' for i in range(max(num_answers) + 1)], rotation=90)\n",
        "    plt.legend([dataset_name])\n",
        "    plt.savefig(file_name, format='pdf', bbox_inches='tight')\n",
        "    plt.show()\n",
        "\n",
        "def print_dataset_statistics(dataset, dataset_name):\n",
        "    total_questions = 0\n",
        "    total_unanswerable_questions = 0\n",
        "    total_paragraph_tokens = 0\n",
        "    total_question_tokens = 0\n",
        "    total_answer_tokens = 0\n",
        "\n",
        "    for data in dataset['data']:\n",
        "        total_questions += len(data['paragraphs'])\n",
        "        for paragraph in data['paragraphs']:\n",
        "            for qa in paragraph['qas']:\n",
        "                if qa['is_impossible']:\n",
        "                    total_unanswerable_questions += 1\n",
        "                total_paragraph_tokens += len(paragraph['context'].split())\n",
        "                total_question_tokens += len(qa['question'].split())\n",
        "                total_answer_tokens += sum([len(answer['text'].split()) for answer in qa['answers']])\n",
        "\n",
        "    mean_paragraph_tokens = total_paragraph_tokens / total_questions\n",
        "    mean_question_tokens = total_question_tokens / total_questions\n",
        "    mean_answer_tokens = total_answer_tokens / total_questions\n",
        "\n",
        "    print(f\"Dataset Statistics ({dataset_name}):\")\n",
        "    print(f\"Total Questions: {total_questions}\")\n",
        "    print(f\"Total Unanswerable Questions: {total_unanswerable_questions}\")\n",
        "    print(f\"Mean # of Paragraph Tokens: {mean_paragraph_tokens:.2f}\")\n",
        "    print(f\"Mean # of Question Tokens: {mean_question_tokens:.2f}\")\n",
        "    print(f\"Mean # of Answer Tokens: {mean_answer_tokens:.2f}\")\n",
        "    print()\n",
        "\n",
        "# Load the train dataset\n",
        "train_dataset = load_dataset(train_file_path)\n",
        "\n",
        "# Plot the distribution of the number of answers for train questions and save as PDF\n",
        "plot_answer_distribution(train_dataset, 'Train', 'traindistribution.pdf')\n",
        "\n",
        "# Print the statistical information of the train dataset\n",
        "print_dataset_statistics(train_dataset, 'Train')\n",
        "\n",
        "# Load the test dataset\n",
        "test_dataset = load_dataset(test_file_path)\n",
        "\n",
        "# Plot the distribution of the number of answers for test questions and save as PDF\n",
        "plot_answer_distribution(test_dataset, 'Test', 'testdistribution.pdf')\n",
        "\n",
        "# Print the statistical information of the test dataset\n",
        "print_dataset_statistics(test_dataset, 'Test')\n",
        "\n",
        "# Load the validation dataset\n",
        "val_dataset = load_dataset(val_file_path)\n",
        "\n",
        "# Plot the distribution of the number of answers for validation questions and save as PDF\n",
        "plot_answer_distribution(val_dataset, 'Validation', 'validationdistribution.pdf')\n",
        "\n",
        "# Print the statistical information of the validation dataset\n",
        "print_dataset_statistics(val_dataset, 'Validation')\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "rgGxss1u39qH"
      },
      "outputs": [],
      "source": [
        "# Set the minimum and maximum length for data\n",
        "min_len, max_len = 128, 256\n",
        "\n",
        "# Set the value of N\n",
        "N = 3"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 241
        },
        "id": "faaSJSCzCq7A",
        "outputId": "be92918b-dbb4-4cd0-a575-144d8f31b68b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Size of the train DataFrame before concatenation is 63994\n",
            "Size of the train DataFrame after concatenation is 63994\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "       Id   title                                            context  \\\n",
              "0  101001  آرسنال  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "1  101002  آرسنال  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "2  101003  آرسنال  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "3  101004  آرسنال  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "4  101005  آرسنال  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "\n",
              "                                            question  ans_start  \\\n",
              "0      موقعیت جغرافی باشگاه فوتبال آرسنال را بگویید؟         86   \n",
              "1  لیگ برتر انگلستان موفق به کسب چند عنوان قهرمان...        173   \n",
              "2  بیشترین بازی بدون باخت پیاپی متعلق به کدام باش...         61   \n",
              "3  باشگاه فوتبال آرسنال موفق به کسب چند عنوان قهر...        119   \n",
              "4  باشگاه فوتبال آرسنال چند عنوان قهرمانی در جام ...        214   \n",
              "\n",
              "                    text  \n",
              "0          شمال شهر لندن  \n",
              "1                     ۱۴  \n",
              "2  باشگاه فوتبال انگلیسی  \n",
              "3                     ۱۳  \n",
              "4                     ۱۶  "
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-69394ac5-269f-4da6-9d7e-efc612cc8548\">\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>Id</th>\n",
              "      <th>title</th>\n",
              "      <th>context</th>\n",
              "      <th>question</th>\n",
              "      <th>ans_start</th>\n",
              "      <th>text</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>101001</td>\n",
              "      <td>آرسنال</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>موقعیت جغرافی باشگاه فوتبال آرسنال را بگویید؟</td>\n",
              "      <td>86</td>\n",
              "      <td>شمال شهر لندن</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>101002</td>\n",
              "      <td>آرسنال</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>لیگ برتر انگلستان موفق به کسب چند عنوان قهرمان...</td>\n",
              "      <td>173</td>\n",
              "      <td>۱۴</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>101003</td>\n",
              "      <td>آرسنال</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>بیشترین بازی بدون باخت پیاپی متعلق به کدام باش...</td>\n",
              "      <td>61</td>\n",
              "      <td>باشگاه فوتبال انگلیسی</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>101004</td>\n",
              "      <td>آرسنال</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>باشگاه فوتبال آرسنال موفق به کسب چند عنوان قهر...</td>\n",
              "      <td>119</td>\n",
              "      <td>۱۳</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>101005</td>\n",
              "      <td>آرسنال</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>باشگاه فوتبال آرسنال چند عنوان قهرمانی در جام ...</td>\n",
              "      <td>214</td>\n",
              "      <td>۱۶</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-69394ac5-269f-4da6-9d7e-efc612cc8548')\"\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-69394ac5-269f-4da6-9d7e-efc612cc8548 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-69394ac5-269f-4da6-9d7e-efc612cc8548');\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": 7
        }
      ],
      "source": [
        "def json_to_dataframe(file):\n",
        "    # Open the JSON file\n",
        "    f = open(file, \"r\")\n",
        "    data = json.loads(f.read())  # Load the JSON file\n",
        "\n",
        "    # Create empty lists to store values\n",
        "    ids = []\n",
        "    titles = []\n",
        "    contexts = []\n",
        "    questions = []\n",
        "    ans_starts = []\n",
        "    texts = []\n",
        "\n",
        "    # Iterate over the JSON data\n",
        "    for i in range(len(data['data'])):\n",
        "        title = data['data'][i]['title']  # Extract the 'title' value\n",
        "\n",
        "        # Iterate over the paragraphs in the JSON data\n",
        "        for p in range(len(data['data'][i]['paragraphs'])):\n",
        "            context = data['data'][i]['paragraphs'][p]['context']  # Extract the 'context' value\n",
        "\n",
        "            # Iterate over the questions in the JSON data\n",
        "            for q in range(len(data['data'][i]['paragraphs'][p]['qas'])):\n",
        "                question = data['data'][i]['paragraphs'][p]['qas'][q]['question']  # Extract the 'question' value\n",
        "                qid = data['data'][i]['paragraphs'][p]['qas'][q]['id']  # Extract the 'id' value\n",
        "\n",
        "                # Check if the question has answers\n",
        "                if len(data['data'][i]['paragraphs'][p]['qas'][q]['answers']) == 0:\n",
        "                    ans_start = -1\n",
        "                    text = ''\n",
        "\n",
        "                    # Append the values to the lists\n",
        "                    titles.append(title)\n",
        "                    contexts.append(context)\n",
        "                    questions.append(question)\n",
        "                    ids.append(qid)\n",
        "                    ans_starts.append(ans_start)\n",
        "                    texts.append(text)\n",
        "                else:\n",
        "                    # Iterate over the answers in the JSON data\n",
        "                    for a in range(len(data['data'][i]['paragraphs'][p]['qas'][q]['answers'])):\n",
        "                        ans_start = data['data'][i]['paragraphs'][p]['qas'][q]['answers'][a]['answer_start']  # Extract the 'answer_start' value\n",
        "                        text = data['data'][i]['paragraphs'][p]['qas'][q]['answers'][a]['text']  # Extract the 'text' value\n",
        "\n",
        "                        # Append the values to the lists\n",
        "                        titles.append(title)\n",
        "                        contexts.append(context)\n",
        "                        questions.append(question)\n",
        "                        ids.append(qid)\n",
        "                        ans_starts.append(ans_start)\n",
        "                        texts.append(text)\n",
        "\n",
        "    # Create an empty DataFrame\n",
        "    new_df = pd.DataFrame(columns=['Id', 'title', 'context', 'question', 'ans_start', 'text'])\n",
        "\n",
        "    # Set the values of the DataFrame columns\n",
        "    new_df['Id'] = ids\n",
        "    new_df['title'] = titles\n",
        "    new_df['context'] = contexts\n",
        "    new_df['question'] = questions\n",
        "    new_df['ans_start'] = ans_starts\n",
        "    new_df['text'] = texts\n",
        "\n",
        "    # Drop duplicate rows from the DataFrame\n",
        "    final_df = new_df.drop_duplicates(keep='first')\n",
        "\n",
        "    return final_df\n",
        "\n",
        "\n",
        "# Convert the train JSON file to a DataFrame\n",
        "df_train = json_to_dataframe(train_file_path)\n",
        "\n",
        "# Get the number of rows in the train DataFrame\n",
        "train_rows = df_train.shape[0]\n",
        "print('Size of the train DataFrame before concatenation is {}'.format(train_rows))\n",
        "\n",
        "# Convert the test JSON file to a DataFrame\n",
        "df_test = json_to_dataframe(test_file_path)\n",
        "\n",
        "# Convert the validation JSON file to a DataFrame\n",
        "df_validation = json_to_dataframe(val_file_path)\n",
        "\n",
        "# Concatenate the train and validation DataFrames\n",
        "# frames = [df_train, df_validation]\n",
        "# df_train = pd.concat(frames)\n",
        "\n",
        "# Get the number of rows in the concatenated train DataFrame\n",
        "train_rows = df_train.shape[0]\n",
        "print('Size of the train DataFrame after concatenation is {}'.format(train_rows))\n",
        "\n",
        "# Display the first few rows of the train DataFrame\n",
        "df_train.head()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "id": "9fLoT3O2Q6HQ"
      },
      "outputs": [],
      "source": [
        "def add_end_index(answers_text, answers_start, contexts):\n",
        "    new_answers = []\n",
        "\n",
        "    # Loop through each answer-context pair\n",
        "    for answer_text, answer_start, context in tqdm(zip(answers_text, answers_start, contexts)):\n",
        "        start_shift = 0\n",
        "\n",
        "        # Remove start half-spaces from the answer text\n",
        "        text = re.sub(\"^\\u200c\", \"\", answer_text)\n",
        "        \n",
        "        # Check if the length of the text is one less than the length of the original answer\n",
        "        if len(list(text)) == (len(list(answer_text)) - 1):\n",
        "            start_shift += 1\n",
        "        \n",
        "        # Remove end half-spaces from the text\n",
        "        text = re.sub(\"\\u200c$\", \"\", text)\n",
        "        \n",
        "        # Remove leading and trailing whitespaces from the text\n",
        "        text = re.sub(\"^\\s+\", '', text)\n",
        "        text = re.sub(\"\\s+$\", '', text)\n",
        "\n",
        "        # Adjust the answer_start index by the start_shift value\n",
        "        answer_start += start_shift\n",
        "        \n",
        "        # Calculate the end index of the answer\n",
        "        end_idx = answer_start + len(text)\n",
        "\n",
        "        # Check if the answer is correct\n",
        "        if context[answer_start:end_idx] == text:\n",
        "            # If the answer is correct, set the answer_end index to end_idx\n",
        "            answer_end = end_idx\n",
        "        else:\n",
        "            # If the answer is off by 1-2 tokens, adjust the answer_start and answer_end indices\n",
        "            for n in [1, 2]:\n",
        "                if context[answer_start - n:end_idx - n] == text:\n",
        "                    answer_start = answer_start - n\n",
        "                    answer_end = end_idx - n\n",
        "\n",
        "        # Append the modified answer to the new_answers list\n",
        "        new_answers.append({'text': text, 'answer_start': answer_start, 'answer_end': answer_end})\n",
        "\n",
        "    return new_answers\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "2zt3FsiwRvwR"
      },
      "outputs": [],
      "source": [
        "def prepare_data(dataset):\n",
        "    # Extract necessary columns from the dataset\n",
        "    answer_start = dataset['ans_start'].tolist()\n",
        "    text = dataset['text'].tolist()\n",
        "    questions = dataset['question'].tolist()\n",
        "    contexts = dataset['context'].tolist()\n",
        "\n",
        "    # Call the add_end_index function to process answers\n",
        "    answers = add_end_index(text, answer_start, contexts)\n",
        "\n",
        "    # Return a dictionary with the prepared data\n",
        "    return {\n",
        "        'question': questions,\n",
        "        'context': contexts,\n",
        "        'answers': answers\n",
        "    }\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7l41LIciVZY5",
        "outputId": "4c9ef808-5296-4af8-f9b4-2a68e705002b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "63994it [00:00, 150254.59it/s]\n",
            "10379it [00:00, 116380.74it/s]\n",
            "10417it [00:00, 155083.23it/s]\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'text': 'پنج عنوان قهرمانی در چمپیونشیپ و دو قهرمانی در جام حذفی',\n",
              " 'answer_start': 288,\n",
              " 'answer_end': 343}"
            ]
          },
          "metadata": {},
          "execution_count": 10
        }
      ],
      "source": [
        "# Prepare the training dataset\n",
        "train_dataset = prepare_data(df_train)\n",
        "\n",
        "# Prepare the validation dataset\n",
        "val_dataset = prepare_data(df_validation)\n",
        "\n",
        "# Prepare the test dataset\n",
        "test_dataset = prepare_data(df_test)\n",
        "\n",
        "# Access the second answer in the training dataset\n",
        "answer = train_dataset['answers'][10]\n",
        "train_dataset['answers'][10]"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# 1.3. ALBERT (Data Processing and Training)"
      ],
      "metadata": {
        "id": "6eQpbr2EoZtt"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install -q sentencepiece"
      ],
      "metadata": {
        "id": "eD_KSl9mxCOL"
      },
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from transformers import AutoConfig, AutoTokenizer\n",
        "\n",
        "config = AutoConfig.from_pretrained(\"m3hrdadfi/albert-fa-base-v2\")\n",
        "tokenizer = AutoTokenizer.from_pretrained(\"m3hrdadfi/albert-fa-base-v2\")\n"
      ],
      "metadata": {
        "id": "-udk7HSpwwG4",
        "outputId": "d2dcd668-4b23-4536-cfb1-6823d6626fdb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 81,
          "referenced_widgets": [
            "fac2919f7a2045a792085d01d84bfa86",
            "908ba1eedfee4bdeb1f35566666609b5",
            "46b514d202bb4959aefa38fe22bd88c0",
            "0dbbb1a1133c44b5bbadfe2c25fea3b9",
            "8d770239e7db43c7a63486d2dcba07ef",
            "877e783b6b4e4a4dbb34234990b2681e",
            "5f5e3fec69004f0ab7135ce44747dd07",
            "0b645147a3bb44379ae50628b933d811",
            "58d83848f06a41b3ac4c952eb913e6f8",
            "762ae5c6efca4a45a487e7157ccd9d1b",
            "6a4c8bc5b6774bd589ec46a530851d39",
            "13e8d36217a946c9b6b4420f6ca2a19c",
            "426de396581945e093160bef6a6d4bd9",
            "d9d46af007b449f79f309bb158dc88a5",
            "b898b80900504a9491e0631351e30188",
            "c4855e4955aa4840ab3d998ec12be433",
            "e58dc03387334b13b567413eed33ec2d",
            "f7f9629826614886b4a2ec9187293620",
            "6f6b64b779664a14bfab0b082847a713",
            "08429e11961447189846b97dee04186c",
            "821cb81683974ea5ba2488ff41aeb3f0",
            "08a01159787440dda7a02041a9c05250"
          ]
        }
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading (…)lve/main/config.json:   0%|          | 0.00/684 [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "fac2919f7a2045a792085d01d84bfa86"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading (…)ve/main/spiece.model:   0%|          | 0.00/1.88M [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "13e8d36217a946c9b6b4420f6ca2a19c"
            }
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PbltbZdlLWmn",
        "outputId": "642d58c0-6f27-4fb0-f34a-ab1d5f6bd4ac"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Total size of data is 63994\n",
            "Number of data without [UNK] and containing 128-256 tokens is 55152\n",
            "Percentage of data without [UNK] and containing 128-256 tokens is 86.18307966371847\n"
          ]
        }
      ],
      "source": [
        "# Create empty lists to store the new tokenized data\n",
        "new_context, new_question, new_answer = [[] for _ in range(3)]\n",
        "\n",
        "# Get the total size of the data\n",
        "len_data = len(train_dataset['answers'])\n",
        "print('Total size of data is {}'.format(len_data))\n",
        "\n",
        "# Iterate through each data instance\n",
        "for i in range(len_data):\n",
        "    # Tokenize the answer, context, and question\n",
        "    tokenized_answer = tokenizer.tokenize(train_dataset['answers'][i]['text'])\n",
        "    tokenized_context = tokenizer.tokenize(train_dataset['context'][i])\n",
        "    tokenized_question = tokenizer.tokenize(train_dataset['question'][i])\n",
        "    \n",
        "    # Calculate the total number of tokens\n",
        "    num = len(tokenized_context) + len(tokenized_question)\n",
        "    \n",
        "    # Check if the total number of tokens is within the desired range\n",
        "    if num > (min_len - 3) and num <= (max_len - 3):  # 3 for three special tokens: 1 for [CLS] and 2 for [SEP]\n",
        "        if '[UNK]' not in tokenized_answer:\n",
        "            # Append the tokenized data to the new lists\n",
        "            new_context.append(train_dataset['context'][i])\n",
        "            new_question.append(train_dataset['question'][i])\n",
        "            new_answer.append({\n",
        "                'text': train_dataset['answers'][i]['text'],\n",
        "                'answer_start': train_dataset['answers'][i]['answer_start'],\n",
        "                'answer_end': train_dataset['answers'][i]['answer_end']\n",
        "            })\n",
        "\n",
        "# Print the number of data without [UNK] and containing 128-256 tokens\n",
        "print('Number of data without [UNK] and containing 128-256 tokens is {}'.format(len(new_context)))\n",
        "\n",
        "# Print the percentage of data without [UNK] and containing 128-256 tokens\n",
        "print('Percentage of data without [UNK] and containing 128-256 tokens is {}'.format(100 * len(new_context) / len_data))\n",
        "\n",
        "# Create a new training dataset with the filtered data\n",
        "new_train_dataset = {\n",
        "    'question': new_question,\n",
        "    'context': new_context,\n",
        "    'answers': new_answer\n",
        "}\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "id": "4MGbeO5u2zio",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ed0a6119-b4f8-4b79-e37c-b07bbb060c83"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Total size of data is 10379\n",
            "Number of data without [UNK] and containing 128-256 tokens is 9054\n",
            "Percentage of data without [UNK] and containing 128-256 tokens is 87.23383755660468\n"
          ]
        }
      ],
      "source": [
        "# Create empty lists to store the new tokenized data\n",
        "new_context, new_question, new_answer = [[] for _ in range(3)]\n",
        "\n",
        "# Get the total size of the data\n",
        "len_data = len(val_dataset['answers'])\n",
        "print('Total size of data is {}'.format(len_data))\n",
        "\n",
        "# Iterate through each data instance\n",
        "for i in range(len_data):\n",
        "    # Tokenize the answer, context, and question\n",
        "    tokenized_answer = tokenizer.tokenize(val_dataset['answers'][i]['text'])\n",
