HopeSpeechEDI_BERT_English _v2.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "HopeSpeechEDI_BERT_English.ipynb",
      "provenance": [],
      "toc_visible": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU",
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "fa9fc5d63de741b79b5a1e551a33f8ec": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "state": {
            "_view_name": "HBoxView",
            "_dom_classes": [],
            "_model_name": "HBoxModel",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "box_style": "",
            "layout": "IPY_MODEL_27996e6c85474da4b82bb07d95152be4",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_0a8891f0d23b403d953e48255296e209",
              "IPY_MODEL_274e163afdb2477d9e2fd806347591e9"
            ]
          }
        },
        "27996e6c85474da4b82bb07d95152be4": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "0a8891f0d23b403d953e48255296e209": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_9a4a8053c28e41769b7912b5c2b40b0e",
            "_dom_classes": [],
            "description": "Downloading: 100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 433,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 433,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_ecf66b7de19e496c8c055ef846a6d5f9"
          }
        },
        "274e163afdb2477d9e2fd806347591e9": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_0e379111c44e4fa9a994f6917dbec311",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 433/433 [00:08&lt;00:00, 48.8B/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_98ea15bc5fc64ed683b9a90fa1a077c1"
          }
        },
        "9a4a8053c28e41769b7912b5c2b40b0e": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "ProgressStyleModel",
            "description_width": "initial",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "bar_color": null,
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "ecf66b7de19e496c8c055ef846a6d5f9": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "0e379111c44e4fa9a994f6917dbec311": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "DescriptionStyleModel",
            "description_width": "",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "98ea15bc5fc64ed683b9a90fa1a077c1": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "b0162a1dc13e4a03b9abdc95d3821e5b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "state": {
            "_view_name": "HBoxView",
            "_dom_classes": [],
            "_model_name": "HBoxModel",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "box_style": "",
            "layout": "IPY_MODEL_516716eaf84448928c0a1af754560d4b",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_c8a30b358cec4057b27e023eb5c38814",
              "IPY_MODEL_3fcf4328921c47a892ec171420e0c14e"
            ]
          }
        },
        "516716eaf84448928c0a1af754560d4b": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "c8a30b358cec4057b27e023eb5c38814": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_bfd6c70f296843deb56abf8a1aa33d94",
            "_dom_classes": [],
            "description": "Downloading: 100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 231508,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 231508,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_fb475406e0154be3903a6e475b8d0d13"
          }
        },
        "3fcf4328921c47a892ec171420e0c14e": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_64ddcf2ec31a4f959282b4e788a3386c",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 232k/232k [00:08&lt;00:00, 26.4kB/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_d22b6683ba684fa090cd31b20456f8c9"
          }
        },
        "bfd6c70f296843deb56abf8a1aa33d94": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "ProgressStyleModel",
            "description_width": "initial",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "bar_color": null,
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "fb475406e0154be3903a6e475b8d0d13": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "64ddcf2ec31a4f959282b4e788a3386c": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "DescriptionStyleModel",
            "description_width": "",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "d22b6683ba684fa090cd31b20456f8c9": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "b211934af71b4c9f96bc335f15c89a3f": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "state": {
            "_view_name": "HBoxView",
            "_dom_classes": [],
            "_model_name": "HBoxModel",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "box_style": "",
            "layout": "IPY_MODEL_cab29ff05da3498e8c7997e1eb5f3167",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_d1dc2dc61a754cf68078a8803fe05ef9",
              "IPY_MODEL_e63cdfdb93904aed9ccc71868f7e473c"
            ]
          }
        },
        "cab29ff05da3498e8c7997e1eb5f3167": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "d1dc2dc61a754cf68078a8803fe05ef9": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_4266bd9ccf944b43b8dcaddf738c054b",
            "_dom_classes": [],
            "description": "Downloading: 100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 466062,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 466062,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_89bfa715ad4f4599a07f59dd57fd348c"
          }
        },
        "e63cdfdb93904aed9ccc71868f7e473c": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_113904683b65459ba4ef274563a12051",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 466k/466k [00:08&lt;00:00, 53.7kB/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_0c3b3575d6354568ada3f6ed39a15fce"
          }
        },
        "4266bd9ccf944b43b8dcaddf738c054b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "ProgressStyleModel",
            "description_width": "initial",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "bar_color": null,
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "89bfa715ad4f4599a07f59dd57fd348c": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "113904683b65459ba4ef274563a12051": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "DescriptionStyleModel",
            "description_width": "",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "0c3b3575d6354568ada3f6ed39a15fce": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "18c4291e51e140d9a85ea1fe8d50ecc3": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "state": {
            "_view_name": "HBoxView",
            "_dom_classes": [],
            "_model_name": "HBoxModel",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "box_style": "",
            "layout": "IPY_MODEL_c3454a2687be43a1aa053816d60ba180",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_300a8f8d8b814fb798a020dffa57d955",
              "IPY_MODEL_f4912fb547764567ab9728e0b22eb3db"
            ]
          }
        },
        "c3454a2687be43a1aa053816d60ba180": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "300a8f8d8b814fb798a020dffa57d955": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_28906f1fd0f140f88a93c0ecf74749e9",
            "_dom_classes": [],
            "description": "Downloading: 100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 440473133,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 440473133,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_7a9432d5bd8b4739879b5eafe10614c9"
          }
        },
        "f4912fb547764567ab9728e0b22eb3db": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_8b5800c073bc4d379ccb18a369ab3482",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 440M/440M [00:09&lt;00:00, 46.4MB/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_05863112e5b746bd80b5b037c7417610"
          }
        },
        "28906f1fd0f140f88a93c0ecf74749e9": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "ProgressStyleModel",
            "description_width": "initial",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "bar_color": null,
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "7a9432d5bd8b4739879b5eafe10614c9": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "8b5800c073bc4d379ccb18a369ab3482": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "DescriptionStyleModel",
            "description_width": "",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "05863112e5b746bd80b5b037c7417610": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        }
      }
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "elqe9PQrVH6P"
      },
      "source": [
        "### Performing checks for the resources available"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fSELLChxa5su",
        "outputId": "26b6ed27-e46e-4efb-dde8-848d9ca57f96"
      },
      "source": [
        "import torch\r\n",
        "\r\n",
        "# If there's a GPU available...\r\n",
        "if torch.cuda.is_available():    \r\n",
        "\r\n",
        "    # Tell PyTorch to use the GPU.    \r\n",
        "    device = torch.device(\"cuda\")\r\n",
        "\r\n",
        "    print('There are %d GPU(s) available.' % torch.cuda.device_count())\r\n",
        "\r\n",
        "    print('We will use the GPU:', torch.cuda.get_device_name(0))\r\n",
        "\r\n",
        "# If not...\r\n",
        "else:\r\n",
        "    print('No GPU available, using the CPU instead.')\r\n",
        "    device = torch.device(\"cpu\")"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "There are 1 GPU(s) available.\n",
            "We will use the GPU: Tesla T4\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8SokVowWo3cc",
        "outputId": "5ff9cce2-74f4-4223-cb23-bfe5c2668b0c"
      },
      "source": [
        "!nvidia-smi"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Sun Jan 10 07:58:41 2021       \n",