        "    tokenized_context = tokenizer.tokenize(val_dataset['context'][i])\n",
        "    tokenized_question = tokenizer.tokenize(val_dataset['question'][i])\n",
        "    \n",
        "    # Calculate the total number of tokens\n",
        "    num = len(tokenized_context) + len(tokenized_question)\n",
        "    \n",
        "    # Check if the total number of tokens is within the desired range\n",
        "    if num > (min_len - 3) and num <= (max_len - 3):  # 3 for three special tokens, [CLS] and 2 [SEP]\n",
        "        if '[UNK]' not in tokenized_answer:\n",
        "            # Append the tokenized data to the new lists\n",
        "            new_context.append(val_dataset['context'][i])\n",
        "            new_question.append(val_dataset['question'][i])\n",
        "            new_answer.append({\n",
        "                'text': val_dataset['answers'][i]['text'],\n",
        "                'answer_start': val_dataset['answers'][i]['answer_start'],\n",
        "                'answer_end': val_dataset['answers'][i]['answer_end']\n",
        "            })\n",
        "\n",
        "# Print the number of data without [UNK] and containing 128-256 tokens\n",
        "print('Number of data without [UNK] and containing 128-256 tokens is {}'.format(len(new_context)))\n",
        "\n",
        "# Print the percentage of data without [UNK] and containing 128-256 tokens\n",
        "print('Percentage of data without [UNK] and containing 128-256 tokens is {}'.format(100 * len(new_context) / len_data))\n",
        "\n",
        "# Create a new validation dataset with the filtered data\n",
        "new_val_dataset = {\n",
        "    'question': new_question,\n",
        "    'context': new_context,\n",
        "    'answers': new_answer\n",
        "}\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {
        "id": "qcrbV1AYyqpL",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e56e6bce-3a00-4e1b-8dd0-ff49e771d4f1"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Total size of data is 10417\n",
            "Number of data without [UNK] and containing 128-256 tokens is 8818\n",
            "Percentage of data without [UNK] and containing 128-256 tokens is 84.6500911970817\n"
          ]
        }
      ],
      "source": [
        "# Create empty lists to store the new tokenized data\n",
        "new_context, new_question, new_answer = [[] for _ in range(3)]\n",
        "\n",
        "# Get the total size of the data\n",
        "len_data = len(test_dataset['answers'])\n",
        "print('Total size of data is {}'.format(len_data))\n",
        "\n",
        "# Iterate through each data instance\n",
        "for i in range(len_data):\n",
        "    # Tokenize the answer, context, and question\n",
        "    tokenized_answer = tokenizer.tokenize(test_dataset['answers'][i]['text'])\n",
        "    tokenized_context = tokenizer.tokenize(test_dataset['context'][i])\n",
        "    tokenized_question = tokenizer.tokenize(test_dataset['question'][i])\n",
        "    \n",
        "    # Calculate the total number of tokens\n",
        "    num = len(tokenized_context) + len(tokenized_question)\n",
        "    \n",
        "    # Check if the total number of tokens is within the desired range\n",
        "    if num > (min_len - 3) and num <= (max_len - 3):  # 3 for three special tokens, [CLS] and 2 [SEP]\n",
        "        if '[UNK]' not in tokenized_answer:\n",
        "            # Append the tokenized data to the new lists\n",
        "            new_context.append(test_dataset['context'][i])\n",
        "            new_question.append(test_dataset['question'][i])\n",
        "            new_answer.append({\n",
        "                'text': test_dataset['answers'][i]['text'],\n",
        "                'answer_start': test_dataset['answers'][i]['answer_start'],\n",
        "                'answer_end': test_dataset['answers'][i]['answer_end']\n",
        "            })\n",
        "\n",
        "# Print the number of data without [UNK] and containing 128-256 tokens\n",
        "print('Number of data without [UNK] and containing 128-256 tokens is {}'.format(len(new_context)))\n",
        "\n",
        "# Print the percentage of data without [UNK] and containing 128-256 tokens\n",
        "print('Percentage of data without [UNK] and containing 128-256 tokens is {}'.format(100 * len(new_context) / len_data))\n",
        "\n",
        "# Create a new validation dataset with the filtered data\n",
        "new_test_dataset = {\n",
        "    'question': new_question,\n",
        "    'context': new_context,\n",
        "    'answers': new_answer\n",
        "}\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "id": "YMem82rmMZmD"
      },
      "outputs": [],
      "source": [
        "# Create a DataFrame for the test dataset\n",
        "test_df = pd.DataFrame.from_dict(new_test_dataset)\n",
        "\n",
        "# Create a DataFrame for the validation dataset\n",
        "validation_df = pd.DataFrame.from_dict(new_val_dataset)\n",
        "\n",
        "# Create a DataFrame for the training dataset\n",
        "train_df = pd.DataFrame.from_dict(new_train_dataset)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "id": "mAHfF-0Vy-Wa",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 241
        },
        "outputId": "075b3064-1615-4d11-9484-0b652059f42c"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Length of the training dataset: 55152\n",
            "Training dataset:\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                            question  \\\n",
              "0      موقعیت جغرافی باشگاه فوتبال آرسنال را بگویید؟   \n",
              "1  لیگ برتر انگلستان موفق به کسب چند عنوان قهرمان...   \n",
              "2  بیشترین بازی بدون باخت پیاپی متعلق به کدام باش...   \n",
              "3  باشگاه فوتبال آرسنال موفق به کسب چند عنوان قهر...   \n",
              "4  باشگاه فوتبال آرسنال چند عنوان قهرمانی در جام ...   \n",
              "\n",
              "                                             context  \\\n",
              "0  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "1  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "2  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "3  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "4  باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...   \n",
              "\n",
              "                                             answers  \n",
              "0  {'text': 'شمال شهر لندن', 'answer_start': 86, ...  \n",
              "1  {'text': '۱۴', 'answer_start': 173, 'answer_en...  \n",
              "2  {'text': 'باشگاه فوتبال انگلیسی', 'answer_star...  \n",
              "3  {'text': '۱۳', 'answer_start': 119, 'answer_en...  \n",
              "4  {'text': '۱۶', 'answer_start': 214, 'answer_en...  "
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-d9b533d3-d3ad-4e74-aaf3-0042edc2f613\">\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>question</th>\n",
              "      <th>context</th>\n",
              "      <th>answers</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>موقعیت جغرافی باشگاه فوتبال آرسنال را بگویید؟</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>{'text': 'شمال شهر لندن', 'answer_start': 86, ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>لیگ برتر انگلستان موفق به کسب چند عنوان قهرمان...</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>{'text': '۱۴', 'answer_start': 173, 'answer_en...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>بیشترین بازی بدون باخت پیاپی متعلق به کدام باش...</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>{'text': 'باشگاه فوتبال انگلیسی', 'answer_star...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>باشگاه فوتبال آرسنال موفق به کسب چند عنوان قهر...</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>{'text': '۱۳', 'answer_start': 119, 'answer_en...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>باشگاه فوتبال آرسنال چند عنوان قهرمانی در جام ...</td>\n",
              "      <td>باشگاه فوتبال آرسنال (به انگلیسی: Arsenal Foo...</td>\n",
              "      <td>{'text': '۱۶', 'answer_start': 214, 'answer_en...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-d9b533d3-d3ad-4e74-aaf3-0042edc2f613')\"\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-d9b533d3-d3ad-4e74-aaf3-0042edc2f613 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-d9b533d3-d3ad-4e74-aaf3-0042edc2f613');\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": 17
        }
      ],
      "source": [
        "# Print the length of the training dataset\n",
        "print(\"Length of the training dataset: {}\".format(len(train_df)))\n",
        "\n",
        "# Display the first few rows of the training dataset\n",
        "print(\"Training dataset:\")\n",
        "train_df.head()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 18,
      "metadata": {
        "id": "JgUCQvNJHvVA",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 241
        },
        "outputId": "b4510745-39d8-407a-bccd-69425fedb932"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Length of the validation dataset: 9054\n",
            "Validation dataset:\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                            question  \\\n",
              "0  روند مناقصه برای بازی‌های المپیک تابستانی در چ...   \n",
              "1                رای‌گیری نهایی در چه سالی انجام شد؟   \n",
              "2                رای‌گیری نهایی در چه سالی انجام شد؟   \n",
              "3       رای‌گیری نهایی در کدام شهرستان‌ها برگزار شد؟   \n",
              "4       رای‌گیری نهایی در کدام شهرستان‌ها برگزار شد؟   \n",
              "\n",
              "                                             context  \\\n",
              "0  روند مناقصه برای بازی‌های المپیک تابستانی ۲۰۱۶...   \n",
              "1  روند مناقصه برای بازی‌های المپیک تابستانی ۲۰۱۶...   \n",
              "2  روند مناقصه برای بازی‌های المپیک تابستانی ۲۰۱۶...   \n",
              "3  روند مناقصه برای بازی‌های المپیک تابستانی ۲۰۱۶...   \n",
              "4  روند مناقصه برای بازی‌های المپیک تابستانی ۲۰۱۶...   \n",
              "\n",
              "                                             answers  \n",
              "0  {'text': '۱۶ مه ۲۰۰۷', 'answer_start': 56, 'an...  \n",
              "1  {'text': '۲ اکتبر ۲۰۰۹', 'answer_start': 196, ...  \n",
              "2  {'text': 'در ۲ اکتبر ۲۰۰۹', 'answer_start': 19...  \n",
              "3  {'text': 'کپنهاگ و مادرید', 'answer_start': 21...  \n",
              "4  {'text': 'در کپنهاگ و مادرید', 'answer_start':...  "
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-9c342705-324e-4664-9459-de5aa8729315\">\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>question</th>\n",
              "      <th>context</th>\n",
              "      <th>answers</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>روند مناقصه برای بازی‌های المپیک تابستانی در چ...</td>\n",
              "      <td>روند مناقصه برای بازی‌های المپیک تابستانی ۲۰۱۶...</td>\n",
              "      <td>{'text': '۱۶ مه ۲۰۰۷', 'answer_start': 56, 'an...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>رای‌گیری نهایی در چه سالی انجام شد؟</td>\n",
              "      <td>روند مناقصه برای بازی‌های المپیک تابستانی ۲۰۱۶...</td>\n",
              "      <td>{'text': '۲ اکتبر ۲۰۰۹', 'answer_start': 196, ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>رای‌گیری نهایی در چه سالی انجام شد؟</td>\n",
              "      <td>روند مناقصه برای بازی‌های المپیک تابستانی ۲۰۱۶...</td>\n",
              "      <td>{'text': 'در ۲ اکتبر ۲۰۰۹', 'answer_start': 19...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>رای‌گیری نهایی در کدام شهرستان‌ها برگزار شد؟</td>\n",
              "      <td>روند مناقصه برای بازی‌های المپیک تابستانی ۲۰۱۶...</td>\n",
              "      <td>{'text': 'کپنهاگ و مادرید', 'answer_start': 21...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>رای‌گیری نهایی در کدام شهرستان‌ها برگزار شد؟</td>\n",
              "      <td>روند مناقصه برای بازی‌های المپیک تابستانی ۲۰۱۶...</td>\n",
              "      <td>{'text': 'در کپنهاگ و مادرید', 'answer_start':...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-9c342705-324e-4664-9459-de5aa8729315')\"\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-9c342705-324e-4664-9459-de5aa8729315 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-9c342705-324e-4664-9459-de5aa8729315');\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": 18
        }
      ],
      "source": [
        "# Print the length of the validation dataset\n",
        "print(\"Length of the validation dataset: {}\".format(len(validation_df)))\n",
        "\n",
        "# Display the first few rows of the validation dataset\n",
        "print(\"Validation dataset:\")\n",
        "validation_df.head()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 19,
      "metadata": {
        "id": "-_kRkrolzfO6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 241
        },
        "outputId": "26f0a360-84c9-49e3-e0da-e3db4d3679c4"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Length of the test dataset: 8818\n",
            "Test dataset:\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                            question  \\\n",
              "0                          کتاب مقدس دین اسلام چیست؟   \n",
              "1  قرآن به صورت وحی توسط چه کسی بر پیامبر اسلام ...   \n",
              "2  قرآن به صورت وحی توسط چه کسی بر پیامبر اسلام ...   \n",
              "3                          پیامبر اسلام چه نام دارد؟   \n",
              "4  بزرگترین معجزهٔ محمد و روشن‌ترین دلیل بر پیامب...   \n",
              "\n",
              "                                             context  \\\n",
              "0  قرآن کتاب مقدس دین اسلام است و در باور مسلمان...   \n",
              "1  قرآن کتاب مقدس دین اسلام است و در باور مسلمان...   \n",
              "2  قرآن کتاب مقدس دین اسلام است و در باور مسلمان...   \n",
              "3  قرآن کتاب مقدس دین اسلام است و در باور مسلمان...   \n",
              "4  قرآن کتاب مقدس دین اسلام است و در باور مسلمان...   \n",
              "\n",
              "                                             answers  \n",
              "0  {'text': 'قرآن', 'answer_start': 0, 'answer_e...  \n",
              "1  {'text': 'جبرئیل', 'answer_start': 91, 'answe...  \n",
              "2  {'text': 'توسط جبرئیل', 'answer_start': 86, '...  \n",
              "3  {'text': 'محمد بن عبدالله', 'answer_start': 11...  \n",
              "4  {'text': 'قرآن', 'answer_start': 158, 'answer...  "
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-f1a07808-ddfa-47fa-b205-3d1ef47a1a3d\">\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>question</th>\n",
              "      <th>context</th>\n",
              "      <th>answers</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>کتاب مقدس دین اسلام چیست؟</td>\n",
              "      <td>قرآن کتاب مقدس دین اسلام است و در باور مسلمان...</td>\n",
              "      <td>{'text': 'قرآن', 'answer_start': 0, 'answer_e...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>قرآن به صورت وحی توسط چه کسی بر پیامبر اسلام ...</td>\n",
              "      <td>قرآن کتاب مقدس دین اسلام است و در باور مسلمان...</td>\n",
              "      <td>{'text': 'جبرئیل', 'answer_start': 91, 'answe...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>قرآن به صورت وحی توسط چه کسی بر پیامبر اسلام ...</td>\n",
              "      <td>قرآن کتاب مقدس دین اسلام است و در باور مسلمان...</td>\n",
              "      <td>{'text': 'توسط جبرئیل', 'answer_start': 86, '...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>پیامبر اسلام چه نام دارد؟</td>\n",
              "      <td>قرآن کتاب مقدس دین اسلام است و در باور مسلمان...</td>\n",
              "      <td>{'text': 'محمد بن عبدالله', 'answer_start': 11...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>بزرگترین معجزهٔ محمد و روشن‌ترین دلیل بر پیامب...</td>\n",
              "      <td>قرآن کتاب مقدس دین اسلام است و در باور مسلمان...</td>\n",
              "      <td>{'text': 'قرآن', 'answer_start': 158, 'answer...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-f1a07808-ddfa-47fa-b205-3d1ef47a1a3d')\"\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-f1a07808-ddfa-47fa-b205-3d1ef47a1a3d 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-f1a07808-ddfa-47fa-b205-3d1ef47a1a3d');\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": 19
        }
      ],
      "source": [
        "# Print the length of the test dataset\n",
        "print(\"Length of the test dataset: {}\".format(len(test_df)))\n",
        "\n",
        "# Display the first few rows of the test dataset\n",
        "print(\"Test dataset:\")\n",
        "test_df.head()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 20,
      "metadata": {
        "id": "8ctcIO2IP1zy"
      },
      "outputs": [],
      "source": [
        "def prepare_data_2(dataset):\n",
        "    # Extract the questions, contexts, and answers from the dataset\n",
        "    questions = dataset['question'].tolist()\n",
        "    contexts = dataset['context'].tolist()\n",
        "    answers = dataset['answers'].tolist()\n",
        "    \n",
        "    # Return a dictionary containing the prepared data\n",
        "    return {\n",
        "        'question': questions,\n",
        "        'context': contexts,\n",
        "        'answers': answers\n",
        "    }\n",
        "\n",
        "# Prepare the training dataset\n",
        "train_dataset = prepare_data_2(train_df)\n",
        "\n",
        "# Prepare the validation dataset\n",
        "val_dataset = prepare_data_2(validation_df)\n",
        "\n",
        "# Prepare the test dataset\n",
        "test_dataset = prepare_data_2(test_df)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 21,
      "metadata": {
        "id": "5-f2BVphsNqD"
      },
      "outputs": [],
      "source": [
        "# Tokenize the training data using the tokenizer\n",
        "train_data = tokenizer(train_dataset['context'], train_dataset['question'],\n",
        "                       truncation=False, padding='max_length',\n",
        "                       max_length=max_len, return_tensors='pt')\n",
        "\n",
        "# Tokenize the validation data using the tokenizer\n",
        "validation_data = tokenizer(val_dataset['context'], val_dataset['question'],\n",
        "                            truncation=False, padding='max_length',\n",
        "                            max_length=max_len, return_tensors='pt')\n",
        "\n",
        "# Tokenize the test data using the tokenizer\n",
        "test_data = tokenizer(test_dataset['context'], test_dataset['question'],\n",
        "                            truncation=False, padding='max_length',\n",
        "                            max_length=max_len, return_tensors='pt')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 22,
      "metadata": {
        "id": "wOXrEpStv3yH",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "525f446d-72bf-46fe-f3ed-ddb45319a775"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 55152/55152 [01:30<00:00, 608.04it/s]\n",
            "100%|██████████| 9054/9054 [00:15<00:00, 574.46it/s]\n",
            "100%|██████████| 8818/8818 [00:15<00:00, 563.72it/s]\n"
          ]
        }
      ],
      "source": [
        "# Function to convert character index to token index\n",
        "def char_idx_to_token_idx(tokenizer, char_idx, sentence):\n",
        "    # Create a list of binary values indicating whether each character is a space or a half-space\n",
        "    char = [0 if sentence[i] == ' ' or sentence[i] == '‌' else 1 for i in range(len(sentence))]\n",
        "    tokens = tokenizer.tokenize(sentence)\n",
        "    index = char_idx\n",
        "\n",
        "    # Adjust the index to consider half-spaces\n",
        "    for i in range(index):\n",
        "        if char[i] != 1:\n",
        "            index -= 1\n",
        "\n",
        "    counter = 0\n",
        "    flag_continue = True\n",
        "    token_index = 0\n",
        "\n",
        "    # Find the token index corresponding to the character index\n",
        "    for i in range(len(tokens)):\n",
        "        if tokens[i].startswith('##'):\n",
        "            tokens[i] = tokens[i][2:]\n",
        "        for j in range(len(tokens[i])):\n",
        "            counter += 1\n",
        "            if counter == index:\n",
        "                flag_continue = False\n",
        "                token_index = i\n",
        "                break\n",
        "        if not flag_continue:\n",
        "            break\n",
        "    \n",
        "    return token_index             \n",
        "\n",
        "# Function to add token positions for answer start and end\n",
        "def add_token_positions(tokenizer, encodings, answers, contexts):\n",
        "    target_start_list = []\n",
        "    target_end_list = []\n",
        "    for i in tqdm(range(len(answers))):\n",
        "        target_start = [0] * max_len\n",
        "        target_end = [0] * max_len\n",
        "        start_idx = answers[i]['answer_start']\n",
        "        end_idx = answers[i]['answer_end']\n",
        "        \n",
        "        if start_idx <= len(contexts[i]) and end_idx <= len(contexts[i]):\n",
        "            # Answerable question\n",
        "            if start_idx != -1 and end_idx != -1:\n",
        "                start_token_idx = char_idx_to_token_idx(tokenizer, start_idx, contexts[i]) + 1\n",
        "                end_token_idx = char_idx_to_token_idx(tokenizer, end_idx, contexts[i]) + 1\n",
        "                target_start[start_token_idx] = 1\n",
        "                target_end[end_token_idx] = 1\n",
        "                target_start_list.append(target_start)\n",
        "                target_end_list.append(target_end)\n",
        "\n",
        "            # Unanswerable question\n",
        "            else:\n",
        "                target_start[0] = 1\n",
        "                target_end[0] = 1\n",
        "                target_start_list.append(target_start)\n",
        "                target_end_list.append(target_end)\n",
        "        else:\n",
        "            continue\n",
        "    \n",
        "    # Update the encodings with the target start and end lists\n",
        "    encodings.update({'targets_start': target_start_list, 'targets_end': target_end_list})\n",
        "\n",
        "# Add token positions to the train_data encodings\n",
        "add_token_positions(tokenizer, train_data, train_dataset['answers'], train_dataset['context'])\n",
        "\n",
        "# Add token positions to the validation_data encodings\n",
        "add_token_positions(tokenizer, validation_data, val_dataset['answers'], val_dataset['context'])\n",
        "\n",
        "# Add token positions to the test_data encodings\n",
        "add_token_positions(tokenizer, test_data, test_dataset['answers'], test_dataset['context'])\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 23,
      "metadata": {
        "id": "CSTIkvemgsHg"
      },
      "outputs": [],
      "source": [
        "# Custom Dataset class\n",
        "class CustomDataset(torch.utils.data.Dataset):\n",
        "    def __init__(self, data):\n",
        "        \"\"\"\n",
        "        Initialize the CustomDataset.\n",
        "\n",
        "        Args:\n",
        "            data (dict): The data dictionary containing input tensors.\n",
        "        \"\"\"\n",
        "        self.data = data\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        \"\"\"\n",
        "        Get an item from the dataset.\n",
        "\n",
        "        Args:\n",
        "            idx (int): The index of the item.\n",
        "\n",
        "        Returns:\n",
        "            dict: A dictionary containing input tensors.\n",
        "        \"\"\"\n",
        "        return {key: torch.tensor(val[idx]) for key, val in self.data.items()}\n",
        "    \n",
        "    def __len__(self):\n",
        "        \"\"\"\n",
        "        Get the length of the dataset.\n",
        "\n",
        "        Returns:\n",
        "            int: The length of the dataset.\n",
        "        \"\"\"\n",
        "        return len(self.data['input_ids'])\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 24,
      "metadata": {
        "id": "ZlYJSrQGnc25",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a3a0a1d3-f3d3-4ba3-e5c4-801d100a0ea6"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Length of training dataset: 55152\n",
            "Length of validation dataset: 9054\n",
            "Length of test dataset: 8818\n"
          ]
        }
      ],
      "source": [
        "# Create a custom dataset for training\n",
        "train_datas = CustomDataset(train_data)\n",
        "print(\"Length of training dataset: {}\".format(len(train_datas)))\n",
        "\n",
        "# Create a data loader for training\n",
        "train_loader = torch.utils.data.DataLoader(train_datas, batch_size=32, shuffle=True)\n",
        "\n",
        "# Create a custom dataset for validation\n",
        "validation_datas = CustomDataset(validation_data)\n",
        "print(\"Length of validation dataset: {}\".format(len(validation_datas)))\n",
        "\n",
        "# Create a data loader for validation\n",
        "validation_loader = torch.utils.data.DataLoader(validation_datas, batch_size=32, shuffle=False)\n",
        "\n",
        "# Create a custom dataset for test\n",
        "test_datas = CustomDataset(test_data)\n",
        "print(\"Length of test dataset: {}\".format(len(test_datas)))\n",
        "\n",
        "# Create a data loader for test\n",
        "test_loader = torch.utils.data.DataLoader(test_datas, batch_size=32, shuffle=False)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 25,
      "metadata": {
        "id": "pMlBS8uYnyop"
      },
      "outputs": [],
      "source": [
        "MODEL_NAME_OR_PATH = 'm3hrdadfi/albert-fa-base-v2'\n",
        "\n",
        "class QAModel(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(QAModel, self).__init__()\n",
        "\n",
        "        # Initialize the BERT model\n",
        "        self.bert = BertModel.from_pretrained(MODEL_NAME_OR_PATH, return_dict=False)\n",
        "        # Add a linear layer for classification\n",
        "        self.classifier = nn.Linear(768, 2)\n",
        "    \n",
        "    def forward(self, input_ids, attention_mask, token_type_ids):\n",
        "        # Pass the input through the BERT model\n",
        "        sequence_output, pooled_output = self.bert(\n",
        "            input_ids=input_ids, \n",
        "            attention_mask=attention_mask, \n",
        "            token_type_ids=token_type_ids)\n",
        "        \n",
        "        # Apply linear layer to the BERT output\n",
        "        # Shape: (batch_size, num_tokens, 768)\n",
        "        logits = self.classifier(sequence_output)\n",
        "        # Shape: (batch_size, num_tokens, 2)\n",
        "        \n",
        "        # Split the logits into start and end logits\n",
        "        start_logits, end_logits = logits.split(1, dim=-1)\n",
        "        start_logits = start_logits.squeeze(-1)\n",
        "        end_logits = end_logits.squeeze(-1)\n",
        "        # Shape: (batch_size, num_tokens), (batch_size, num_tokens)\n",
        "\n",
        "        return start_logits, end_logits\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 26,
      "metadata": {
        "id": "l1Phrs8rs5Vo"
      },
      "outputs": [],
      "source": [
        "def loss_fn(start_logits, end_logits, start_targets, end_targets):\n",
        "    \"\"\"\n",
        "    Compute the loss function given the predicted start and end logits and the target start and end positions.\n",
        "\n",
        "    Args:\n",
        "        start_logits (torch.Tensor): Predicted start logits of shape (batch_size, num_tokens).\n",
        "        end_logits (torch.Tensor): Predicted end logits of shape (batch_size, num_tokens).\n",
        "        start_targets (torch.Tensor): Target start positions of shape (batch_size, num_tokens).\n",
        "        end_targets (torch.Tensor): Target end positions of shape (batch_size, num_tokens).\n",
        "\n",
        "    Returns:\n",
        "        torch.Tensor: Loss value.\n",
        "    \"\"\"\n",
        "    # Compute the binary cross-entropy loss for start and end logits\n",
        "    start_loss = nn.BCEWithLogitsLoss()(start_logits, start_targets.float())\n",
        "    end_loss = nn.BCEWithLogitsLoss()(end_logits, end_targets.float())\n",
        "\n",
        "    # Return the sum of the two losses\n",
        "    return start_loss + end_loss\n",
        "\n",
        "def evaluate_f1(start_pred, start_target, end_pred, end_target):\n",
        "    \"\"\"\n",
        "    Compute the F1 score given the predicted start and end positions and the target start and end positions.\n",
        "\n",
        "    Args:\n",
        "        start_pred (int): Predicted start position.\n",
        "        start_target (int): Target start position.\n",
        "        end_pred (int): Predicted end position.\n",
        "        end_target (int): Target end position.\n",
        "\n",
        "    Returns:\n",
        "        float: F1 score.\n",
        "    \"\"\"\n",
        "    # Generate arrays of tokens for prediction and target spans\n",
        "    pred = np.arange(start_pred, end_pred + 1)\n",
        "    tar = np.arange(start_target, end_target + 1)\n",
        "\n",
        "    # Compute the number of tokens shared between prediction and target\n",
        "    tp_list = list(set.intersection(*map(set, [pred, tar])))\n",
        "\n",
        "    # Compute the number of tokens in prediction not in target\n",
        "    fp_list = list(set(pred).symmetric_difference(set(tp_list)))\n",
        "\n",
        "    # Compute the number of tokens in target not in prediction\n",
        "    fn_list = list(set(tar).symmetric_difference(set(tp_list)))\n",
        "\n",
        "    tp, fp, fn = len(tp_list), len(fp_list), len(fn_list)\n",
        "\n",
        "    # Compute precision, recall, and F1 score\n",
        "    if (tp + fp) != 0:\n",
        "        precision = tp / (tp + fp)\n",
        "    else:\n",
        "        precision = 0\n",
        "\n",
        "    if (tp + fn) != 0:\n",
        "        recall = tp / (tp + fn)\n",
        "    else:\n",
        "        recall = 0\n",
        "\n",
        "    if (precision + recall) != 0:\n",
        "        f1 = (2 * precision * recall) / (precision + recall)\n",
        "    else:\n",
        "        f1 = 0\n",
        "\n",
        "    return f1\n",
        "\n",
        "\n",
        "def generate_indexes(start_logits, end_logits, N, max_index_list):\n",
        "    \"\"\"\n",
        "    Generate the start and end indexes for the predicted spans.\n",
        "\n",
        "    Args:\n",
        "        start_logits (numpy.ndarray): Predicted start logits of shape (batch_size, num_tokens).\n",
        "        end_logits (numpy.ndarray): Predicted end logits of shape (batch_size, num_tokens).\n",
        "        N (int): Number of top start and end indexes to consider.\n",
        "        max_index_list (list): List of maximum indexes for each example.\n",
        "\n",
        "    Returns:\n",
        "        tuple: Final start and end indexes for the predicted spans.\n",
        "    \"\"\"\n",
        "    output_start = start_logits\n",
        "    output_end = end_logits\n",
        "    dimension = output_start.shape[1]\n",
        "\n",
        "    list_start, list_end = [], []\n",
        "    for n in range(output_start.shape[0]):\n",
        "        start_indexes = np.arange(output_start.shape[1])\n",
        "        start_probs = output_start[n]\n",
        "        list_start.append(dict(zip(start_indexes, start_probs)))\n",
        "\n",
        "        end_indexes = np.arange(output_start.shape[1])\n",
        "        end_probs = output_end[n]\n",
        "        list_end.append(dict(zip(end_indexes, end_probs)))\n",
        "\n",
        "    sorted_start_list, sorted_end_list = [], []\n",
        "    for j in range(len(list_start)):\n",
        "        sort_start_probs = sorted(list_start[j].items(), key=lambda x: x[1], reverse=True)\n",
        "        sort_end_probs = sorted(list_end[j].items(), key=lambda x: x[1], reverse=True)\n",
        "        sorted_start_list.append(sort_start_probs)\n",
        "        sorted_end_list.append(sort_end_probs)\n",
        "\n",
        "    final_start_idx, final_end_idx = [], []\n",
        "\n",
        "    for c in range(len(list_start)):\n",
        "        start_idx, end_idx, prob = 0, 0, 0\n",
        "        for a in range(N):\n",
        "            for b in range(N):\n",
        "                if (sorted_start_list[c][a][1] + sorted_end_list[c][b][1]) > prob:\n",
        "                    if (sorted_start_list[c][a][0] <= sorted_end_list[c][b][0]) and (\n",
        "                            sorted_end_list[c][b][0] < max_index_list[c]):\n",
        "                        prob = sorted_start_list[c][a][1] + sorted_end_list[c][b][1]\n",
        "                        start_idx = sorted_start_list[c][a][0]\n",
        "                        end_idx = sorted_end_list[c][b][0]\n",
        "        final_start_idx.append(start_idx)\n",
        "        final_end_idx.append(end_idx)\n",
        "\n",
        "    return final_start_idx, final_end_idx\n",
        "\n",
        "\n",
        "def evaluate_model(start_logits, end_logits, N, max_index_list, target_start, target_end):\n",
        "    \"\"\"\n",
        "    Evaluate the model by computing the F1 score.\n",
        "\n",
        "    Args:\n",
        "        start_logits (numpy.ndarray): Predicted start logits of shape (batch_size, num_tokens).\n",
        "        end_logits (numpy.ndarray): Predicted end logits of shape (batch_size, num_tokens).\n",
        "        N (int): Number of top start and end indexes to consider.\n",
        "        max_index_list (list): List of maximum indexes for each example.\n",
        "        target_start (list): List of target start positions.\n",
        "        target_end (list): List of target end positions.\n",
        "\n",
        "    Returns:\n",
        "        float: Mean F1 score.\n",
        "    \"\"\"\n",
        "    final_start_idx, final_end_idx = generate_indexes(start_logits, end_logits, N, max_index_list)\n",
        "    f1 = []\n",
        "    for i in range(len(final_start_idx)):\n",
        "        f1.append(evaluate_f1(final_start_idx[i], target_start[i], final_end_idx[i], target_end[i]))\n",
        "\n",
        "    return np.mean(f1)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 27,
      "metadata": {
        "id": "LC-hOeXazuyi",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 191,
          "referenced_widgets": [
            "d924d290ec2c488f98d29a034c4308fe",
            "8acd136ef48044d884b133b027c07a89",
            "7e76938cec534176ba49da562c332e11",
            "f19dfe296e1f4a93bd19bccb63cc69fc",
            "ead5fbeaec0641069ff245bf7828a6db",
            "910dafa67e0949b0a6371084e808b192",
            "0d350638b1214cd5953059a9f8ae0ab4",
            "4630ace2df564055a22aa1e1b4d043c0",
            "85168b8f8b6140a7ba53e595f4575b89",
            "f8263188caf145e8a20f56459bb4f1e0",
            "022f884ecd0e48c3a451d72412eb58ed"
          ]
        },
        "outputId": "446261a9-7fdd-4673-ba57-1003c6649057"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Device: cuda:0\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "You are using a model of type albert to instantiate a model of type bert. This is not supported for all configurations of models and can yield errors.\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading pytorch_model.bin:   0%|          | 0.00/73.1M [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "d924d290ec2c488f98d29a034c4308fe"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Some weights of the model checkpoint at m3hrdadfi/albert-fa-base-v2 were not used when initializing BertModel: ['albert.encoder.albert_layer_groups.0.albert_layers.0.ffn.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.query.weight', 'albert.encoder.embedding_hidden_mapping_in.weight', 'predictions.decoder.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.key.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.query.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.key.bias', 'albert.pooler.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.LayerNorm.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.full_layer_layer_norm.bias', 'albert.encoder.embedding_hidden_mapping_in.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.dense.weight', 'albert.embeddings.LayerNorm.bias', 'predictions.dense.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.ffn_output.bias', 'albert.embeddings.token_type_embeddings.weight', 'albert.pooler.weight', 'predictions.decoder.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.dense.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.full_layer_layer_norm.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.value.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.ffn.bias', 'predictions.LayerNorm.weight', 'sop_classifier.classifier.weight', 'predictions.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.ffn_output.weight', 'sop_classifier.classifier.bias', 'albert.embeddings.position_embeddings.weight', 'predictions.LayerNorm.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.LayerNorm.bias', 'predictions.dense.bias', 'albert.embeddings.LayerNorm.weight', 'albert.embeddings.word_embeddings.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.value.weight']\n",
            "- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
            "- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
            "Some weights of BertModel were not initialized from the model checkpoint at m3hrdadfi/albert-fa-base-v2 and are newly initialized: ['encoder.layer.0.attention.output.dense.bias', 'encoder.layer.2.output.LayerNorm.weight', 'encoder.layer.0.attention.output.LayerNorm.weight', 'encoder.layer.10.attention.self.query.weight', 'encoder.layer.1.attention.self.key.weight', 'encoder.layer.5.attention.self.value.bias', 'encoder.layer.10.output.dense.weight', 'encoder.layer.11.attention.self.value.bias', 'encoder.layer.7.output.dense.weight', 'encoder.layer.5.attention.output.dense.bias', 'encoder.layer.4.attention.self.key.bias', 'encoder.layer.11.output.LayerNorm.weight', 'encoder.layer.8.attention.self.query.weight', 'encoder.layer.10.attention.self.key.weight', 'encoder.layer.7.attention.output.LayerNorm.weight', 'encoder.layer.4.intermediate.dense.bias', 'encoder.layer.0.output.dense.weight', 'encoder.layer.5.output.LayerNorm.weight', 'encoder.layer.10.attention.self.key.bias', 'encoder.layer.11.output.LayerNorm.bias', 'encoder.layer.3.attention.output.dense.bias', 'encoder.layer.9.attention.output.LayerNorm.bias', 'encoder.layer.3.output.dense.weight', 'encoder.layer.9.output.dense.bias', 'encoder.layer.0.attention.self.query.weight', 'embeddings.LayerNorm.bias', 'embeddings.token_type_embeddings.weight', 'encoder.layer.6.output.dense.weight', 'encoder.layer.9.attention.output.LayerNorm.weight', 'encoder.layer.1.attention.self.query.weight', 'encoder.layer.1.attention.output.dense.weight', 'encoder.layer.5.output.LayerNorm.bias', 'encoder.layer.6.attention.output.LayerNorm.weight', 'encoder.layer.8.output.dense.weight', 'encoder.layer.9.attention.output.dense.weight', 'encoder.layer.9.output.LayerNorm.bias', 'encoder.layer.2.attention.self.query.weight', 'encoder.layer.3.attention.self.key.weight', 'encoder.layer.0.attention.output.dense.weight', 'encoder.layer.2.output.dense.weight', 'encoder.layer.5.attention.output.dense.weight', 'encoder.layer.1.output.dense.weight', 'encoder.layer.8.attention.output.LayerNorm.bias', 'encoder.layer.4.attention.output.dense.weight', 'encoder.layer.3.attention.self.value.bias', 'encoder.layer.11.attention.output.LayerNorm.weight', 'encoder.layer.7.attention.self.value.bias', 'encoder.layer.4.attention.output.LayerNorm.weight', 'encoder.layer.1.attention.output.LayerNorm.bias', 'encoder.layer.7.attention.self.query.bias', 'embeddings.LayerNorm.weight', 'encoder.layer.2.attention.output.LayerNorm.weight', 'encoder.layer.5.intermediate.dense.bias', 'encoder.layer.3.attention.self.query.bias', 'encoder.layer.6.output.LayerNorm.weight', 'encoder.layer.8.intermediate.dense.bias', 'encoder.layer.9.attention.self.key.weight', 'encoder.layer.4.output.dense.weight', 'encoder.layer.2.attention.self.key.bias', 'encoder.layer.11.intermediate.dense.weight', 'encoder.layer.4.attention.output.LayerNorm.bias', 'encoder.layer.1.output.LayerNorm.bias', 'encoder.layer.6.attention.self.query.weight', 'encoder.layer.4.attention.self.query.bias', 'encoder.layer.6.output.dense.bias', 'encoder.layer.8.attention.self.key.bias', 'encoder.layer.3.attention.self.key.bias', 'encoder.layer.3.intermediate.dense.bias', 'encoder.layer.6.attention.self.query.bias', 'encoder.layer.2.attention.self.value.bias', 'encoder.layer.8.attention.self.value.bias', 'encoder.layer.0.attention.self.key.bias', 'encoder.layer.2.attention.self.value.weight', 'encoder.layer.1.attention.self.value.bias', 'encoder.layer.11.attention.self.key.bias', 'encoder.layer.8.attention.output.LayerNorm.weight', 'encoder.layer.1.attention.self.value.weight', 'encoder.layer.5.output.dense.bias', 'encoder.layer.8.output.LayerNorm.bias', 'encoder.layer.10.attention.output.LayerNorm.weight', 'encoder.layer.11.intermediate.dense.bias', 'encoder.layer.11.output.dense.weight', 'encoder.layer.10.attention.output.LayerNorm.bias', 'encoder.layer.6.attention.self.value.bias', 'encoder.layer.9.attention.self.key.bias', 'encoder.layer.10.intermediate.dense.bias', 'encoder.layer.2.output.dense.bias', 'encoder.layer.0.attention.self.key.weight', 'pooler.dense.bias', 'encoder.layer.7.output.LayerNorm.weight', 'encoder.layer.0.intermediate.dense.bias', 'encoder.layer.6.intermediate.dense.bias', 'encoder.layer.9.output.dense.weight', 'encoder.layer.5.attention.output.LayerNorm.bias', 'encoder.layer.2.intermediate.dense.bias', 'encoder.layer.6.attention.output.dense.bias', 'encoder.layer.1.attention.output.dense.bias', 'encoder.layer.6.intermediate.dense.weight', 'encoder.layer.6.attention.self.key.bias', 'encoder.layer.6.attention.output.dense.weight', 'encoder.layer.4.attention.self.query.weight', 'encoder.layer.7.attention.self.key.weight', 'encoder.layer.2.intermediate.dense.weight', 'encoder.layer.7.output.dense.bias', 'encoder.layer.6.attention.output.LayerNorm.bias', 'encoder.layer.0.attention.self.value.bias', 'encoder.layer.4.attention.self.value.weight', 'encoder.layer.1.attention.output.LayerNorm.weight', 'encoder.layer.9.intermediate.dense.bias', 'encoder.layer.1.intermediate.dense.bias', 'encoder.layer.2.output.LayerNorm.bias', 'encoder.layer.4.output.LayerNorm.bias', 'encoder.layer.8.output.dense.bias', 'embeddings.word_embeddings.weight', 'encoder.layer.5.output.dense.weight', 'encoder.layer.11.attention.output.dense.bias', 'encoder.layer.10.output.LayerNorm.weight', 'encoder.layer.0.attention.output.LayerNorm.bias', 'encoder.layer.7.intermediate.dense.bias', 'encoder.layer.10.output.LayerNorm.bias', 'encoder.layer.6.attention.self.key.weight', 'encoder.layer.2.attention.output.dense.weight', 'encoder.layer.0.output.LayerNorm.weight', 'encoder.layer.8.intermediate.dense.weight', 'encoder.layer.7.attention.output.dense.weight', 'encoder.layer.3.attention.output.LayerNorm.weight', 'encoder.layer.3.output.dense.bias', 'encoder.layer.1.attention.self.query.bias', 'encoder.layer.2.attention.self.key.weight', 'encoder.layer.3.attention.self.query.weight', 'encoder.layer.4.intermediate.dense.weight', 'encoder.layer.5.attention.self.query.bias', 'encoder.layer.5.intermediate.dense.weight', 'encoder.layer.7.attention.output.dense.bias', 'encoder.layer.9.attention.self.value.bias', 'encoder.layer.10.attention.self.query.bias', 'encoder.layer.8.attention.output.dense.bias', 'encoder.layer.0.intermediate.dense.weight', 'encoder.layer.7.output.LayerNorm.bias', 'encoder.layer.10.attention.self.value.weight', 'encoder.layer.7.attention.self.value.weight', 'encoder.layer.1.output.LayerNorm.weight', 'encoder.layer.1.attention.self.key.bias', 'encoder.layer.6.output.LayerNorm.bias', 'encoder.layer.9.attention.self.query.weight', 'encoder.layer.5.attention.self.key.weight', 'encoder.layer.9.attention.self.query.bias', 'encoder.layer.4.attention.self.key.weight', 'encoder.layer.4.output.dense.bias', 'encoder.layer.10.attention.output.dense.bias', 'encoder.layer.11.attention.output.LayerNorm.bias', 'encoder.layer.8.attention.self.query.bias', 'encoder.layer.0.attention.self.query.bias', 'encoder.layer.5.attention.output.LayerNorm.weight', 'encoder.layer.11.attention.output.dense.weight', 'encoder.layer.7.attention.self.query.weight', 'encoder.layer.6.attention.self.value.weight', 'encoder.layer.9.intermediate.dense.weight', 'pooler.dense.weight', 'embeddings.position_embeddings.weight', 'encoder.layer.0.output.dense.bias', 'encoder.layer.3.attention.output.dense.weight', 'encoder.layer.3.output.LayerNorm.weight', 'encoder.layer.4.attention.output.dense.bias', 'encoder.layer.2.attention.self.query.bias', 'encoder.layer.3.attention.self.value.weight', 'encoder.layer.7.attention.self.key.bias', 'encoder.layer.9.attention.self.value.weight', 'encoder.layer.10.output.dense.bias', 'encoder.layer.10.attention.output.dense.weight', 'encoder.layer.0.attention.self.value.weight', 'encoder.layer.9.attention.output.dense.bias', 'encoder.layer.5.attention.self.query.weight', 'encoder.layer.8.attention.self.value.weight', 'encoder.layer.11.attention.self.key.weight', 'encoder.layer.10.intermediate.dense.weight', 'encoder.layer.4.attention.self.value.bias', 'encoder.layer.2.attention.output.dense.bias', 'encoder.layer.7.intermediate.dense.weight', 'encoder.layer.4.output.LayerNorm.weight', 'encoder.layer.8.output.LayerNorm.weight', 'encoder.layer.5.attention.self.key.bias', 'encoder.layer.5.attention.self.value.weight', 'encoder.layer.1.intermediate.dense.weight', 'encoder.layer.2.attention.output.LayerNorm.bias', 'encoder.layer.3.intermediate.dense.weight', 'encoder.layer.11.attention.self.query.weight', 'encoder.layer.8.attention.self.key.weight', 'encoder.layer.11.output.dense.bias', 'encoder.layer.3.output.LayerNorm.bias', 'encoder.layer.7.attention.output.LayerNorm.bias', 'encoder.layer.11.attention.self.query.bias', 'encoder.layer.8.attention.output.dense.weight', 'encoder.layer.11.attention.self.value.weight', 'encoder.layer.0.output.LayerNorm.bias', 'encoder.layer.10.attention.self.value.bias', 'encoder.layer.3.attention.output.LayerNorm.bias', 'encoder.layer.9.output.LayerNorm.weight', 'encoder.layer.1.output.dense.bias']\n",