            "+-----------------------------------------------------------------------------+\n",
            "| NVIDIA-SMI 460.27.04    Driver Version: 418.67       CUDA Version: 10.1     |\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   49C    P8    11W /  70W |     10MiB / 15079MiB |      0%      Default |\n",
            "|                               |                      |                 ERR! |\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"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "K-gbOgflVRxg"
      },
      "source": [
        "### Installing huggingface transformers library"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DbDuxKy0a9Zl",
        "outputId": "02e8e01c-27b0-4176-f2de-2636ff7919e5"
      },
      "source": [
        "!pip install transformers"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting transformers\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/50/0c/7d5950fcd80b029be0a8891727ba21e0cd27692c407c51261c3c921f6da3/transformers-4.1.1-py3-none-any.whl (1.5MB)\n",
            "\u001b[K     |████████████████████████████████| 1.5MB 16.4MB/s \n",
            "\u001b[?25hRequirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.6/dist-packages (from transformers) (4.41.1)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.6/dist-packages (from transformers) (3.0.12)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.6/dist-packages (from transformers) (2.23.0)\n",
            "Requirement already satisfied: packaging in /usr/local/lib/python3.6/dist-packages (from transformers) (20.8)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from transformers) (1.19.4)\n",
            "Collecting sacremoses\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/7d/34/09d19aff26edcc8eb2a01bed8e98f13a1537005d31e95233fd48216eed10/sacremoses-0.0.43.tar.gz (883kB)\n",
            "\u001b[K     |████████████████████████████████| 890kB 44.5MB/s \n",
            "\u001b[?25hRequirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.6/dist-packages (from transformers) (2019.12.20)\n",
            "Collecting tokenizers==0.9.4\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/0f/1c/e789a8b12e28be5bc1ce2156cf87cb522b379be9cadc7ad8091a4cc107c4/tokenizers-0.9.4-cp36-cp36m-manylinux2010_x86_64.whl (2.9MB)\n",
            "\u001b[K     |████████████████████████████████| 2.9MB 54.7MB/s \n",
            "\u001b[?25hRequirement already satisfied: dataclasses; python_version < \"3.7\" in /usr/local/lib/python3.6/dist-packages (from transformers) (0.8)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests->transformers) (3.0.4)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests->transformers) (2.10)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests->transformers) (2020.12.5)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests->transformers) (1.24.3)\n",
            "Requirement already satisfied: pyparsing>=2.0.2 in /usr/local/lib/python3.6/dist-packages (from packaging->transformers) (2.4.7)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from sacremoses->transformers) (1.15.0)\n",
            "Requirement already satisfied: click in /usr/local/lib/python3.6/dist-packages (from sacremoses->transformers) (7.1.2)\n",
            "Requirement already satisfied: joblib in /usr/local/lib/python3.6/dist-packages (from sacremoses->transformers) (1.0.0)\n",
            "Building wheels for collected packages: sacremoses\n",
            "  Building wheel for sacremoses (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for sacremoses: filename=sacremoses-0.0.43-cp36-none-any.whl size=893261 sha256=735710731e5d503e908d52a366f9861de68e31045516564d989f0556fbb49c14\n",
            "  Stored in directory: /root/.cache/pip/wheels/29/3c/fd/7ce5c3f0666dab31a50123635e6fb5e19ceb42ce38d4e58f45\n",
            "Successfully built sacremoses\n",
            "Installing collected packages: sacremoses, tokenizers, transformers\n",
            "Successfully installed sacremoses-0.0.43 tokenizers-0.9.4 transformers-4.1.1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bPwaYLiRSNPC"
      },
      "source": [
        "## Importing necessary packages"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0XdFrhr8SRG0"
      },
      "source": [
        "import pandas as pd\r\n",
        "import numpy as np \r\n",
        "import random\r\n",
        "from sklearn.preprocessing import LabelEncoder\r\n",
        "from sklearn.utils.class_weight import compute_class_weight\r\n",
        "import time\r\n",
        "import datetime\r\n",
        "from sklearn.metrics import classification_report,confusion_matrix\r\n",
        "import random\r\n",
        "import time\r\n",
        "import torch.nn as nn\r\n",
        "from transformers import AutoModel , AutoModelForSequenceClassification, AutoConfig , AutoTokenizer , AdamW ,get_linear_schedule_with_warmup \r\n",