            "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
          ]
        }
      ],
      "source": [
        "import warnings\n",
        "warnings.filterwarnings('ignore')\n",
        "\n",
        "# Clear GPU cache\n",
        "torch.cuda.empty_cache()\n",
        "\n",
        "# Check if CUDA is available and set the device accordingly\n",
        "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
        "print(\"Device:\", device)\n",
        "\n",
        "# Create an instance of the QAModel\n",
        "qa_model = QAModel()\n",
        "qa_model.to(device)  # Move the model to the appropriate device\n",
        "qa_model.train()  # Set the model in training mode\n",
        "\n",
        "# Create an instance of the AdamW optimizer with a learning rate of 5e-5\n",
        "optimizer = AdamW(qa_model.parameters(), lr=5e-5)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 28,
      "metadata": {
        "id": "TjziJAFq4Wpa"
      },
      "outputs": [],
      "source": [
        "MODEL_NAME_OR_PATH = 'm3hrdadfi/albert-fa-base-v2'\n",
        "\n",
        "from transformers import PreTrainedModel\n",
        "\n",
        "class QAModel2(PreTrainedModel):\n",
        "    \"\"\"\n",
        "    QA Model based on BERT for sequence classification.\n",
        "    \"\"\"\n",
        "\n",
        "    def __init__(self, config):\n",
        "        super(QAModel2, self).__init__(config)\n",
        "        self.bert = BertModel.from_pretrained(MODEL_NAME_OR_PATH, return_dict=False)\n",
        "        # self.dropout = nn.Dropout(p=dropout_rate)\n",
        "        self.classifier = nn.Linear(config.hidden_size, config.num_labels)\n",
        "\n",
        "    def forward(self, input_ids, attention_mask, token_type_ids):\n",
        "        \"\"\"\n",
        "        Forward pass of the QA model.\n",
        "        \n",
        "        Args:\n",
        "            input_ids (tensor): Input tensor of shape (batch_size, sequence_length).\n",
        "            attention_mask (tensor): Attention mask tensor of shape (batch_size, sequence_length).\n",
        "            token_type_ids (tensor): Token type IDs tensor of shape (batch_size, sequence_length).\n",
        "        \n",
        "        Returns:\n",
        "            start_logits (tensor): Start logits tensor of shape (batch_size, sequence_length).\n",
        "            end_logits (tensor): End logits tensor of shape (batch_size, sequence_length).\n",
        "        \"\"\"\n",
        "        sequence_output, pooled_output = self.bert(\n",
        "            input_ids=input_ids,\n",
        "            attention_mask=attention_mask,\n",
        "            token_type_ids=token_type_ids\n",
        "        )\n",
        "        \n",
        "        # Apply dropout to the sequence output\n",
        "        # sequence_output = self.dropout(sequence_output)\n",
        "\n",
        "        # (batch_size, num_tokens, 768)\n",
        "        logits = self.classifier(sequence_output)\n",
        "        # (batch_size, num_tokens, 2)\n",
        "        \n",
        "        # Split the logits into start and end logits\n",
        "        start_logits, end_logits = logits.split(1, dim=-1)\n",
        "        start_logits = start_logits.squeeze(-1)\n",
        "        end_logits = end_logits.squeeze(-1)\n",
        "        # (batch_size, num_tokens), (batch_size, num_tokens)\n",
        "\n",
        "        return start_logits, end_logits\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 29,
      "metadata": {
        "id": "8b_THEfbYvjt",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7d259e77-6052-47a4-b480-08c833913cc1"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "You are using a model of type albert to instantiate a model of type bert. This is not supported for all configurations of models and can yield errors.\n",
            "Some weights of the model checkpoint at m3hrdadfi/albert-fa-base-v2 were not used when initializing BertModel: ['albert.encoder.albert_layer_groups.0.albert_layers.0.ffn.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.query.weight', 'albert.encoder.embedding_hidden_mapping_in.weight', 'predictions.decoder.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.key.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.query.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.key.bias', 'albert.pooler.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.LayerNorm.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.full_layer_layer_norm.bias', 'albert.encoder.embedding_hidden_mapping_in.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.dense.weight', 'albert.embeddings.LayerNorm.bias', 'predictions.dense.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.ffn_output.bias', 'albert.embeddings.token_type_embeddings.weight', 'albert.pooler.weight', 'predictions.decoder.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.dense.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.full_layer_layer_norm.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.value.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.ffn.bias', 'predictions.LayerNorm.weight', 'sop_classifier.classifier.weight', 'predictions.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.ffn_output.weight', 'sop_classifier.classifier.bias', 'albert.embeddings.position_embeddings.weight', 'predictions.LayerNorm.bias', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.LayerNorm.bias', 'predictions.dense.bias', 'albert.embeddings.LayerNorm.weight', 'albert.embeddings.word_embeddings.weight', 'albert.encoder.albert_layer_groups.0.albert_layers.0.attention.value.weight']\n",
            "- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
            "- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
            "Some weights of BertModel were not initialized from the model checkpoint at m3hrdadfi/albert-fa-base-v2 and are newly initialized: ['encoder.layer.0.attention.output.dense.bias', 'encoder.layer.2.output.LayerNorm.weight', 'encoder.layer.0.attention.output.LayerNorm.weight', 'encoder.layer.10.attention.self.query.weight', 'encoder.layer.1.attention.self.key.weight', 'encoder.layer.5.attention.self.value.bias', 'encoder.layer.10.output.dense.weight', 'encoder.layer.11.attention.self.value.bias', 'encoder.layer.7.output.dense.weight', 'encoder.layer.5.attention.output.dense.bias', 'encoder.layer.4.attention.self.key.bias', 'encoder.layer.11.output.LayerNorm.weight', 'encoder.layer.8.attention.self.query.weight', 'encoder.layer.10.attention.self.key.weight', 'encoder.layer.7.attention.output.LayerNorm.weight', 'encoder.layer.4.intermediate.dense.bias', 'encoder.layer.0.output.dense.weight', 'encoder.layer.5.output.LayerNorm.weight', 'encoder.layer.10.attention.self.key.bias', 'encoder.layer.11.output.LayerNorm.bias', 'encoder.layer.3.attention.output.dense.bias', 'encoder.layer.9.attention.output.LayerNorm.bias', 'encoder.layer.3.output.dense.weight', 'encoder.layer.9.output.dense.bias', 'encoder.layer.0.attention.self.query.weight', 'embeddings.LayerNorm.bias', 'embeddings.token_type_embeddings.weight', 'encoder.layer.6.output.dense.weight', 'encoder.layer.9.attention.output.LayerNorm.weight', 'encoder.layer.1.attention.self.query.weight', 'encoder.layer.1.attention.output.dense.weight', 'encoder.layer.5.output.LayerNorm.bias', 'encoder.layer.6.attention.output.LayerNorm.weight', 'encoder.layer.8.output.dense.weight', 'encoder.layer.9.attention.output.dense.weight', 'encoder.layer.9.output.LayerNorm.bias', 'encoder.layer.2.attention.self.query.weight', 'encoder.layer.3.attention.self.key.weight', 'encoder.layer.0.attention.output.dense.weight', 'encoder.layer.2.output.dense.weight', 'encoder.layer.5.attention.output.dense.weight', 'encoder.layer.1.output.dense.weight', 'encoder.layer.8.attention.output.LayerNorm.bias', 'encoder.layer.4.attention.output.dense.weight', 'encoder.layer.3.attention.self.value.bias', 'encoder.layer.11.attention.output.LayerNorm.weight', 'encoder.layer.7.attention.self.value.bias', 'encoder.layer.4.attention.output.LayerNorm.weight', 'encoder.layer.1.attention.output.LayerNorm.bias', 'encoder.layer.7.attention.self.query.bias', 'embeddings.LayerNorm.weight', 'encoder.layer.2.attention.output.LayerNorm.weight', 'encoder.layer.5.intermediate.dense.bias', 'encoder.layer.3.attention.self.query.bias', 'encoder.layer.6.output.LayerNorm.weight', 'encoder.layer.8.intermediate.dense.bias', 'encoder.layer.9.attention.self.key.weight', 'encoder.layer.4.output.dense.weight', 'encoder.layer.2.attention.self.key.bias', 'encoder.layer.11.intermediate.dense.weight', 'encoder.layer.4.attention.output.LayerNorm.bias', 'encoder.layer.1.output.LayerNorm.bias', 'encoder.layer.6.attention.self.query.weight', 'encoder.layer.4.attention.self.query.bias', 'encoder.layer.6.output.dense.bias', 'encoder.layer.8.attention.self.key.bias', 'encoder.layer.3.attention.self.key.bias', 'encoder.layer.3.intermediate.dense.bias', 'encoder.layer.6.attention.self.query.bias', 'encoder.layer.2.attention.self.value.bias', 'encoder.layer.8.attention.self.value.bias', 'encoder.layer.0.attention.self.key.bias', 'encoder.layer.2.attention.self.value.weight', 'encoder.layer.1.attention.self.value.bias', 'encoder.layer.11.attention.self.key.bias', 'encoder.layer.8.attention.output.LayerNorm.weight', 'encoder.layer.1.attention.self.value.weight', 'encoder.layer.5.output.dense.bias', 'encoder.layer.8.output.LayerNorm.bias', 'encoder.layer.10.attention.output.LayerNorm.weight', 'encoder.layer.11.intermediate.dense.bias', 'encoder.layer.11.output.dense.weight', 'encoder.layer.10.attention.output.LayerNorm.bias', 'encoder.layer.6.attention.self.value.bias', 'encoder.layer.9.attention.self.key.bias', 'encoder.layer.10.intermediate.dense.bias', 'encoder.layer.2.output.dense.bias', 'encoder.layer.0.attention.self.key.weight', 'pooler.dense.bias', 'encoder.layer.7.output.LayerNorm.weight', 'encoder.layer.0.intermediate.dense.bias', 'encoder.layer.6.intermediate.dense.bias', 'encoder.layer.9.output.dense.weight', 'encoder.layer.5.attention.output.LayerNorm.bias', 'encoder.layer.2.intermediate.dense.bias', 'encoder.layer.6.attention.output.dense.bias', 'encoder.layer.1.attention.output.dense.bias', 'encoder.layer.6.intermediate.dense.weight', 'encoder.layer.6.attention.self.key.bias', 'encoder.layer.6.attention.output.dense.weight', 'encoder.layer.4.attention.self.query.weight', 'encoder.layer.7.attention.self.key.weight', 'encoder.layer.2.intermediate.dense.weight', 'encoder.layer.7.output.dense.bias', 'encoder.layer.6.attention.output.LayerNorm.bias', 'encoder.layer.0.attention.self.value.bias', 'encoder.layer.4.attention.self.value.weight', 'encoder.layer.1.attention.output.LayerNorm.weight', 'encoder.layer.9.intermediate.dense.bias', 'encoder.layer.1.intermediate.dense.bias', 'encoder.layer.2.output.LayerNorm.bias', 'encoder.layer.4.output.LayerNorm.bias', 'encoder.layer.8.output.dense.bias', 'embeddings.word_embeddings.weight', 'encoder.layer.5.output.dense.weight', 'encoder.layer.11.attention.output.dense.bias', 'encoder.layer.10.output.LayerNorm.weight', 'encoder.layer.0.attention.output.LayerNorm.bias', 'encoder.layer.7.intermediate.dense.bias', 'encoder.layer.10.output.LayerNorm.bias', 'encoder.layer.6.attention.self.key.weight', 'encoder.layer.2.attention.output.dense.weight', 'encoder.layer.0.output.LayerNorm.weight', 'encoder.layer.8.intermediate.dense.weight', 'encoder.layer.7.attention.output.dense.weight', 'encoder.layer.3.attention.output.LayerNorm.weight', 'encoder.layer.3.output.dense.bias', 'encoder.layer.1.attention.self.query.bias', 'encoder.layer.2.attention.self.key.weight', 'encoder.layer.3.attention.self.query.weight', 'encoder.layer.4.intermediate.dense.weight', 'encoder.layer.5.attention.self.query.bias', 'encoder.layer.5.intermediate.dense.weight', 'encoder.layer.7.attention.output.dense.bias', 'encoder.layer.9.attention.self.value.bias', 'encoder.layer.10.attention.self.query.bias', 'encoder.layer.8.attention.output.dense.bias', 'encoder.layer.0.intermediate.dense.weight', 'encoder.layer.7.output.LayerNorm.bias', 'encoder.layer.10.attention.self.value.weight', 'encoder.layer.7.attention.self.value.weight', 'encoder.layer.1.output.LayerNorm.weight', 'encoder.layer.1.attention.self.key.bias', 'encoder.layer.6.output.LayerNorm.bias', 'encoder.layer.9.attention.self.query.weight', 'encoder.layer.5.attention.self.key.weight', 'encoder.layer.9.attention.self.query.bias', 'encoder.layer.4.attention.self.key.weight', 'encoder.layer.4.output.dense.bias', 'encoder.layer.10.attention.output.dense.bias', 'encoder.layer.11.attention.output.LayerNorm.bias', 'encoder.layer.8.attention.self.query.bias', 'encoder.layer.0.attention.self.query.bias', 'encoder.layer.5.attention.output.LayerNorm.weight', 'encoder.layer.11.attention.output.dense.weight', 'encoder.layer.7.attention.self.query.weight', 'encoder.layer.6.attention.self.value.weight', 'encoder.layer.9.intermediate.dense.weight', 'pooler.dense.weight', 'embeddings.position_embeddings.weight', 'encoder.layer.0.output.dense.bias', 'encoder.layer.3.attention.output.dense.weight', 'encoder.layer.3.output.LayerNorm.weight', 'encoder.layer.4.attention.output.dense.bias', 'encoder.layer.2.attention.self.query.bias', 'encoder.layer.3.attention.self.value.weight', 'encoder.layer.7.attention.self.key.bias', 'encoder.layer.9.attention.self.value.weight', 'encoder.layer.10.output.dense.bias', 'encoder.layer.10.attention.output.dense.weight', 'encoder.layer.0.attention.self.value.weight', 'encoder.layer.9.attention.output.dense.bias', 'encoder.layer.5.attention.self.query.weight', 'encoder.layer.8.attention.self.value.weight', 'encoder.layer.11.attention.self.key.weight', 'encoder.layer.10.intermediate.dense.weight', 'encoder.layer.4.attention.self.value.bias', 'encoder.layer.2.attention.output.dense.bias', 'encoder.layer.7.intermediate.dense.weight', 'encoder.layer.4.output.LayerNorm.weight', 'encoder.layer.8.output.LayerNorm.weight', 'encoder.layer.5.attention.self.key.bias', 'encoder.layer.5.attention.self.value.weight', 'encoder.layer.1.intermediate.dense.weight', 'encoder.layer.2.attention.output.LayerNorm.bias', 'encoder.layer.3.intermediate.dense.weight', 'encoder.layer.11.attention.self.query.weight', 'encoder.layer.8.attention.self.key.weight', 'encoder.layer.11.output.dense.bias', 'encoder.layer.3.output.LayerNorm.bias', 'encoder.layer.7.attention.output.LayerNorm.bias', 'encoder.layer.11.attention.self.query.bias', 'encoder.layer.8.attention.output.dense.weight', 'encoder.layer.11.attention.self.value.weight', 'encoder.layer.0.output.LayerNorm.bias', 'encoder.layer.10.attention.self.value.bias', 'encoder.layer.3.attention.output.LayerNorm.bias', 'encoder.layer.9.output.LayerNorm.weight', 'encoder.layer.1.output.dense.bias']\n",
            "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
          ]
        }
      ],
      "source": [
        "from transformers import AutoConfig, AutoTokenizer\n",
        "\n",
        "# Clear GPU cache\n",
        "torch.cuda.empty_cache()\n",
        "\n",
        "# Determine the device (GPU or CPU)\n",
        "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
        "\n",
        "# Load the configuration for the BERT model\n",
        "config = AutoConfig.from_pretrained(\"m3hrdadfi/albert-fa-base-v2\")\n",
        "\n",
        "# Create an instance of the QAModel2 using the loaded configuration\n",
        "model = QAModel2(config)\n",
        "\n",
        "# Move the model to the appropriate device (GPU or CPU)\n",
        "model.to(device)\n",
        "\n",
        "# Set the model to training mode\n",
        "model.train()\n",
        "\n",
        "# Define the optimizer\n",
        "optim = AdamW(model.parameters(), lr=5e-5)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 30,
      "metadata": {
        "id": "zyabIN0CfhiA",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 537
        },
        "outputId": "04e2f263-e2cb-4d46-ef9d-825332a41fda"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 1 - Training: 100%|██████████| 1724/1724 [41:45<00:00,  1.45s/it]\n",
            "Epoch 1 - Validation: 100%|██████████| 283/283 [12:17<00:00,  2.61s/it]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model Saved in Google Drive!\n",
            "Epoch 1: Train Loss: 0.04189, Valid Loss: 0.04076, Exact Match: 19.16%, F1: 19.16%\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 2 - Training: 100%|██████████| 1724/1724 [41:45<00:00,  1.45s/it]\n",
            "Epoch 2 - Validation: 100%|██████████| 283/283 [12:10<00:00,  2.58s/it]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model Saved in Google Drive!\n",
            "Epoch 2: Train Loss: 0.03870, Valid Loss: 0.04065, Exact Match: 19.16%, F1: 19.16%\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 3 - Training:   4%|▍         | 75/1724 [01:50<40:37,  1.48s/it]\n"
          ]
        },
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-30-f0ac671e906e>\u001b[0m in \u001b[0;36m<cell line: 21>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     47\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     48\u001b[0m         \u001b[0;31m# Backward pass\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 49\u001b[0;31m         \u001b[0mloss\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     50\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     51\u001b[0m         \u001b[0;31m# Update weights\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/torch/_tensor.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(self, gradient, retain_graph, create_graph, inputs)\u001b[0m\n\u001b[1;32m    485\u001b[0m                 \u001b[0minputs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    486\u001b[0m             )\n\u001b[0;32m--> 487\u001b[0;31m         torch.autograd.backward(\n\u001b[0m\u001b[1;32m    488\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgradient\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    489\u001b[0m         )\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/torch/autograd/__init__.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(tensors, grad_tensors, retain_graph, create_graph, grad_variables, inputs)\u001b[0m\n\u001b[1;32m    198\u001b[0m     \u001b[0;31m# some Python versions print out the first line of a multi-line function\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    199\u001b[0m     \u001b[0;31m# calls in the traceback and some print out the last line\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 200\u001b[0;31m     Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass\n\u001b[0m\u001b[1;32m    201\u001b[0m         \u001b[0mtensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad_tensors_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    202\u001b[0m         allow_unreachable=True, accumulate_grad=True)  # Calls into the C++ engine to run the backward pass\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Set the number of epochs\n",
        "n_epochs = 5\n",
        "\n",
        "# Get the number of batches in the training and validation sets\n",
        "n_train_batches = len(train_loader)\n",
        "n_validation_batches = len(validation_loader)\n",
        "\n",
        "# Initialize softmax function\n",
        "softmax = torch.nn.Softmax(dim=1)\n",
        "\n",
        "# Initialize best validation loss\n",
        "best_valid_loss = float('inf')\n",
        "\n",
        "# Initialize lists to store loss, EM, and F1 scores for each epoch\n",
        "train_losses, valid_losses = [], []\n",
        "exact_match_scores, f1_scores = [], []\n",
        "\n",
        "# Main training loop\n",
        "for epoch in range(n_epochs):\n",
        "  \n",
        "    loop = tqdm(train_loader)\n",
        "    train_running_loss, validation_running_loss = 0.0, 0.0\n",
        "    exact_match = []\n",
        "    f1 = []\n",
        "    \n",
        "    # Training phase\n",
        "    for batch in loop:\n",
        "       \n",
        "        # Zero the gradients\n",
        "        optim.zero_grad()\n",
        "        \n",
        "        # Move data to device\n",
        "        input_ids = batch['input_ids'].to(device)\n",
        "        attention_mask = batch['attention_mask'].to(device)\n",
        "        token_type_ids = batch['token_type_ids'].to(device)\n",
        "        y_start = batch['targets_start'].to(device)\n",
        "        y_end = batch['targets_end'].to(device)\n",
        "        \n",
        "        # Forward pass\n",
        "        out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)  \n",