        "import torch\r\n",
        "import matplotlib.pyplot as plt\r\n",
        "import seaborn as sns\r\n",
        "from matplotlib import rc\r\n",
        "from matplotlib.ticker import MaxNLocator"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dROX3juhVCxC"
      },
      "source": [
        "## Setting some configuration values"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Kzafv4BzVCC9"
      },
      "source": [
        "# Use plot styling from seaborn.\r\n",
        "sns.set(style='darkgrid')\r\n",
        "\r\n",
        "# Increase the plot size and font size.\r\n",
        "sns.set(font_scale=1.5)\r\n",
        "plt.rcParams[\"figure.figsize\"] = (16,12)\r\n",
        "\r\n",
        "# Set the seed value all over the place to make this reproducible.\r\n",
        "seed_val = 42\r\n",
        "\r\n",
        "random.seed(seed_val)\r\n",
        "np.random.seed(seed_val)\r\n",
        "torch.manual_seed(seed_val)\r\n",
        "torch.cuda.manual_seed_all(seed_val)"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2v1AIN_OVX4t"
      },
      "source": [
        "## Importing drive into the colaboratory"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Lofimyh2a_tp",
        "outputId": "0fe5da27-6df7-4130-f3bf-504ae93542b0"
      },
      "source": [
        "from google.colab import drive \r\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yOK-M00dVeW2"
      },
      "source": [
        "## Importing dataset directories"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-y-JRPKBbCTv"
      },
      "source": [
        "from os import getcwd , listdir \r\n",
        "from os.path import join\r\n",
        "curr_dir = getcwd()\r\n",
        "drive_dir = join(curr_dir , 'drive','MyDrive')\r\n",
        "dataset_dir = join(drive_dir,'Codalab','HopeSpeech Detection For Equality,DiversityAnd Inclusion','Dataset')"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7YU1h8WhmqvK",
        "outputId": "5642ab9b-1581-4988-fea5-ab77b23902f7"
      },
      "source": [
        "listdir(dataset_dir)"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['english_hope_dev.csv',\n",
              " 'english_hope_train.csv',\n",
              " 'malayalam_hope_dev.csv',\n",
              " 'malayalam_hope_train.csv',\n",
              " 'tamil_hope_first_dev.csv',\n",
              " 'tamil_hope_first_train.csv',\n",
              " 'tamil_hope_test.csv',\n",
              " 'malayalam_hope_test.csv',\n",
              " 'english_hope_test.csv']"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FUPkZxjLVjLK"
      },
      "source": [
        "## Loading training data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 377
        },
        "id": "joone3H4bINR",
        "outputId": "192718c5-b09b-4ec6-d67e-bbdc38a6de45"
      },
      "source": [
        "train_df_path = join(dataset_dir,'english_hope_train.csv')\r\n",
        "# Load the dataset into a pandas dataframe.\r\n",
        "train_df = pd.read_csv(train_df_path)\r\n",
        "\r\n",
        "# Report the number of sentences.\r\n",
        "print('Number of training sentences: {:,}\\n'.format(train_df.shape[0]))\r\n",
        "\r\n",
        "# Display 10 random rows from the data.\r\n",
        "train_df.sample(10)"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of training sentences: 22,762\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<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>text</th>\n",
              "      <th>label</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>22434</th>\n",
              "      <td>I want you to know that I watched this video w...</td>\n",
              "      <td>Hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9303</th>\n",
              "      <td>2020 is definitely the worst year I've witness...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>14935</th>\n",
              "      <td>My grandmother was superbutch and straight.</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9722</th>\n",
              "      <td>@Doc LMAO Liberals to have to fight alongside ...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>13345</th>\n",
              "      <td>​@Everything EVERYTHING!! Ironic how black peo...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7445</th>\n",
              "      <td>Me and my frnd literally had this discussion d...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5391</th>\n",
              "      <td>OMG sounds like a bunch of unattractive women ...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>18821</th>\n",
              "      <td>To the ones who think or feel... not all relig...</td>\n",
              "      <td>Hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1847</th>\n",
              "      <td>that girl that doesn’t “agree” with homosexual...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>11902</th>\n",
              "      <td>It would have been nice. I'm a female engineer...</td>\n",
              "      <td>Hope_speech</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                                    text            label\n",
              "22434  I want you to know that I watched this video w...      Hope_speech\n",
              "9303   2020 is definitely the worst year I've witness...  Non_hope_speech\n",
              "14935        My grandmother was superbutch and straight.  Non_hope_speech\n",
              "9722   @Doc LMAO Liberals to have to fight alongside ...  Non_hope_speech\n",
              "13345  ​@Everything EVERYTHING!! Ironic how black peo...  Non_hope_speech\n",
              "7445   Me and my frnd literally had this discussion d...  Non_hope_speech\n",
              "5391   OMG sounds like a bunch of unattractive women ...  Non_hope_speech\n",
              "18821  To the ones who think or feel... not all relig...      Hope_speech\n",
              "1847   that girl that doesn’t “agree” with homosexual...  Non_hope_speech\n",
              "11902  It would have been nice. I'm a female engineer...      Hope_speech"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "w4VW351225YK",
        "outputId": "114e699f-a39c-463c-ff9f-5f50868bde7f"
      },
      "source": [
        "train_df['label'].value_counts()"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Non_hope_speech    20778\n",
              "Hope_speech         1962\n",
              "not-English           22\n",
              "Name: label, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ooOv9QZtnLIF"
      },
      "source": [
        "\r\n",
        "le = LabelEncoder()\r\n",
        "train_df['label'] = le.fit_transform(train_df['label'])"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "a89Bf4LpnigU",
        "outputId": "6cfaff92-75af-4613-a0e9-f51cbb561def"
      },
      "source": [
        "train_df['label'].value_counts()"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1    20778\n",
              "0     1962\n",
              "2       22\n",
              "Name: label, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bvNeR4lTVnK0"
      },
      "source": [
        "### Loading `train_sentences` and `train_labels`"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EufQZx2dbKMT"
      },
      "source": [
        "# Get the lists of sentences and their labels.\r\n",
        "train_sentences = train_df['text'].values\r\n",
        "train_labels = train_df['label'].values"
      ],
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wnZSg5f7ZWs_"
      },
      "source": [
        "## Saving Class Names"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "83tTT-wbZa1o"
      },
      "source": [
        "class_names = ['Hope_speech' , 'Non-hope-speech'  , 'not-English']"
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ToUbAgZ9ap63"
      },
      "source": [
        "## Loading the validation data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 377
        },
        "id": "Dle8VzoAjOLS",
        "outputId": "b5e310d3-ed4d-47aa-a41c-bbcffb70b9e1"
      },
      "source": [
        "val_df_path = join(dataset_dir,'english_hope_dev.csv')\r\n",
        "# Load the dataset into a pandas dataframe.\r\n",
        "val_df = pd.read_csv(val_df_path)\r\n",
        "\r\n",
        "# Report the number of sentences.\r\n",
        "print('Number of training sentences: {:,}\\n'.format(val_df.shape[0]))\r\n",
        "\r\n",
        "# Display 10 random rows from the data.\r\n",
        "val_df.sample(10)"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of training sentences: 2,843\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<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>text</th>\n",
              "      <th>label</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>2834</th>\n",
              "      <td>She is the last truly 'intellectual' pop star</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1189</th>\n",
              "      <td>Yeah bicth and ur coming with us hahaha</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1426</th>\n",
              "      <td>BLM are certainly doing their best to exaspera...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1620</th>\n",
              "      <td>Professor the police where called on him for a...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2687</th>\n",
              "      <td>WHY COULDN'T THEY FIX HER MIC</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2240</th>\n",
              "      <td>@Phoebe Reynolds Also</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>566</th>\n",
              "      <td>God does not hate gaysnHe does not hate anyone...</td>\n",
              "      <td>Hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2420</th>\n",