        "        \n",
        "        # Compute loss\n",
        "        loss = loss_fn(out_start, out_end, y_start, y_end)\n",
        "        train_running_loss += loss\n",
        "        \n",
        "        # Backward pass\n",
        "        loss.backward()\n",
        "        \n",
        "        # Update weights\n",
        "        optim.step()\n",
        "        \n",
        "        loop.set_description(f'Epoch {epoch+1} - Training')\n",
        "    \n",
        "    # Validation phase\n",
        "    with torch.no_grad():  \n",
        "        loop2 = tqdm(validation_loader)\n",
        "        for content in loop2:\n",
        "            \n",
        "            # Move data to device\n",
        "            input_ids = content['input_ids'].to(device)\n",
        "            temp_ids = input_ids.cpu().data.numpy().tolist()\n",
        "            max_ind = [temp_ids[i].index(4) if 4 in temp_ids[i] else -1 for i in range(len(temp_ids))]  # index of first [sep]\n",
        "            attention_mask = content['attention_mask'].to(device)\n",
        "            token_type_ids = content['token_type_ids'].to(device)\n",
        "            y_start = content['targets_start'].to(device)\n",
        "            y_end = content['targets_end'].to(device)\n",
        "            \n",
        "            # Forward pass\n",
        "            out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)\n",
        "            start_pred = softmax(out_start)\n",
        "            end_pred = softmax(out_end)\n",
        "            \n",
        "            # Compute loss\n",
        "            loss2 = loss_fn(out_start, out_end, y_start, y_end)\n",
        "            validation_running_loss += loss2\n",
        "            \n",
        "            # Compute EM and F1 scores\n",
        "            start_label = np.nonzero(y_start).cpu().data.numpy()\n",
        "            end_label = np.nonzero(y_end).cpu().data.numpy()\n",
        "            tensor_label = torch.stack((torch.tensor(start_label[:, 1]), torch.tensor(end_label[:, 1])), -1)\n",
        "            start_log, end_log = generate_indexes(start_pred, end_pred, N, max_ind)\n",
        "            start_log = np.array(start_log)\n",
        "            end_log = np.array(end_log)\n",
        "            tensor_pred = torch.stack((torch.tensor(start_log), torch.tensor(end_log)), -1)\n",
        "            ex_ma = sum([1 if (tensor_pred[i][0] == tensor_label[i][0] and tensor_pred[i][1] == tensor_label[i][1]) else 0 for i in range(0, len(tensor_pred))]) / len(tensor_pred)\n",
        "            exact_match.append(ex_ma)\n",
        "            f1.append(evaluate_model(start_pred.cpu().data.numpy(), end_pred.cpu().data.numpy(), N, max_ind, start_label[:, 1], end_label[:, 1]))\n",
        "            loop2.set_description(f'Epoch {epoch+1} - Validation')\n",
        "\n",
        "    # Compute average losses and scores\n",
        "    train_loss = train_running_loss / n_train_batches\n",
        "    valid_loss = validation_running_loss / n_validation_batches\n",
        "    \n",
        "    # Append losses to the lists\n",
        "    train_losses.append(train_loss)\n",
        "    valid_losses.append(valid_loss)\n",
        "    \n",
        "    # Update best validation loss and save model weights\n",
        "    if valid_loss < best_valid_loss:\n",
        "        best_valid_loss = valid_loss\n",
        "        torch.save(model.state_dict(), '/content/drive/MyDrive/HW5/Q2/Models/ALBERT1.pt')\n",
        "        print('Model Saved in Google Drive!')\n",
        "    \n",
        "    # Compute and append average EM and F1 scores\n",
        "    exact_match_score = 100 * np.mean(exact_match)\n",
        "    f1_score = 100 * np.mean(f1)\n",
        "    exact_match_scores.append(exact_match_score)\n",
        "    f1_scores.append(f1_score)\n",
        "    \n",
        "    # Print epoch statistics\n",
        "    print('Epoch {}: Train Loss: {:.5f}, Valid Loss: {:.5f}, Exact Match: {:.2f}%, F1: {:.2f}%'.format(epoch + 1,\n",
        "                                                                                                     train_loss,\n",
        "                                                                                                     valid_loss,\n",
        "                                                                                                     exact_match_score,\n",
        "                                                                                                     f1_score))\n",
        "\n",
        "# # Plotting the results\n",
        "# epochs = range(1, n_epochs + 1)\n",
        "\n",
        "# # Plotting training and validation losses\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, train_losses, label='Train')\n",
        "# plt.plot(epochs, valid_losses, label='Validation')\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('Loss')\n",
        "# plt.title('Training and Validation Loss')\n",
        "# plt.legend()\n",
        "# plt.savefig('loss_plot.pdf')\n",
        "# plt.show()\n",
        "\n",
        "# # Plotting Exact Match scores\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, exact_match_scores)\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('Exact Match (%)')\n",
        "# plt.title('Exact Match Score')\n",
        "# plt.savefig('exact_match_plot.pdf')\n",
        "# plt.show()\n",
        "\n",
        "# # Plotting F1 scores\n",
        "# plt.figure()\n",
        "# plt.plot(epochs, f1_scores)\n",
        "# plt.xlabel('Epochs')\n",
        "# plt.ylabel('F1 Score (%)')\n",
        "# plt.title('F1 Score')\n",
        "# plt.savefig('f1_score_plot.pdf')\n",
        "# plt.show()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# 1.4. ALBERT (Evaluating)"
      ],
      "metadata": {
        "id": "nliKHfSAojOT"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Softmax function for probability calculation\n",
        "softmax = torch.nn.Softmax(dim=1)\n",
        "\n",
        "# Set the model in evaluation mode\n",
        "model.eval()\n",
        "\n",
        "# Initialize variables for loss and evaluation metrics\n",
        "n_test_batches = len(test_loader)\n",
        "loss = 0\n",
        "exact_match = []\n",
        "f1_scores = []\n",
        "\n",
        "# Iterate over the test data\n",
        "with torch.no_grad():\n",
        "    loop = tqdm(test_loader)\n",
        "    for content in loop:\n",
        "        # Move input tensors to the appropriate device\n",
        "        input_ids = content['input_ids'].to(device)\n",
        "        temp_ids = input_ids.cpu().data.numpy().tolist()\n",
        "        max_ind = [temp_ids[i].index(3) for i in range(0, len(temp_ids))]  # index of first [SEP]\n",
        "        attention_mask = content['attention_mask'].to(device)\n",
        "        token_type_ids = content['token_type_ids'].to(device)\n",
        "        y_start = content['targets_start'].to(device)\n",
        "        y_end = content['targets_end'].to(device)\n",
        "\n",
        "        # Forward pass\n",
        "        out_start, out_end = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)\n",
        "        start_pred = softmax(out_start)\n",
        "        end_pred = softmax(out_end)\n",
        "\n",
        "        # Calculate the loss\n",
        "        loss_batch = loss_fn(out_start, out_end, y_start, y_end)\n",
        "        loss += loss_batch.item()\n",
        "\n",
        "        # Convert labels and predictions to numpy arrays\n",
        "        start_label = np.nonzero(y_start).cpu().data.numpy()\n",
        "        end_label = np.nonzero(y_end).cpu().data.numpy()\n",
        "        tensor_label = torch.stack((torch.tensor(start_label[:, 1]), torch.tensor(end_label[:, 1])), -1)\n",
        "        start_log, end_log = generate_indexes(start_pred, end_pred, N, max_ind)\n",
        "        start_log = np.array(start_log)\n",
        "        end_log = np.array(end_log)\n",
        "        tensor_pred = torch.stack((torch.tensor(start_log), torch.tensor(end_log)), -1)\n",
        "\n",
        "        # Calculate exact match and F1 scores\n",
        "        ex_ma = sum([1 if (tensor_pred[i][0] == tensor_label[i][0] and tensor_pred[i][1] == tensor_label[i][1]) else 0\n",
        "                     for i in range(0, len(tensor_pred))]) / len(tensor_pred)\n",
        "        exact_match.append(ex_ma)\n",
        "\n",
        "        f1_score = evaluate_model(start_pred.cpu().data.numpy(), end_pred.cpu().data.numpy(),\n",
        "                                  N, max_ind, start_label[:, 1], end_label[:, 1])\n",
        "        f1_scores.append(f1_score)\n",
        "\n",
        "    # Calculate average loss and metrics\n",
        "    valid_loss = loss / n_test_batches\n",
        "    em_accuracy = 100 * np.mean(exact_match)\n",
        "    f1_mean = 100 * np.mean(f1_scores)\n",
        "    print('Test Loss: {:.4f}, Exact Match: {:.2f}%, F1 Score: {:.2f}%'.format(valid_loss, em_accuracy, f1_mean))"
      ],
      "metadata": {
        "id": "IKajRCWD5D3Q",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "73349bb6-0a3b-4f64-b451-764fb2035889"
      },
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 276/276 [12:17<00:00,  2.67s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Test Loss: 0.0416, Exact Match: 18.59%, F1 Score: 19.24%\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# ALBERT - Method 2"
      ],
      "metadata": {
        "id": "VbqfD24VIT_g"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "! nvidia-smi\n",
        "!pip install transformers[torch]  # Install the transformers package with PyTorch support\n",
        "!pip install accelerate -U  # Install the accelerate package\n",
        "! pip install transformers datasets\n",
        "! pip install huggingface_hub\n",
        "! pip install sentencepiece\n",
        "# ! gdown 1SLkoO3mD7bpV8B7UpcDeuPHIOF3bFLPp\n",
        "! unzip Dataset.zip\n",
        "from datasets import load_dataset, load_metric\n",
        "from transformers import AutoTokenizer, DefaultDataCollator, AutoModelForQuestionAnswering\n",
        "from transformers import TrainingArguments, Trainer, create_optimizer, pipeline\n",
        "import json\n",
        "import pandas as pd\n",
        "import datasets\n",
        "from tqdm.auto import tqdm\n",
        "from matplotlib import pyplot as plt\n",
        "from huggingface_hub import notebook_login\n",
        "notebook_login()"
      ],
      "metadata": {
        "id": "7DgqE92Jo6ur",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000,
          "referenced_widgets": [
            "89aa2fe8d7e5444bbf1ed02b4b52307d",
            "6ba684d6f6444062a97663e58a1f90c9",
            "9d7d6c0882e2411e994016ab626ef47c",
            "835354b9a1714113bb4316cdf06cc9cb",
            "e303e0323b06400f87ef04359b10ef1c",
            "6bf56af69b8b45229242cd7532531976",
            "7c4da9c2b4f9453e8f5b5d63b16f315a",
            "9df3669f33644a52b5c2ecf5f6b45b7d",
            "05c856c20da045ae8fdf79cba9da708a",
            "5c91635d7239403494ca820b7cb1db0b",
            "7caaa50cab264b429a2eee589eaac064",
            "2404a3ce16a54ce3a239f0f14db7d13e",
            "a167df040c9441b9ae26b7e37fa75bfa",
            "0c2a3eb042f44ca98ce160c184d61a81",
            "c61bd82dfa404a5f840ebb4e080f4bc8",
            "54d485902bb64812b297ee0a0746a4f1",
            "2494eb315625447287ba0913588b3dcf",
            "3eb42853e4854bfaabf4c3f0821f428b",
            "3baaecb5bb3c475dbda32e9a841de295",
            "8d55040a1b9440229940c4fc0272cd3b",
            "1634761578cf43e183a0d188f142be76",
            "368a1768378b4c678cfc92a1c9c55e8b",
            "bea474f4786e4595884f69a4eb7314fb",
            "22c2a688e5874b8ebe22574580dca5e6",
            "aa91bc2d637d4013ac971b51ba13485b",
            "86b1f7d3829d4fd7be62d3c4f7663c67",
            "1618d0c730554e80aa758d10e49596f4",
            "39e66bcd4d7d4b88a927f274a4e6b5be",
            "724b5e5cda2f4519abc2302854beb982",
            "fd15e87132f04ddc91e00ff38c892ef2",
            "dab2d9544f8941ccbf015327e7b89f53",
            "569526c5dcca48aeac72ffe6bf781340",
            "5b95de48161945a0ab3937c691faafc5",
            "eb68f70d02dc4c8eaffa02e8e74927da",
            "1842ec8566a74be19eb31355840e6faa",
            "83c39a519d27437aaddc7c00612e694f",
            "be76a584312a43d798000a425c026866",
            "2aa6e41f81744c9f9c021817b93789e8",
            "083bc732b50d466ea8b9c8174175daa4",
            "8a0923ac084249f995651b7cce797938",
            "63bbce8baa814624b143f4388f6c5a04",
            "ee854fd0f91145679962780e56a6dd56",
            "db1534149dd94b369828fe2c261c8250",
            "6062a44fb9a24e73b12b035433f3ffbc",
            "9076fdec70224c9d821095a4a097d679",
            "65b3277270cf46739de33d5fa244f032",
            "493bc9228e5d43a5829c95e6dc60719c"
          ]
        },
        "outputId": "9b1c72e1-9600-42e1-be12-0cc18ceabc0e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Thu Jun  8 22:10:18 2023       \n",
            "+-----------------------------------------------------------------------------+\n",
            "| NVIDIA-SMI 525.85.12    Driver Version: 525.85.12    CUDA Version: 12.0     |\n",
            "|-------------------------------+----------------------+----------------------+\n",
            "| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |\n",
            "| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |\n",
            "|                               |                      |               MIG M. |\n",
            "|===============================+======================+======================|\n",
            "|   0  Tesla T4            Off  | 00000000:00:04.0 Off |                    0 |\n",
            "| N/A   59C    P8    12W /  70W |      0MiB / 15360MiB |      0%      Default |\n",
            "|                               |                      |                  N/A |\n",
            "+-------------------------------+----------------------+----------------------+\n",
            "                                                                               \n",
            "+-----------------------------------------------------------------------------+\n",
            "| Processes:                                                                  |\n",
            "|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |\n",
            "|        ID   ID                                                   Usage      |\n",
            "|=============================================================================|\n",
            "|  No running processes found                                                 |\n",
            "+-----------------------------------------------------------------------------+\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting transformers[torch]\n",
            "  Using cached transformers-4.30.0-py3-none-any.whl (7.2 MB)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (3.12.0)\n",
            "Collecting huggingface-hub<1.0,>=0.14.1 (from transformers[torch])\n",
            "  Using cached huggingface_hub-0.15.1-py3-none-any.whl (236 kB)\n",
            "Requirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (1.22.4)\n",
            "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (23.1)\n",
            "Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (6.0)\n",
            "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (2022.10.31)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (2.27.1)\n",
            "Collecting tokenizers!=0.11.3,<0.14,>=0.11.1 (from transformers[torch])\n",
            "  Using cached tokenizers-0.13.3-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (7.8 MB)\n",
            "Collecting safetensors>=0.3.1 (from transformers[torch])\n",
            "  Using cached safetensors-0.3.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.3 MB)\n",
            "Requirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (4.65.0)\n",
            "Requirement already satisfied: torch!=1.12.0,>=1.9 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (2.0.1+cu118)\n",
            "Collecting accelerate>=0.20.2 (from transformers[torch])\n",
            "  Downloading accelerate-0.20.3-py3-none-any.whl (227 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m227.6/227.6 kB\u001b[0m \u001b[31m18.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: psutil in /usr/local/lib/python3.10/dist-packages (from accelerate>=0.20.2->transformers[torch]) (5.9.5)\n",
            "Requirement already satisfied: fsspec in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.14.1->transformers[torch]) (2023.4.0)\n",
            "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.14.1->transformers[torch]) (4.5.0)\n",
            "Requirement already satisfied: sympy in /usr/local/lib/python3.10/dist-packages (from torch!=1.12.0,>=1.9->transformers[torch]) (1.11.1)\n",
            "Requirement already satisfied: networkx in /usr/local/lib/python3.10/dist-packages (from torch!=1.12.0,>=1.9->transformers[torch]) (3.1)\n",
            "Requirement already satisfied: jinja2 in /usr/local/lib/python3.10/dist-packages (from torch!=1.12.0,>=1.9->transformers[torch]) (3.1.2)\n",
            "Requirement already satisfied: triton==2.0.0 in /usr/local/lib/python3.10/dist-packages (from torch!=1.12.0,>=1.9->transformers[torch]) (2.0.0)\n",
            "Requirement already satisfied: cmake in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch!=1.12.0,>=1.9->transformers[torch]) (3.25.2)\n",
            "Requirement already satisfied: lit in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch!=1.12.0,>=1.9->transformers[torch]) (16.0.5)\n",
            "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->transformers[torch]) (1.26.15)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->transformers[torch]) (2022.12.7)\n",
            "Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.10/dist-packages (from requests->transformers[torch]) (2.0.12)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->transformers[torch]) (3.4)\n",
            "Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.10/dist-packages (from jinja2->torch!=1.12.0,>=1.9->transformers[torch]) (2.1.2)\n",
            "Requirement already satisfied: mpmath>=0.19 in /usr/local/lib/python3.10/dist-packages (from sympy->torch!=1.12.0,>=1.9->transformers[torch]) (1.3.0)\n",
            "Installing collected packages: tokenizers, safetensors, huggingface-hub, transformers, accelerate\n",
            "Successfully installed accelerate-0.20.3 huggingface-hub-0.15.1 safetensors-0.3.1 tokenizers-0.13.3 transformers-4.30.0\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: accelerate in /usr/local/lib/python3.10/dist-packages (0.20.3)\n",
            "Requirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.10/dist-packages (from accelerate) (1.22.4)\n",
            "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.10/dist-packages (from accelerate) (23.1)\n",
            "Requirement already satisfied: psutil in /usr/local/lib/python3.10/dist-packages (from accelerate) (5.9.5)\n",
            "Requirement already satisfied: pyyaml in /usr/local/lib/python3.10/dist-packages (from accelerate) (6.0)\n",
            "Requirement already satisfied: torch>=1.6.0 in /usr/local/lib/python3.10/dist-packages (from accelerate) (2.0.1+cu118)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from torch>=1.6.0->accelerate) (3.12.0)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.10/dist-packages (from torch>=1.6.0->accelerate) (4.5.0)\n",
            "Requirement already satisfied: sympy in /usr/local/lib/python3.10/dist-packages (from torch>=1.6.0->accelerate) (1.11.1)\n",
            "Requirement already satisfied: networkx in /usr/local/lib/python3.10/dist-packages (from torch>=1.6.0->accelerate) (3.1)\n",
            "Requirement already satisfied: jinja2 in /usr/local/lib/python3.10/dist-packages (from torch>=1.6.0->accelerate) (3.1.2)\n",
            "Requirement already satisfied: triton==2.0.0 in /usr/local/lib/python3.10/dist-packages (from torch>=1.6.0->accelerate) (2.0.0)\n",
            "Requirement already satisfied: cmake in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch>=1.6.0->accelerate) (3.25.2)\n",
            "Requirement already satisfied: lit in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch>=1.6.0->accelerate) (16.0.5)\n",
            "Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.10/dist-packages (from jinja2->torch>=1.6.0->accelerate) (2.1.2)\n",
            "Requirement already satisfied: mpmath>=0.19 in /usr/local/lib/python3.10/dist-packages (from sympy->torch>=1.6.0->accelerate) (1.3.0)\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: transformers in /usr/local/lib/python3.10/dist-packages (4.30.0)\n",
            "Collecting datasets\n",
            "  Using cached datasets-2.12.0-py3-none-any.whl (474 kB)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from transformers) (3.12.0)\n",
            "Requirement already satisfied: huggingface-hub<1.0,>=0.14.1 in /usr/local/lib/python3.10/dist-packages (from transformers) (0.15.1)\n",
            "Requirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.10/dist-packages (from transformers) (1.22.4)\n",
            "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.10/dist-packages (from transformers) (23.1)\n",
            "Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.10/dist-packages (from transformers) (6.0)\n",
            "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.10/dist-packages (from transformers) (2022.10.31)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from transformers) (2.27.1)\n",
            "Requirement already satisfied: tokenizers!=0.11.3,<0.14,>=0.11.1 in /usr/local/lib/python3.10/dist-packages (from transformers) (0.13.3)\n",
            "Requirement already satisfied: safetensors>=0.3.1 in /usr/local/lib/python3.10/dist-packages (from transformers) (0.3.1)\n",
            "Requirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.10/dist-packages (from transformers) (4.65.0)\n",
            "Requirement already satisfied: pyarrow>=8.0.0 in /usr/local/lib/python3.10/dist-packages (from datasets) (9.0.0)\n",
            "Collecting dill<0.3.7,>=0.3.0 (from datasets)\n",
            "  Using cached dill-0.3.6-py3-none-any.whl (110 kB)\n",