              "      <td>@TheMighty Derp nnStop trying to convince your...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>842</th>\n",
              "      <td>@CLIP IT! woah woah woahhhhh chill. I'm bisexu...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1862</th>\n",
              "      <td>The brain mainly develops in the first 5 years...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                                   text            label\n",
              "2834      She is the last truly 'intellectual' pop star  Non_hope_speech\n",
              "1189            Yeah bicth and ur coming with us hahaha  Non_hope_speech\n",
              "1426  BLM are certainly doing their best to exaspera...  Non_hope_speech\n",
              "1620  Professor the police where called on him for a...  Non_hope_speech\n",
              "2687                      WHY COULDN'T THEY FIX HER MIC  Non_hope_speech\n",
              "2240                              @Phoebe Reynolds Also  Non_hope_speech\n",
              "566   God does not hate gaysnHe does not hate anyone...      Hope_speech\n",
              "2420  @TheMighty Derp nnStop trying to convince your...  Non_hope_speech\n",
              "842   @CLIP IT! woah woah woahhhhh chill. I'm bisexu...  Non_hope_speech\n",
              "1862  The brain mainly develops in the first 5 years...  Non_hope_speech"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AypCmtBcDV6v",
        "outputId": "d70e9c61-ba5a-4855-b68f-bbc08dd33475"
      },
      "source": [
        "val_df['label'].value_counts()"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Non_hope_speech    2569\n",
              "Hope_speech         272\n",
              "not-English           2\n",
              "Name: label, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-CgK81lGDPKC"
      },
      "source": [
        "val_df['label'] = le.transform(val_df['label'])"
      ],
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Yfgn-E6JDZQZ",
        "outputId": "8bab8747-d0e9-4316-89f9-fd592a65f0f4"
      },
      "source": [
        "val_df['label'].value_counts()"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1    2569\n",
              "0     272\n",
              "2       2\n",
              "Name: label, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DMdNNVqrDCuN"
      },
      "source": [
        "### Loading `val_sentences` and `val_labels`"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LvGhjUkaDCuO"
      },
      "source": [
        "# Get the lists of sentences and their labels.\r\n",
        "val_sentences = val_df['text'].values\r\n",
        "val_labels = val_df['label'].values"
      ],
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fS2UcOiEVuhm"
      },
      "source": [
        "## Handling class imbalance using `sklearn's compute_class_weight`\r\n",
        "\r\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WL4tP_YkI1mY"
      },
      "source": [
        "\r\n",
        "# #compute the class weights\r\n",
        "# class_wts = compute_class_weight('balanced', np.unique(train_labels), train_labels)\r\n",
        "\r\n",
        "# print(class_wts)"
      ],
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z_WjMTuibsrf"
      },
      "source": [
        "## Helper Functions"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Bef28eHNV-_d"
      },
      "source": [
        "#### 1) Update Interval Function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lBK0ujfEbMir"
      },
      "source": [
        "def good_update_interval(total_iters, num_desired_updates):\r\n",
        "    '''\r\n",
        "    This function will try to pick an intelligent progress update interval \r\n",
        "    based on the magnitude of the total iterations.\r\n",
        "\r\n",
        "    Parameters:\r\n",
        "      `total_iters` - The number of iterations in the for-loop.\r\n",
        "      `num_desired_updates` - How many times we want to see an update over the \r\n",
        "                              course of the for-loop.\r\n",
        "    '''\r\n",
        "    # Divide the total iterations by the desired number of updates. Most likely\r\n",
        "    # this will be some ugly number.\r\n",
        "    exact_interval = total_iters / num_desired_updates\r\n",
        "\r\n",
        "    # The `round` function has the ability to round down a number to, e.g., the\r\n",
        "    # nearest thousandth: round(exact_interval, -3)\r\n",
        "    #\r\n",
        "    # To determine the magnitude to round to, find the magnitude of the total,\r\n",
        "    # and then go one magnitude below that.\r\n",
        "\r\n",
        "    # Get the order of magnitude of the total.\r\n",
        "    order_of_mag = len(str(total_iters)) - 1\r\n",
        "\r\n",
        "    # Our update interval should be rounded to an order of magnitude smaller. \r\n",
        "    round_mag = order_of_mag - 1\r\n",