            "Requirement already satisfied: pandas in /usr/local/lib/python3.10/dist-packages (from datasets) (1.5.3)\n",
            "Collecting xxhash (from datasets)\n",
            "  Using cached xxhash-3.2.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (212 kB)\n",
            "Collecting multiprocess (from datasets)\n",
            "  Using cached multiprocess-0.70.14-py310-none-any.whl (134 kB)\n",
            "Requirement already satisfied: fsspec[http]>=2021.11.1 in /usr/local/lib/python3.10/dist-packages (from datasets) (2023.4.0)\n",
            "Collecting aiohttp (from datasets)\n",
            "  Downloading aiohttp-3.8.4-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.0 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.0/1.0 MB\u001b[0m \u001b[31m54.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting responses<0.19 (from datasets)\n",
            "  Downloading responses-0.18.0-py3-none-any.whl (38 kB)\n",
            "Requirement already satisfied: attrs>=17.3.0 in /usr/local/lib/python3.10/dist-packages (from aiohttp->datasets) (23.1.0)\n",
            "Requirement already satisfied: charset-normalizer<4.0,>=2.0 in /usr/local/lib/python3.10/dist-packages (from aiohttp->datasets) (2.0.12)\n",
            "Collecting multidict<7.0,>=4.5 (from aiohttp->datasets)\n",
            "  Downloading multidict-6.0.4-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (114 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m114.5/114.5 kB\u001b[0m \u001b[31m17.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting async-timeout<5.0,>=4.0.0a3 (from aiohttp->datasets)\n",
            "  Downloading async_timeout-4.0.2-py3-none-any.whl (5.8 kB)\n",
            "Collecting yarl<2.0,>=1.0 (from aiohttp->datasets)\n",
            "  Downloading yarl-1.9.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (268 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m268.8/268.8 kB\u001b[0m \u001b[31m30.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting frozenlist>=1.1.1 (from aiohttp->datasets)\n",
            "  Downloading frozenlist-1.3.3-cp310-cp310-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl (149 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m149.6/149.6 kB\u001b[0m \u001b[31m21.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting aiosignal>=1.1.2 (from aiohttp->datasets)\n",
            "  Downloading aiosignal-1.3.1-py3-none-any.whl (7.6 kB)\n",
            "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.14.1->transformers) (4.5.0)\n",
            "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (1.26.15)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (2022.12.7)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (3.4)\n",
            "Requirement already satisfied: python-dateutil>=2.8.1 in /usr/local/lib/python3.10/dist-packages (from pandas->datasets) (2.8.2)\n",
            "Requirement already satisfied: pytz>=2020.1 in /usr/local/lib/python3.10/dist-packages (from pandas->datasets) (2022.7.1)\n",
            "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.10/dist-packages (from python-dateutil>=2.8.1->pandas->datasets) (1.16.0)\n",
            "Installing collected packages: xxhash, multidict, frozenlist, dill, async-timeout, yarl, responses, multiprocess, aiosignal, aiohttp, datasets\n",
            "Successfully installed aiohttp-3.8.4 aiosignal-1.3.1 async-timeout-4.0.2 datasets-2.12.0 dill-0.3.6 frozenlist-1.3.3 multidict-6.0.4 multiprocess-0.70.14 responses-0.18.0 xxhash-3.2.0 yarl-1.9.2\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: huggingface_hub in /usr/local/lib/python3.10/dist-packages (0.15.1)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from huggingface_hub) (3.12.0)\n",
            "Requirement already satisfied: fsspec in /usr/local/lib/python3.10/dist-packages (from huggingface_hub) (2023.4.0)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from huggingface_hub) (2.27.1)\n",
            "Requirement already satisfied: tqdm>=4.42.1 in /usr/local/lib/python3.10/dist-packages (from huggingface_hub) (4.65.0)\n",
            "Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.10/dist-packages (from huggingface_hub) (6.0)\n",
            "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.10/dist-packages (from huggingface_hub) (4.5.0)\n",
            "Requirement already satisfied: packaging>=20.9 in /usr/local/lib/python3.10/dist-packages (from huggingface_hub) (23.1)\n",
            "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->huggingface_hub) (1.26.15)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->huggingface_hub) (2022.12.7)\n",
            "Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.10/dist-packages (from requests->huggingface_hub) (2.0.12)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->huggingface_hub) (3.4)\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting sentencepiece\n",
            "  Using cached sentencepiece-0.1.99-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.3 MB)\n",
            "Installing collected packages: sentencepiece\n",
            "Successfully installed sentencepiece-0.1.99\n",
            "Downloading...\n",
            "From: https://drive.google.com/uc?id=1SLkoO3mD7bpV8B7UpcDeuPHIOF3bFLPp\n",
            "To: /content/Dataset.zip\n",
            "100% 13.0M/13.0M [00:00<00:00, 23.5MB/s]\n",
            "Archive:  Dataset.zip\n",
            "   creating: Dataset/\n",
            "   creating: Dataset/NLU dataset (Farsi MASSIVE)/\n",
            "  inflating: Dataset/NLU dataset (Farsi MASSIVE)/fa-IR.jsonl  \n",
            "   creating: Dataset/ParSQuAD/\n",
            "  inflating: Dataset/ParSQuAD/ParSQuAD-automatic-dev.json  \n",
            "  inflating: Dataset/ParSQuAD/ParSQuAD-automatic-train.json  \n",
            "  inflating: Dataset/ParSQuAD/ParSQuAD-manual-dev.json  \n",
            "  inflating: Dataset/ParSQuAD/ParSQuAD-manual-train.json  \n",
            "   creating: Dataset/PersianQA/\n",
            "  inflating: Dataset/PersianQA/pqa_test.json  \n",
            "  inflating: Dataset/PersianQA/pqa_train.json  \n",
            "   creating: Dataset/pquad_public/\n",
            "  inflating: Dataset/pquad_public/test_samples.json  \n",
            "  inflating: Dataset/pquad_public/train_samples.json  \n",
            "  inflating: Dataset/pquad_public/validation_samples.json  \n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "VBox(children=(HTML(value='<center> <img\\nsrc=https://huggingface.co/front/assets/huggingface_logo-noborder.sv…"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "89aa2fe8d7e5444bbf1ed02b4b52307d"
            }
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Clone the PQuAD dataset from GitHub\n",
        "!git clone https://github.com/AUT-NLP/PQuAD.git\n",
        "\n",
        "# Load the PQuad dataset\n",
        "train_file_path = '/content/PQuAD/Dataset/Train.json'\n",
        "test_file_path = '/content/PQuAD/Dataset/Test.json'\n",
        "val_file_path = '/content/PQuAD/Dataset/Validation.json'"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7mNYI4o_E70H",
        "outputId": "4034c9e4-1395-4734-e55d-9534fb573ecb"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Cloning into 'PQuAD'...\n",
            "remote: Enumerating objects: 27, done.\u001b[K\n",
            "remote: Counting objects: 100% (27/27), done.\u001b[K\n",
            "remote: Compressing objects: 100% (22/22), done.\u001b[K\n",
            "remote: Total 27 (delta 9), reused 15 (delta 3), pack-reused 0\u001b[K\n",
            "Unpacking objects: 100% (27/27), 5.71 MiB | 6.60 MiB/s, done.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "DATASETS = {\n",
        "    \"pquad\": 0,  # Dataset key for pquad\n",
        "    \"persian_qa\": 1,  # Dataset key for PersianQA\n",
        "    \"parsquad\": 2,  # Dataset key for ParSQuAD\n",
        "    \"pquad_and_persian_qa\": 3  # Dataset key for both pquad and PersianQA\n",
        "}\n",
        "\n",
        "class DatasetLoader:\n",
        "    def __init__(self, dataset, tokenizer):\n",
        "        self.tokenizer = tokenizer  # Tokenizer for preprocessing\n",
        "        self.dataset = datasets.DatasetDict()  # Dictionary to store the datasets\n",
        "        dataset_to_loader = {\n",
        "            DATASETS[\"pquad\"]: self.__load_pquad_public,  # Load the pquad_public dataset\n",
        "            DATASETS[\"persian_qa\"]: self.__load_persian_qa,  # Load the PersianQA dataset\n",
        "            DATASETS[\"parsquad\"]: self.__load_parsquad,  # Load the ParSQuAD dataset\n",
        "            DATASETS[\"pquad_and_persian_qa\"]: self.__load_pquad_and_persian_qa  # Load both pquad_public and PersianQA datasets\n",
        "        }\n",
        "        self.dataset = dataset_to_loader[dataset]()  # Call the corresponding loader based on the selected dataset\n",
        "        self.tokenized_dataset = self.dataset.map(self.preprocess_function,  # Apply the preprocess_function to tokenize and preprocess the dataset\n",
        "                                                  batched=True,\n",
        "                                                  remove_columns=self.dataset[\"train\"].column_names)\n",
        "        \n",
        "    def __extract_entries(self, data, limit=200000000):\n",
        "        df_list = []  # List to store the extracted entries\n",
        "        c = 0  # Counter to keep track of the number of entries processed\n",
        "        length_distribution = []  # List to store the length of contexts\n",
        "        for d in tqdm(data['data'], desc=\"Converting json to dataset\"):  # Iterate over the data entries\n",
        "            for p in d['paragraphs']:  # Iterate over the paragraphs in the data entry\n",
        "                length_distribution.append(len(p['context'].split()))  # Record the length of the context\n",
        "                for qas in p['qas']:  # Iterate over the questions and answers in the paragraph\n",
        "                    c += 1  # Increment the counter\n",
        "                    if c > limit:\n",
        "                        return df_list  # Return the extracted entries if the limit is reached\n",
        "                    if qas[\"is_impossible\"]:\n",
        "                        continue  # Skip the entry if it is impossible to answer\n",
        "                        df_list.append({\n",
        "                            \"id\": str(qas['id']),\n",
        "                            \"title\": d['title'],\n",
        "                            \"context\": p['context'],\n",
        "                            \"question\": qas['question'],\n",
        "                            \"answers\": {\"text\": \"\", \"answer_start\": 0}\n",
        "                        })\n",
        "                    else:\n",
        "                        for answer in qas['answers']:  # Iterate over the answers\n",
        "                            df_list.append({\n",
        "                                \"id\": str(qas['id']),\n",
        "                                \"title\": d['title'],\n",
        "                                \"context\": p['context'],\n",
        "                                \"question\": qas['question'],\n",
        "                                \"answers\": {\"text\": answer[\"text\"], \"answer_start\": answer[\"answer_start\"]}\n",
        "                            })\n",
        "        plt.hist(length_distribution)  # Plot the length distribution of contexts\n",
        "        plt.title(\"Length Distribution\")\n",
        "        plt.show()\n",
        "        return df_list\n",
        "        \n",
        "    def __load_pquad_public(self):\n",
        "        self.dataset = datasets.DatasetDict()  # Clear the dataset dictionary\n",
        "        for part in [\"train\", \"validation\", \"test\"]:\n",
        "            with open(f\"PQuAD/Dataset/{part}.json\", 'r', encoding='utf-8') as f:\n",
        "                data = json.load(f)  # Load the data from the JSON file\n",
        "            df_list = self.__extract_entries(data)  # Extract entries from the data\n",
        "            self.dataset[part] = datasets.Dataset.from_pandas(pd.DataFrame.from_dict(df_list))  # Convert the extracted entries to a dataset\n",
        "        return self.dataset\n",
        "    \n",
        "    def __load_persian_qa(self):\n",
        "        self.dataset = datasets.DatasetDict()  # Clear the dataset dictionary\n",
        "        part_to_path = {\n",
        "            \"train\": f\"Dataset/PersianQA/pqa_train.json\",\n",
        "            \"validation\": f\"Dataset/PersianQA/pqa_test.json\",\n",
        "            \"test\": f\"Dataset/pquad_public/test_samples.json\"\n",
        "        }\n",
        "        for part in [\"train\", \"validation\", \"test\"]:\n",
        "            with open(part_to_path[part], 'r', encoding='utf-8') as f:\n",
        "                data = json.load(f)  # Load the data from the JSON file\n",
        "            df_list = self.__extract_entries(data)  # Extract entries from the data\n",
        "            self.dataset[part] = datasets.Dataset.from_pandas(pd.DataFrame.from_dict(df_list))  # Convert the extracted entries to a dataset\n",
        "        return self.dataset\n",
        "    \n",
        "    def __load_parsquad(self):\n",
        "        self.dataset = datasets.DatasetDict()  # Clear the dataset dictionary\n",
        "        part_to_path = {\n",
        "            \"train\": f\"Dataset/ParSQuAD/ParSQuAD-manual-train.json\",\n",
        "            \"validation\": f\"Dataset/ParSQuAD/ParSQuAD-manual-dev.json\",\n",
        "            \"test\": f\"Dataset/pquad_public/test_samples.json\"\n",
        "        }\n",
        "        for part in [\"train\", \"validation\", \"test\"]:\n",
        "            with open(part_to_path[part], 'r', encoding='utf-8') as f:\n",
        "                data = json.load(f)  # Load the data from the JSON file\n",
        "            df_list = self.__extract_entries(data)  # Extract entries from the data\n",
        "            self.dataset[part] = datasets.Dataset.from_pandas(pd.DataFrame.from_dict(df_list))  # Convert the extracted entries to a dataset\n",
        "        return self.dataset\n",
        "    \n",
        "    def __load_pquad_and_persian_qa(self):\n",
        "        self.dataset = datasets.DatasetDict()  # Clear the dataset dictionary\n",
        "        part_to_path = {\n",
        "            \"train\": f\"Dataset/PersianQA/pqa_train.json\",\n",
        "            \"validation\": f\"Dataset/PersianQA/pqa_test.json\",\n",
        "            \"test\": f\"Dataset/pquad_public/test_samples.json\"\n",
        "        }\n",
        "        for part in [\"train\", \"validation\", \"test\"]:\n",
        "            with open(f\"PQuAD/Dataset/{part}.json\", 'r', encoding='utf-8') as f:\n",
        "                data = json.load(f)  # Load the data from the JSON file\n",
        "            df_list = self.__extract_entries(data)  # Extract entries from the data\n",
        "            if part != \"test\":\n",
        "                with open(part_to_path[part], 'r', encoding='utf-8') as f:\n",
        "                    data = json.load(f)  # Load the additional data from the JSON file\n",
        "                df_list.extend(self.__extract_entries(data))  # Extend the extracted entries with additional entries\n",
        "            self.dataset[part] = datasets.Dataset.from_pandas(pd.DataFrame.from_dict(df_list))  # Convert the extracted entries to a dataset\n",
        "        return self.dataset\n",
        "    \n",
        "    def preprocess_function(self, examples):\n",
        "        questions = [q.strip() for q in examples[\"question\"]]  # Extract the questions\n",
        "        inputs = self.tokenizer(\n",
        "            questions,\n",
        "            examples[\"context\"],\n",
        "            max_length=400,  # Maximum length of the tokenized inputs\n",
        "#             max_length=250,\n",
        "            truncation=\"only_second\",  # Truncate the second sequence (context)\n",
        "            # return_overflowing_tokens=True,\n",
        "            return_offsets_mapping=True,  # Return the character offsets of the tokens\n",
        "#             padding=\"max_length\",\n",
        "        )\n",
        "\n",
        "        offset_mapping = inputs.pop(\"offset_mapping\")  # Extract the offset mappings\n",
        "        answers = examples[\"answers\"]  # Extract the answers\n",
        "        start_positions = []  # List to store the start positions of the answers\n",
        "        end_positions = []  # List to store the end positions of the answers\n",
        "\n",
        "        for i, offset in enumerate(offset_mapping):  # Iterate over the offset mappings\n",
        "            answer = answers[i]  # Get the answer\n",
        "            start_char = answer[\"answer_start\"]  # Get the start character position of the answer\n",
        "            end_char = answer[\"answer_start\"] + len(answer[\"text\"])  # Get the end character position of the answer\n",
        "            sequence_ids = inputs.sequence_ids(i)  # Get the sequence IDs of the input tokens\n",
        "\n",
        "            # Find the start and end of the context\n",
        "            idx = 0\n",
        "            while sequence_ids[idx] != 1:  # Find the start of the context\n",
        "                idx += 1\n",
        "            context_start = idx\n",
        "            while sequence_ids[idx] == 1:  # Find the end of the context\n",
        "                idx += 1\n",
        "            context_end = idx - 1\n",
        "\n",
        "            # If the answer is not fully inside the context, label it (0, 0)\n",
        "            if offset[context_start][0] > end_char or offset[context_end][1] < start_char:\n",
        "                start_positions.append(0)\n",
        "                end_positions.append(0)\n",
        "            else:\n",
        "                # Otherwise it's the start and end token positions\n",
        "                idx = context_start\n",
        "                while idx <= context_end and offset[idx][0] <= start_char:\n",
        "                    idx += 1\n",
        "                start_positions.append(idx - 1)\n",
        "\n",
        "                idx = context_end\n",
        "                while idx >= context_start and offset[idx][1] >= end_char:\n",
        "                    idx -= 1\n",
        "                end_positions.append(idx + 1)\n",
        "            \n",
        "        inputs[\"start_positions\"] = start_positions  # Add the start positions to the inputs\n",
        "        inputs[\"end_positions\"] = end_positions  # Add the end positions to the inputs\n",
        "        return inputs\n"
      ],
      "metadata": {
        "id": "7LaBXJZS_mSC"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "class TrainerQA:\n",
        "    def __init__(self, model_checkpoint, dataset):\n",
        "        print(\"### Loading Tokenizer ###\")\n",
        "        self.tokenizer = AutoTokenizer.from_pretrained(model_checkpoint)  # Load the tokenizer\n",
        "        print(\"### Loading Dataset ###\")\n",
        "        self.dataset_loader = DatasetLoader(dataset, self.tokenizer)  # Load the dataset\n",
        "        print(self.dataset_loader.dataset)  # Print the loaded dataset\n",
        "        print(self.dataset_loader.tokenized_dataset)  # Print the tokenized dataset\n",
        "        print(\"### Loading Model ###\")\n",
        "        self.model = AutoModelForQuestionAnswering.from_pretrained(model_checkpoint)  # Load the model\n",
        "        \n",
        "    def train(self, num_train_epochs=3, learning_rate=2e-5):\n",
        "        print(\"### Training Model ###\")\n",
        "        data_collator = DefaultDataCollator()  # Data collator for handling the training data\n",
        "        training_args = TrainingArguments(\n",
        "            output_dir=\"./results\",\n",
        "            evaluation_strategy=\"epoch\",\n",
        "            learning_rate=learning_rate,\n",
        "            per_device_train_batch_size=16,\n",
        "            per_device_eval_batch_size=16,\n",
        "            num_train_epochs=num_train_epochs,\n",
        "            weight_decay=0.01,\n",
        "            group_by_length=True,\n",
        "            logging_steps=20\n",
        "        )\n",
        "        self.trainer = Trainer(\n",
        "            model=self.model,\n",
        "            args=training_args,\n",
        "            train_dataset=self.dataset_loader.tokenized_dataset[\"train\"],\n",
        "            eval_dataset=self.dataset_loader.tokenized_dataset[\"validation\"],\n",
        "            tokenizer=self.tokenizer,\n",
        "#             data_collator=data_collator,\n",
        "        )  # Initialize the Trainer with the model, training arguments, datasets, and tokenizer\n",
        "        self.trainer.train()  # Train the model\n",
        "        \n",
        "    def evaluate(self):\n",
        "        print(\"### Evaluating Model ###\")\n",
        "        qa_model = pipeline(\"question-answering\", model=self.model, tokenizer=self.tokenizer, device=0)  # Create a question-answering pipeline with the model and tokenizer\n",
        "        questions = self.dataset_loader.dataset[\"test\"][\"question\"]  # Get the questions from the test dataset\n",
        "        contexts = self.dataset_loader.dataset[\"test\"][\"context\"]  # Get the contexts from the test dataset\n",
        "        preds = qa_model(question=questions, context=contexts, device=\"cuda\")  # Generate predictions using the question-answering pipeline\n",
        "        print(questions[0])  # Print the first question\n",
        "        print(contexts[0])  # Print the first context\n",
        "        print(preds[0])  # Print the prediction for the first question-context pair\n",
        "        \n",
        "        metric = load_metric(\"squad\")  # Load the SQuAD metric\n",
        "        last_id = -1\n",
        "        predictions, references = [], []\n",
        "        for i, answers in tqdm(enumerate(self.dataset_loader.dataset[\"test\"][\"answers\"])):\n",
        "            if len(answers[\"text\"]) < 1:\n",
        "                continue\n",
        "            id = self.dataset_loader.dataset[\"test\"][i][\"id\"]\n",
        "            if id != last_id:\n",
        "                predictions.append({\n",
        "                    \"id\": id,\n",
        "                    \"prediction_text\": preds[i][\"answer\"].strip()\n",
        "                })\n",
        "                references.append({\n",
        "                    \"id\": id,\n",
        "                    \"answers\": []\n",
        "                })\n",
        "                last_id = id\n",
        "            references[-1][\"answers\"].append(answers)\n",
        "#         predictions = [{\"id\": i, \"prediction_text\": pred[\"answer\"].strip()} for i, pred in enumerate(preds) if len(trainer.dataset_loader.dataset[\"test\"][i][\"answers\"][\"text\"]) > 0]\n",
        "#         references=[{\"id\": i, \"answers\": [answers]} for i, answers in enumerate(trainer.dataset_loader.dataset[\"test\"][\"answers\"]) if len(answers[\"text\"]) > 0]\n",
        "        results = metric.compute(predictions=predictions, references=references)  # Compute the SQuAD metric using the predictions and references\n",
        "        print(results)  # Print the evaluation results\n",
        "        return results\n",
        "\n",
        "    def push_to_hub(self, name):\n",
        "        self.model.push_to_hub(name)  # Push the model to the Hugging Face Model Hub\n",
        "        self.tokenizer.push_to_hub(name)  # Push the tokenizer to the Hugging Face Model Hub\n"
      ],
      "metadata": {
        "id": "GFRBV5CD_4fJ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000,
          "referenced_widgets": [
            "6142837de5d34b31bc7bc79a797bd3bd",
            "72190fc63eb5442e9375dc21353516e3",
            "f7b0066b000e427c9111a3663770dd5e",
            "33ed4573b8af43efb3d77ce3485e0780",
            "b331299eb6814dcead2046c51cb9e3fd",
            "389f4fbb92de41dda8679c4ea87731f4",
            "5c6f16e7b29a443fb86c41d127285083",
            "13a190c32117478ca3be5178cb81a9f9",
            "3e3fc2f61faf4222b88f0e666b01ab3f",
            "c020c2c7834d46a6a45607c1db3ac6d1",
            "5b4e0433f33b4f119141c710eeeae90d",
            "2778c8c1050c426bbe1e12198f61fa55",
            "dc97625cc9254c2eb6bbfde09bae966a",
            "c4f0fe5acf9c4035a96fec58596165ba",
            "2a25d8ab8f524c10ac612637c947449b",
            "37435be021dc437c9fd7bf8205eb3935",
            "30070a4415ce40d9a4ce5e1be077c72a",
            "1b22f4f30489410b9a8c964fdb4e1e44",
            "f58d5e71fcc848d9a1ac6448d8d633bf",
            "84b6451f72e943ebb699d5b2bd8a3dab",
            "b0fcab47012d46e6a0b8d2989152b2a8",
            "c8c8df95a47d42cbbb5c3cd33d20641d",
            "79e5789a55fe4913a16370b6ac921134",
            "20041fef51c94527acf3b85740739fda",
            "ca73192fb9284bb9bf8a5c1cffb821b3",
            "abebbe62d3e248168e02c7ccc43f307d",
            "26435d0766dd47eabf69112acbe66b78",
            "531e737a241948a6b42d4bafe6ed0852",
            "e440147ead2c442b9d842f0dce6454a5",
            "b37a3faf5baf4345b365699562dbf836",
            "4ab340b4dc284f9f899ce9d1ea10e290",
            "961e833dcf9146fc936fc5d653788999",
            "d3dde5651ae542cdac66666e65f3577c",
            "b0070c039c684f0b874004b703eeff8e",
            "f766568c974e4ff3a7f9eb9b30dec544",
            "802cb023921a4d6eb03dfb8ce73145f7",
            "d98bb3ba7b5646f9a43c747fa862158f",
            "f2b8e7b128ef4d5e9eee10ebf5927da4",
            "041f510d716643019a7486e22058cff1",
            "e6586509431c4f1d845047a4cf3539ea",
            "0f1b6f38d4f2407d895affbb24f159b4",
            "7e17ff1095344bf38bf357fa609d36d0",
            "b5b1e250111f457fbeca0c1cbe4e55d3",
            "71ca20214b464c4182e7e3888fe931d9",
            "81f94fe0b58b455c8d348c4375b6651f",
            "60929a49ebbd4f7e88cce53824737ee3",
            "a5f6f7945ca64d12a3c90910a11a381f",
            "d8ce0db4bd5248778d442778e7d8d28b",
            "bd703c11bdfe4e21b0d4bc46d69e1e13",
            "c88e4d2fcf4f465ca93efe28ec7f6375",
            "5571de0414d4441d979191fda59719ed",
            "2201953e1303486c8b8e4623dbebeefc",
            "6aef847a630945cb93ed699231fd7152",
            "0c8c8251bf0b467786fe4bcb77299af1",
            "61f1bc7ceb2f4dfda846445be8f2480a",
            "2026586018cf48a5ace467024242208b",
            "808b666544404696acd2e4d851c95162",
            "325df36e1a814a2d90e1b0a0768414eb",
            "0caf8fc64dea474a830aa33d6019a78f",
            "83c30579722645488aa4e61193831527",
            "ac4c898badcd4a3785cab6ebfcada393",
            "3c632a086cc944c297e992f4a5347c93",
            "811540681bfc4ecd88dc780371a62ff3",
            "a58767a4a79f420ab6e8b3dab16106d6",
            "d685e0fcb84645f5b67709ea89b32144",
            "9f288ea127934f79a745eab3f335f30f",
            "0426bb72d2f046f3ae889b5a759ed972",
            "126f39dbe0cf493298f084dce23c9b10",
            "e0cd8a8e8a704f9691ef09c38825b4fe",
            "54dc9c07de004d0fa311ff17ab929f2c",
            "50085e8a5e7e48db9c11ecc65f586193",
            "021805464e204c0bbc0e15261500d7a4",
            "6ddfdc4761d44d488fdbf2f6e8c98fe2",
            "ae91094a720243a9a31081afcfb7b569",
            "d2bf3396763044ae8e8dc211d67f5cb7",
            "3a9977fa803e45b19c70fd0723702032",
            "548d8f95bbe244c697eb7cc866b252b4",
            "37e142600bcf461a966a842cbe17957e",
            "5a7e6c16cd4445218c71e69d9727d910",
            "dbbfe021b8e9433f8eb938ae408dd398",
            "94928c1407da4ab5a07f27c0fa043371",
            "305128e8a87a485c824849f799dc1026",
            "619b0b6bd5274d06bb96f39094462d04",
            "64f31a26406843bc884fe696c997498b",
            "dc6318894da6497c81840dacbaf47f93",
            "66f673051d13417aac8102f2020f863a",
            "d088e5500f3849efad31e439989e7a4a",
            "76b641126e5b420fa10099da0c2a5872",
            "e6cca63711144a089091313ca78f6c4e",
            "386cf52f72f141c7bdf0b64b96bf5db3",
            "d204b94c93804ccbbdef8ed362affe5f",
            "22cef5c2a9e44ef8ba0c99883b9fa706",
            "2b1ecfa5b359446cb0ba1779156a45ba",
            "b205769fe0ce41b7be50de7d5f33a440",
            "5d62a05a52f5438585836f7a968c3763",
            "6e29b2e715e34f8fb75e433c37776b71",
            "4c714cc6a9194130ad9d91ce0b3b8676",
            "56e58477e3ee44a8bd91476bd225b806",
            "d30c19f8b8de4d76bd59ab134a765f30"
          ]
        },
        "outputId": "b37c3de6-5b8e-41a9-e0c7-5b35341d209f",
        "id": "602vAnn8CYj3"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "### Loading Tokenizer ###\n",
            "### Loading Dataset ###\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Converting json to dataset:   0%|          | 0/891 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "6142837de5d34b31bc7bc79a797bd3bd"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Converting json to dataset:   0%|          | 0/120 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "2778c8c1050c426bbe1e12198f61fa55"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Converting json to dataset:   0%|          | 0/114 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "79e5789a55fe4913a16370b6ac921134"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Map:   0%|          | 0/48273 [00:00<?, ? examples/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "b0070c039c684f0b874004b703eeff8e"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Map:   0%|          | 0/8402 [00:00<?, ? examples/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "81f94fe0b58b455c8d348c4375b6651f"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Map:   0%|          | 0/8540 [00:00<?, ? examples/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "2026586018cf48a5ace467024242208b"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "DatasetDict({\n",
            "    train: Dataset({\n",
            "        features: ['id', 'title', 'context', 'question', 'answers'],\n",
            "        num_rows: 48273\n",
            "    })\n",
            "    validation: Dataset({\n",
            "        features: ['id', 'title', 'context', 'question', 'answers'],\n",
            "        num_rows: 8402\n",
            "    })\n",
            "    test: Dataset({\n",
            "        features: ['id', 'title', 'context', 'question', 'answers'],\n",
            "        num_rows: 8540\n",
            "    })\n",
            "})\n",
            "DatasetDict({\n",
            "    train: Dataset({\n",
            "        features: ['input_ids', 'token_type_ids', 'attention_mask', 'start_positions', 'end_positions'],\n",
            "        num_rows: 48273\n",
            "    })\n",
            "    validation: Dataset({\n",
            "        features: ['input_ids', 'token_type_ids', 'attention_mask', 'start_positions', 'end_positions'],\n",
            "        num_rows: 8402\n",
            "    })\n",
            "    test: Dataset({\n",
            "        features: ['input_ids', 'token_type_ids', 'attention_mask', 'start_positions', 'end_positions'],\n",
            "        num_rows: 8540\n",
            "    })\n",
            "})\n",
            "### Loading Model ###\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Some weights of the model checkpoint at m3hrdadfi/albert-fa-base-v2 were not used when initializing AlbertForQuestionAnswering: ['predictions.LayerNorm.bias', 'predictions.bias', 'sop_classifier.classifier.bias', 'sop_classifier.classifier.weight', 'predictions.decoder.weight', 'predictions.decoder.bias', 'predictions.LayerNorm.weight', 'predictions.dense.weight', 'predictions.dense.bias']\n",
            "- This IS expected if you are initializing AlbertForQuestionAnswering from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
            "- This IS NOT expected if you are initializing AlbertForQuestionAnswering from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
            "Some weights of AlbertForQuestionAnswering were not initialized from the model checkpoint at m3hrdadfi/albert-fa-base-v2 and are newly initialized: ['qa_outputs.bias', 'qa_outputs.weight']\n",
            "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "### Training Model ###\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.10/dist-packages/transformers/optimization.py:411: FutureWarning: This implementation of AdamW is deprecated and will be removed in a future version. Use the PyTorch implementation torch.optim.AdamW instead, or set `no_deprecation_warning=True` to disable this warning\n",
            "  warnings.warn(\n",
            "You're using a AlbertTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ],
            "text/html": [
              "\n",
              "    <div>\n",
              "      \n",
              "      <progress value='9054' max='9054' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      [9054/9054 1:36:59, Epoch 3/3]\n",
              "    </div>\n",
              "    <table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              " <tr style=\"text-align: left;\">\n",
              "      <th>Epoch</th>\n",
              "      <th>Training Loss</th>\n",
              "      <th>Validation Loss</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>0.852800</td>\n",
              "      <td>1.075374</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>0.686400</td>\n",
              "      <td>1.111044</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>0.455700</td>\n",
              "      <td>1.233197</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table><p>"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "### Evaluating Model ###\n",
            "کتاب مقدس دین اسلام چیست؟\n",
            "قرآن کتاب مقدس دین اسلام است و در باور مسلمانان سخنان خداست که به صورت وحی از سوی او توسط جبرئیل بر پیامبر اسلام، محمد بن عبدالله، نازل شده‌ است. مسلمانان، قرآن را بزرگترین معجزهٔ محمد و روشن‌ترین دلیل بر پیامبری او می‌دانند. قرآن اصلی‌ترین منبع وحی در اسلام به‌شمار می‌آید و به زبان عربی است. کلمهٔ قرآن در لغت به معنی «قرائت کردن» و «خواندن» است و مسلمانان معمولاً به آن با عناوینی مانند «قرآن کریم» و «قرآن مجید» اشاره می‌کنند. قرآن به ۳۰ جزء تقسیم شده و ۱۱۴ سوره دارد.\n",
            "{'score': 0.9982663989067078, 'start': 0, 'end': 5, 'answer': 'قرآن'}\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "<ipython-input-20-dab71847fa9c>:46: FutureWarning: load_metric is deprecated and will be removed in the next major version of datasets. Use 'evaluate.load' instead, from the new library 🤗 Evaluate: https://huggingface.co/docs/evaluate\n",
            "  metric = load_metric(\"squad\")  # Load the SQuAD metric\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading builder script:   0%|          | 0.00/1.72k [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "0426bb72d2f046f3ae889b5a759ed972"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading extra modules:   0%|          | 0.00/1.11k [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "37e142600bcf461a966a842cbe17957e"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "0it [00:00, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "e6cca63711144a089091313ca78f6c4e"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "{'exact_match': 56.48817345597897, 'f1': 77.83653745721735}\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[31m╭─\u001b[0m\u001b[31m──────────────────────────────\u001b[0m\u001b[31m \u001b[0m\u001b[1;31mTraceback \u001b[0m\u001b[1;2;31m(most recent call last)\u001b[0m\u001b[31m \u001b[0m\u001b[31m───────────────────────────────\u001b[0m\u001b[31m─╮\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[2;33m/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/\u001b[0m\u001b[1;33m_errors.py\u001b[0m:\u001b[94m259\u001b[0m in                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[92mhf_raise_for_status\u001b[0m                                                                              \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m256 \u001b[0m\u001b[2;33m│   \u001b[0m\u001b[33m</Tip>\u001b[0m                                                                                 \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m257 \u001b[0m\u001b[2;33m│   \u001b[0m\u001b[33m\"\"\"\u001b[0m                                                                                    \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m258 \u001b[0m\u001b[2m│   \u001b[0m\u001b[94mtry\u001b[0m:                                                                                   \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m259 \u001b[2m│   │   \u001b[0mresponse.raise_for_status()                                                        \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m260 \u001b[0m\u001b[2m│   \u001b[0m\u001b[94mexcept\u001b[0m HTTPError \u001b[94mas\u001b[0m e:                                                                 \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m261 \u001b[0m\u001b[2m│   │   \u001b[0merror_code = response.headers.get(\u001b[33m\"\u001b[0m\u001b[33mX-Error-Code\u001b[0m\u001b[33m\"\u001b[0m)                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m262 \u001b[0m                                                                                           \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[2;33m/usr/local/lib/python3.10/dist-packages/requests/\u001b[0m\u001b[1;33mmodels.py\u001b[0m:\u001b[94m960\u001b[0m in \u001b[92mraise_for_status\u001b[0m               \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m957 \u001b[0m\u001b[2m│   │   │   \u001b[0mhttp_error_msg = \u001b[33mu\u001b[0m\u001b[33m'\u001b[0m\u001b[33m%s\u001b[0m\u001b[33m Server Error: \u001b[0m\u001b[33m%s\u001b[0m\u001b[33m for url: \u001b[0m\u001b[33m%s\u001b[0m\u001b[33m'\u001b[0m % (\u001b[96mself\u001b[0m.status_code, rea   \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m958 \u001b[0m\u001b[2m│   │   \u001b[0m                                                                                   \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m959 \u001b[0m\u001b[2m│   │   \u001b[0m\u001b[94mif\u001b[0m http_error_msg:                                                                 \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m960 \u001b[2m│   │   │   \u001b[0m\u001b[94mraise\u001b[0m HTTPError(http_error_msg, response=\u001b[96mself\u001b[0m)                                 \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m961 \u001b[0m\u001b[2m│   \u001b[0m                                                                                       \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m962 \u001b[0m\u001b[2m│   \u001b[0m\u001b[94mdef\u001b[0m \u001b[92mclose\u001b[0m(\u001b[96mself\u001b[0m):                                                                       \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m963 \u001b[0m\u001b[2;90m│   │   \u001b[0m\u001b[33m\"\"\"Releases the connection back to the pool. Once this method has been\u001b[0m             \u001b[31m│\u001b[0m\n",
              "\u001b[31m╰──────────────────────────────────────────────────────────────────────────────────────────────────╯\u001b[0m\n",
              "\u001b[1;91mHTTPError: \u001b[0m\u001b[1;36m401\u001b[0m Client Error: Unauthorized for url: \u001b[4;94mhttps://huggingface.co/api/repos/create\u001b[0m\n",
              "\n",
              "\u001b[3mThe above exception was the direct cause of the following exception:\u001b[0m\n",
              "\n",
              "\u001b[31m╭─\u001b[0m\u001b[31m──────────────────────────────\u001b[0m\u001b[31m \u001b[0m\u001b[1;31mTraceback \u001b[0m\u001b[1;2;31m(most recent call last)\u001b[0m\u001b[31m \u001b[0m\u001b[31m───────────────────────────────\u001b[0m\u001b[31m─╮\u001b[0m\n",
              "\u001b[31m│\u001b[0m in \u001b[92m<cell line: 8>\u001b[0m:\u001b[94m8\u001b[0m                                                                              \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m in \u001b[92mpush_to_hub\u001b[0m:\u001b[94m71\u001b[0m                                                                                \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[2;33m/usr/local/lib/python3.10/dist-packages/transformers/utils/\u001b[0m\u001b[1;33mhub.py\u001b[0m:\u001b[94m804\u001b[0m in \u001b[92mpush_to_hub\u001b[0m             \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 801 \u001b[0m\u001b[2m│   │   \u001b[0m\u001b[94melse\u001b[0m:                                                                             \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 802 \u001b[0m\u001b[2m│   │   │   \u001b[0mworking_dir = repo_id.split(\u001b[33m\"\u001b[0m\u001b[33m/\u001b[0m\u001b[33m\"\u001b[0m)[-\u001b[94m1\u001b[0m]                                          \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 803 \u001b[0m\u001b[2m│   │   \u001b[0m                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m 804 \u001b[2m│   │   \u001b[0mrepo_id = \u001b[96mself\u001b[0m._create_repo(                                                      \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 805 \u001b[0m\u001b[2m│   │   │   \u001b[0mrepo_id, private=private, use_auth_token=use_auth_token, repo_url=repo_url,   \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 806 \u001b[0m\u001b[2m│   │   \u001b[0m)                                                                                 \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 807 \u001b[0m                                                                                          \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[2;33m/usr/local/lib/python3.10/dist-packages/transformers/utils/\u001b[0m\u001b[1;33mhub.py\u001b[0m:\u001b[94m661\u001b[0m in \u001b[92m_create_repo\u001b[0m            \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 658 \u001b[0m\u001b[2m│   │   │   │   │   \u001b[0mrepo_id = repo_id.split(\u001b[33m\"\u001b[0m\u001b[33m/\u001b[0m\u001b[33m\"\u001b[0m)[-\u001b[94m1\u001b[0m]                                      \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 659 \u001b[0m\u001b[2m│   │   │   │   \u001b[0mrepo_id = \u001b[33mf\u001b[0m\u001b[33m\"\u001b[0m\u001b[33m{\u001b[0morganization\u001b[33m}\u001b[0m\u001b[33m/\u001b[0m\u001b[33m{\u001b[0mrepo_id\u001b[33m}\u001b[0m\u001b[33m\"\u001b[0m                                     \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 660 \u001b[0m\u001b[2m│   │   \u001b[0m                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m 661 \u001b[2m│   │   \u001b[0murl = create_repo(repo_id=repo_id, token=use_auth_token, private=private, exist_  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 662 \u001b[0m\u001b[2m│   │   \u001b[0m                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 663 \u001b[0m\u001b[2m│   │   \u001b[0m\u001b[2m# If the namespace is not there, add it or `upload_file` will complain\u001b[0m            \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m 664 \u001b[0m\u001b[2m│   │   \u001b[0m\u001b[94mif\u001b[0m \u001b[33m\"\u001b[0m\u001b[33m/\u001b[0m\u001b[33m\"\u001b[0m \u001b[95mnot\u001b[0m \u001b[95min\u001b[0m repo_id \u001b[95mand\u001b[0m url != \u001b[33mf\u001b[0m\u001b[33m\"\u001b[0m\u001b[33m{\u001b[0mHUGGINGFACE_CO_RESOLVE_ENDPOINT\u001b[33m}\u001b[0m\u001b[33m/\u001b[0m\u001b[33m{\u001b[0mrepo_id\u001b[33m}\u001b[0m\u001b[33m\"\u001b[0m:  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[2;33m/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/\u001b[0m\u001b[1;33m_validators.py\u001b[0m:\u001b[94m118\u001b[0m in \u001b[92m_inner_fn\u001b[0m    \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m115 \u001b[0m\u001b[2m│   │   \u001b[0m\u001b[94mif\u001b[0m check_use_auth_token:                                                           \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m116 \u001b[0m\u001b[2m│   │   │   \u001b[0mkwargs = smoothly_deprecate_use_auth_token(fn_name=fn.