        "\r\n",
        "    # Round down and cast to an int.\r\n",
        "    update_interval = int(round(exact_interval, -round_mag))\r\n",
        "\r\n",
        "    # Don't allow the interval to be zero!\r\n",
        "    if update_interval == 0:\r\n",
        "        update_interval = 1\r\n",
        "\r\n",
        "    return update_interval"
      ],
      "execution_count": 21,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eMGycwGAWFyO"
      },
      "source": [
        "#### 2) Format time function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "c9zVyNQLm-3q"
      },
      "source": [
        "def format_time(elapsed):\r\n",
        "    '''\r\n",
        "    Takes a time in seconds and returns a string hh:mm:ss\r\n",
        "    '''\r\n",
        "    # Round to the nearest second.\r\n",
        "    elapsed_rounded = int(round((elapsed)))\r\n",
        "    \r\n",
        "    # Format as hh:mm:ss\r\n",
        "    return str(datetime.timedelta(seconds=elapsed_rounded))\r\n"
      ],
      "execution_count": 22,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EBAo04ENWw3n"
      },
      "source": [
        "#### 3) Make Smart Batches function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nof-AUeoiZc2"
      },
      "source": [
        "def make_smart_batches(text_samples, labels, batch_size,tokenizer,max_input_length):\r\n",
        "    '''\r\n",
        "    This function combines all of the required steps to prepare batches.\r\n",
        "    '''\r\n",
        "\r\n",
        "    print('Creating Smart Batches from {:,} examples with batch size {:,}...\\n'.format(len(text_samples), batch_size))\r\n",
        "\r\n",
        "    # =========================\r\n",
        "    #   Tokenize & Truncate\r\n",
        "    # =========================\r\n",
        "\r\n",
        "    full_input_ids = []\r\n",
        "\r\n",
        "    # Tokenize all training examples\r\n",
        "    print('Tokenizing {:,} samples...'.format(len(labels)))\r\n",
        "\r\n",
        "    # Choose an interval on which to print progress updates.\r\n",
        "    update_interval = good_update_interval(total_iters=len(labels), num_desired_updates=10)\r\n",
        "\r\n",
        "    # For each training example...\r\n",
        "    for text in text_samples:\r\n",
        "        \r\n",
        "        # Report progress.\r\n",
        "        if ((len(full_input_ids) % update_interval) == 0):\r\n",
        "            print('  Tokenized {:,} samples.'.format(len(full_input_ids)))\r\n",
        "\r\n",
        "        # Tokenize the sample.\r\n",
        "        input_ids = tokenizer.encode(text=text,              # Text to encode.\r\n",
        "                                    add_special_tokens=True, # Do add specials.\r\n",
        "                                    max_length=max_input_length,      \r\n",
        "                                    truncation=True,         # Do Truncate!\r\n",
        "                                    padding=False)           # DO NOT pad.\r\n",
        "                                    \r\n",
        "        # Add the tokenized result to our list.\r\n",
        "        full_input_ids.append(input_ids)\r\n",
        "        \r\n",
        "    print('DONE.')\r\n",
        "    print('{:>10,} samples\\n'.format(len(full_input_ids)))\r\n",
        "\r\n",
        "    # =========================\r\n",
        "    #      Select Batches\r\n",
        "    # =========================    \r\n",
        "\r\n",
        "    # Sort the two lists together by the length of the input sequence.\r\n",
        "    samples = sorted(zip(full_input_ids, labels), key=lambda x: len(x[0]))\r\n",
        "\r\n",
        "    print('{:>10,} samples after sorting\\n'.format(len(samples)))\r\n",
        "\r\n",
        "    import random\r\n",
        "\r\n",
        "    # List of batches that we'll construct.\r\n",
        "    batch_ordered_sentences = []\r\n",
        "    batch_ordered_labels = []\r\n",
        "\r\n",
        "    print('Creating batches of size {:}...'.format(batch_size))\r\n",
        "\r\n",
        "    # Choose an interval on which to print progress updates.\r\n",
        "    update_interval = good_update_interval(total_iters=len(samples), num_desired_updates=10)\r\n",
        "    \r\n",
        "    # Loop over all of the input samples...    \r\n",
        "    while len(samples) > 0:\r\n",
        "        \r\n",
        "        # Report progress.\r\n",
        "        if ((len(batch_ordered_sentences) % update_interval) == 0 \\\r\n",
        "            and not len(batch_ordered_sentences) == 0):\r\n",
        "            print('  Selected {:,} batches.'.format(len(batch_ordered_sentences)))\r\n",
        "\r\n",
        "        # `to_take` is our actual batch size. It will be `batch_size` until \r\n",