\u001b[91m__name__\u001b[0m, has_token=ha   \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m117 \u001b[0m\u001b[2m│   │   \u001b[0m                                                                                   \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m118 \u001b[2m│   │   \u001b[0m\u001b[94mreturn\u001b[0m fn(*args, **kwargs)                                                         \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m119 \u001b[0m\u001b[2m│   \u001b[0m                                                                                       \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m120 \u001b[0m\u001b[2m│   \u001b[0m\u001b[94mreturn\u001b[0m _inner_fn  \u001b[2m# type: ignore\u001b[0m                                                       \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m121 \u001b[0m                                                                                           \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[2;33m/usr/local/lib/python3.10/dist-packages/huggingface_hub/\u001b[0m\u001b[1;33mhf_api.py\u001b[0m:\u001b[94m2306\u001b[0m in \u001b[92mcreate_repo\u001b[0m            \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m2303 \u001b[0m\u001b[2m│   │   \u001b[0mr = get_session().post(path, headers=headers, json=json)                          \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m2304 \u001b[0m\u001b[2m│   │   \u001b[0m                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m2305 \u001b[0m\u001b[2m│   │   \u001b[0m\u001b[94mtry\u001b[0m:                                                                              \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m2306 \u001b[2m│   │   │   \u001b[0mhf_raise_for_status(r)                                                        \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m2307 \u001b[0m\u001b[2m│   │   \u001b[0m\u001b[94mexcept\u001b[0m HTTPError \u001b[94mas\u001b[0m err:                                                          \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m2308 \u001b[0m\u001b[2m│   │   │   \u001b[0m\u001b[94mif\u001b[0m exist_ok \u001b[95mand\u001b[0m err.response.status_code == \u001b[94m409\u001b[0m:                              \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m2309 \u001b[0m\u001b[2m│   │   │   │   \u001b[0m\u001b[2m# Repo already exists and `exist_ok=True`\u001b[0m                                 \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[2;33m/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/\u001b[0m\u001b[1;33m_errors.py\u001b[0m:\u001b[94m291\u001b[0m in                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[92mhf_raise_for_status\u001b[0m                                                                              \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m                                                                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m288 \u001b[0m\u001b[2m│   │   │   │   \u001b[0m\u001b[33m\"\u001b[0m\u001b[33m `repo_type`.\u001b[0m\u001b[33m\\n\u001b[0m\u001b[33mIf you are trying to access a private or gated repo,\u001b[0m\u001b[33m\"\u001b[0m      \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m289 \u001b[0m\u001b[2m│   │   │   │   \u001b[0m\u001b[33m\"\u001b[0m\u001b[33m make sure you are authenticated.\u001b[0m\u001b[33m\"\u001b[0m                                        \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m290 \u001b[0m\u001b[2m│   │   │   \u001b[0m)                                                                              \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m291 \u001b[2m│   │   │   \u001b[0m\u001b[94mraise\u001b[0m RepositoryNotFoundError(message, response) \u001b[94mfrom\u001b[0m \u001b[4;96me\u001b[0m                        \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m292 \u001b[0m\u001b[2m│   │   \u001b[0m                                                                                   \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m293 \u001b[0m\u001b[2m│   │   \u001b[0m\u001b[94melif\u001b[0m response.status_code == \u001b[94m400\u001b[0m:                                                  \u001b[31m│\u001b[0m\n",
              "\u001b[31m│\u001b[0m   \u001b[2m294 \u001b[0m\u001b[2m│   │   │   \u001b[0mmessage = (                                                                    \u001b[31m│\u001b[0m\n",
              "\u001b[31m╰──────────────────────────────────────────────────────────────────────────────────────────────────╯\u001b[0m\n",
              "\u001b[1;91mRepositoryNotFoundError: \u001b[0m\u001b[1;36m401\u001b[0m Client Error. \u001b[1m(\u001b[0mRequest ID: \u001b[33mRoot\u001b[0m=\u001b[1;36m1\u001b[0m-64826c04-38f378fe7a63953f3fb17607\u001b[1m)\u001b[0m\n",
              "\n",
              "Repository Not Found for url: \u001b[4;94mhttps://huggingface.co/api/repos/create.\u001b[0m\n",
              "Please make sure you specified the correct `repo_id` and `repo_type`.\n",
              "If you are trying to access a private or gated repo, make sure you are authenticated.\n",
              "Invalid username or password.\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #800000; text-decoration-color: #800000\">╭─────────────────────────────── </span><span style=\"color: #800000; text-decoration-color: #800000; font-weight: bold\">Traceback </span><span style=\"color: #bf7f7f; text-decoration-color: #bf7f7f; font-weight: bold\">(most recent call last)</span><span style=\"color: #800000; text-decoration-color: #800000\"> ────────────────────────────────╮</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/</span><span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">_errors.py</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">259</span> in                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #00ff00; text-decoration-color: #00ff00\">hf_raise_for_status</span>                                                                              <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">256 </span><span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">│   </span><span style=\"color: #808000; text-decoration-color: #808000\">&lt;/Tip&gt;</span>                                                                                 <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">257 </span><span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">│   </span><span style=\"color: #808000; text-decoration-color: #808000\">\"\"\"</span>                                                                                    <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">258 │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">try</span>:                                                                                   <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span>259 <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│   │   </span>response.raise_for_status()                                                        <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">260 │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">except</span> HTTPError <span style=\"color: #0000ff; text-decoration-color: #0000ff\">as</span> e:                                                                 <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">261 │   │   </span>error_code = response.headers.get(<span style=\"color: #808000; text-decoration-color: #808000\">\"X-Error-Code\"</span>)                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">262 </span>                                                                                           <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">/usr/local/lib/python3.10/dist-packages/requests/</span><span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">models.py</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">960</span> in <span style=\"color: #00ff00; text-decoration-color: #00ff00\">raise_for_status</span>               <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">957 │   │   │   </span>http_error_msg = <span style=\"color: #808000; text-decoration-color: #808000\">u'%s Server Error: %s for url: %s'</span> % (<span style=\"color: #00ffff; text-decoration-color: #00ffff\">self</span>.status_code, rea   <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">958 │   │   </span>                                                                                   <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">959 │   │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">if</span> http_error_msg:                                                                 <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span>960 <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│   │   │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">raise</span> HTTPError(http_error_msg, response=<span style=\"color: #00ffff; text-decoration-color: #00ffff\">self</span>)                                 <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">961 │   </span>                                                                                       <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">962 │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">def</span> <span style=\"color: #00ff00; text-decoration-color: #00ff00\">close</span>(<span style=\"color: #00ffff; text-decoration-color: #00ffff\">self</span>):                                                                       <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">963 </span><span style=\"color: #bfbfbf; text-decoration-color: #bfbfbf\">│   │   </span><span style=\"color: #808000; text-decoration-color: #808000\">\"\"\"Releases the connection back to the pool. Once this method has been</span>             <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">╰──────────────────────────────────────────────────────────────────────────────────────────────────╯</span>\n",
              "<span style=\"color: #ff0000; text-decoration-color: #ff0000; font-weight: bold\">HTTPError: </span><span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">401</span> Client Error: Unauthorized for url: <span style=\"color: #0000ff; text-decoration-color: #0000ff; text-decoration: underline\">https://huggingface.co/api/repos/create</span>\n",
              "\n",
              "<span style=\"font-style: italic\">The above exception was the direct cause of the following exception:</span>\n",
              "\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">╭─────────────────────────────── </span><span style=\"color: #800000; text-decoration-color: #800000; font-weight: bold\">Traceback </span><span style=\"color: #bf7f7f; text-decoration-color: #bf7f7f; font-weight: bold\">(most recent call last)</span><span style=\"color: #800000; text-decoration-color: #800000\"> ────────────────────────────────╮</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> in <span style=\"color: #00ff00; text-decoration-color: #00ff00\">&lt;cell line: 8&gt;</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">8</span>                                                                              <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> in <span style=\"color: #00ff00; text-decoration-color: #00ff00\">push_to_hub</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">71</span>                                                                                <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">/usr/local/lib/python3.10/dist-packages/transformers/utils/</span><span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">hub.py</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">804</span> in <span style=\"color: #00ff00; text-decoration-color: #00ff00\">push_to_hub</span>             <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 801 │   │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">else</span>:                                                                             <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 802 │   │   │   </span>working_dir = repo_id.split(<span style=\"color: #808000; text-decoration-color: #808000\">\"/\"</span>)[-<span style=\"color: #0000ff; text-decoration-color: #0000ff\">1</span>]                                          <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 803 │   │   </span>                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span> 804 <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│   │   </span>repo_id = <span style=\"color: #00ffff; text-decoration-color: #00ffff\">self</span>._create_repo(                                                      <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 805 │   │   │   </span>repo_id, private=private, use_auth_token=use_auth_token, repo_url=repo_url,   <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 806 │   │   </span>)                                                                                 <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 807 </span>                                                                                          <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">/usr/local/lib/python3.10/dist-packages/transformers/utils/</span><span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">hub.py</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">661</span> in <span style=\"color: #00ff00; text-decoration-color: #00ff00\">_create_repo</span>            <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 658 │   │   │   │   │   </span>repo_id = repo_id.split(<span style=\"color: #808000; text-decoration-color: #808000\">\"/\"</span>)[-<span style=\"color: #0000ff; text-decoration-color: #0000ff\">1</span>]                                      <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 659 │   │   │   │   </span>repo_id = <span style=\"color: #808000; text-decoration-color: #808000\">f\"{</span>organization<span style=\"color: #808000; text-decoration-color: #808000\">}/{</span>repo_id<span style=\"color: #808000; text-decoration-color: #808000\">}\"</span>                                     <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 660 │   │   </span>                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span> 661 <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│   │   </span>url = create_repo(repo_id=repo_id, token=use_auth_token, private=private, exist_  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 662 │   │   </span>                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 663 │   │   # If the namespace is not there, add it or `upload_file` will complain</span>            <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"> 664 │   │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">if</span> <span style=\"color: #808000; text-decoration-color: #808000\">\"/\"</span> <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">not</span> <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">in</span> repo_id <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">and</span> url != <span style=\"color: #808000; text-decoration-color: #808000\">f\"{</span>HUGGINGFACE_CO_RESOLVE_ENDPOINT<span style=\"color: #808000; text-decoration-color: #808000\">}/{</span>repo_id<span style=\"color: #808000; text-decoration-color: #808000\">}\"</span>:  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/</span><span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">_validators.py</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">118</span> in <span style=\"color: #00ff00; text-decoration-color: #00ff00\">_inner_fn</span>    <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">115 │   │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">if</span> check_use_auth_token:                                                           <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">116 │   │   │   </span>kwargs = smoothly_deprecate_use_auth_token(fn_name=fn.<span style=\"color: #ff0000; text-decoration-color: #ff0000\">__name__</span>, has_token=ha   <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">117 │   │   </span>                                                                                   <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span>118 <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│   │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">return</span> fn(*args, **kwargs)                                                         <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">119 │   </span>                                                                                       <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">120 │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">return</span> _inner_fn  <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"># type: ignore</span>                                                       <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">121 </span>                                                                                           <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">/usr/local/lib/python3.10/dist-packages/huggingface_hub/</span><span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">hf_api.py</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">2306</span> in <span style=\"color: #00ff00; text-decoration-color: #00ff00\">create_repo</span>            <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2303 │   │   </span>r = get_session().post(path, headers=headers, json=json)                          <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2304 │   │   </span>                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2305 │   │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">try</span>:                                                                              <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span>2306 <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│   │   │   </span>hf_raise_for_status(r)                                                        <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2307 │   │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">except</span> HTTPError <span style=\"color: #0000ff; text-decoration-color: #0000ff\">as</span> err:                                                          <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2308 │   │   │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">if</span> exist_ok <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">and</span> err.response.status_code == <span style=\"color: #0000ff; text-decoration-color: #0000ff\">409</span>:                              <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2309 │   │   │   │   # Repo already exists and `exist_ok=True`</span>                                 <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/</span><span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">_errors.py</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">291</span> in                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #00ff00; text-decoration-color: #00ff00\">hf_raise_for_status</span>                                                                              <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>                                                                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">288 │   │   │   │   </span><span style=\"color: #808000; text-decoration-color: #808000\">\" `repo_type`.\\nIf you are trying to access a private or gated repo,\"</span>      <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">289 │   │   │   │   </span><span style=\"color: #808000; text-decoration-color: #808000\">\" make sure you are authenticated.\"</span>                                        <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">290 │   │   │   </span>)                                                                              <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span>291 <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│   │   │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">raise</span> RepositoryNotFoundError(message, response) <span style=\"color: #0000ff; text-decoration-color: #0000ff\">from</span> <span style=\"color: #00ffff; text-decoration-color: #00ffff; text-decoration: underline\">e</span>                        <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">292 │   │   </span>                                                                                   <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">293 │   │   </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">elif</span> response.status_code == <span style=\"color: #0000ff; text-decoration-color: #0000ff\">400</span>:                                                  <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">│</span>   <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">294 │   │   │   </span>message = (                                                                    <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
              "<span style=\"color: #800000; text-decoration-color: #800000\">╰──────────────────────────────────────────────────────────────────────────────────────────────────╯</span>\n",
              "<span style=\"color: #ff0000; text-decoration-color: #ff0000; font-weight: bold\">RepositoryNotFoundError: </span><span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">401</span> Client Error. <span style=\"font-weight: bold\">(</span>Request ID: <span style=\"color: #808000; text-decoration-color: #808000\">Root</span>=<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">1</span>-64826c04-38f378fe7a63953f3fb17607<span style=\"font-weight: bold\">)</span>\n",
              "\n",
              "Repository Not Found for url: <span style=\"color: #0000ff; text-decoration-color: #0000ff; text-decoration: underline\">https://huggingface.co/api/repos/create.</span>\n",
              "Please make sure you specified the correct `repo_id` and `repo_type`.\n",
              "If you are trying to access a private or gated repo, make sure you are authenticated.\n",
              "Invalid username or password.\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        }
      ],
      "source": [
        "from transformers import TrainingArguments\n",
        "\n",
        "model_checkpoint = \"m3hrdadfi/albert-fa-base-v2\"  # Specify the model checkpoint to use\n",
        "dataset = DATASETS[\"pquad\"]  # Select the dataset to train on\n",
        "trainer = TrainerQA(model_checkpoint, dataset)  # Create an instance of the TrainerQA class\n",
        "trainer.train()  # Train the model using the specified dataset\n",
        "trainer.evaluate()  # Evaluate the trained model\n",
        "trainer.push_to_hub(\"albert-fa-base-v2_pquad\")  # Push the trained model to the Hugging Face Model Hub\n"
      ]
    }
  ]
}