        "        # we get to the last batch, which may be smaller. \r\n",
        "        to_take = min(batch_size, len(samples))\r\n",
        "\r\n",
        "        # Pick a random index in the list of remaining samples to start\r\n",
        "        # our batch at.\r\n",
        "        select = random.randint(0, len(samples) - to_take)\r\n",
        "\r\n",
        "        # Select a contiguous batch of samples starting at `select`.\r\n",
        "        #print(\"Selecting batch from {:} to {:}\".format(select, select+to_take))\r\n",
        "        batch = samples[select:(select + to_take)]\r\n",
        "\r\n",
        "        #print(\"Batch length:\", len(batch))\r\n",
        "\r\n",
        "        # Each sample is a tuple--split them apart to create a separate list of \r\n",
        "        # sequences and a list of labels for this batch.\r\n",
        "        batch_ordered_sentences.append([s[0] for s in batch])\r\n",
        "        batch_ordered_labels.append([s[1] for s in batch])\r\n",
        "\r\n",
        "        # Remove these samples from the list.\r\n",
        "        del samples[select:select + to_take]\r\n",
        "\r\n",
        "    print('\\n  DONE - Selected {:,} batches.\\n'.format(len(batch_ordered_sentences)))\r\n",
        "\r\n",
        "    # =========================\r\n",
        "    #        Add Padding\r\n",
        "    # =========================    \r\n",
        "\r\n",
        "    print('Padding out sequences within each batch...')\r\n",
        "\r\n",
        "    py_inputs = []\r\n",
        "    py_attn_masks = []\r\n",
        "    py_labels = []\r\n",
        "\r\n",
        "    # For each batch...\r\n",
        "    for (batch_inputs, batch_labels) in zip(batch_ordered_sentences, batch_ordered_labels):\r\n",
        "\r\n",
        "        # New version of the batch, this time with padded sequences and now with\r\n",
        "        # attention masks defined.\r\n",
        "        batch_padded_inputs = []\r\n",
        "        batch_attn_masks = []\r\n",
        "        \r\n",
        "        # First, find the longest sample in the batch. \r\n",
        "        # Note that the sequences do currently include the special tokens!\r\n",
        "        max_size = max([len(sen) for sen in batch_inputs])\r\n",
        "\r\n",
        "        # For each input in this batch...\r\n",
        "        for sen in batch_inputs:\r\n",
        "            \r\n",
        "            # How many pad tokens do we need to add?\r\n",
        "            num_pads = max_size - len(sen)\r\n",
        "\r\n",
        "            # Add `num_pads` padding tokens to the end of the sequence.\r\n",
        "            padded_input = sen + [tokenizer.pad_token_id]*num_pads\r\n",
        "\r\n",
        "            # Define the attention mask--it's just a `1` for every real token\r\n",
        "            # and a `0` for every padding token.\r\n",
        "            attn_mask = [1] * len(sen) + [0] * num_pads\r\n",
        "\r\n",
        "            # Add the padded results to the batch.\r\n",
        "            batch_padded_inputs.append(padded_input)\r\n",
        "            batch_attn_masks.append(attn_mask)\r\n",
        "\r\n",
        "        # Our batch has been padded, so we need to save this updated batch.\r\n",
        "        # We also need the inputs to be PyTorch tensors, so we'll do that here.\r\n",
        "        # Todo - Michael's code specified \"dtype=torch.long\"\r\n",
        "        py_inputs.append(torch.tensor(batch_padded_inputs))\r\n",
        "        py_attn_masks.append(torch.tensor(batch_attn_masks))\r\n",
        "        py_labels.append(torch.tensor(batch_labels))\r\n",
        "    \r\n",
        "    print('  DONE.')\r\n",
        "\r\n",
        "    # Return the smart-batched dataset!\r\n",
        "    return (py_inputs, py_attn_masks, py_labels)\r\n"
      ],
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xo1_bDriha4q"
      },
      "source": [
        "### 4) Make Smart Batches (On Test)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hC-c82iuRlv2"
      },
      "source": [
        "def make_smart_batches_on_test(text_samples, text_ids, batch_size,tokenizer,max_input_length):\r\n",
        "    '''\r\n",
        "    This function combines all of the required steps to prepare batches.\r\n",
        "    '''\r\n",
        "\r\n",
        "    print('Creating Smart Batches from {:,} examples with batch size {:,}...\\n'.format(len(text_samples), batch_size))\r\n",
        "\r\n",
        "    # =========================\r\n",
        "    #   Tokenize & Truncate\r\n",
        "    # =========================\r\n",
        "\r\n",
        "    full_input_ids = []\r\n",
        "\r\n",
        "    # Tokenize all training examples\r\n",
        "    print('Tokenizing {:,} samples...'.format(len(text_samples)))\r\n",
        "\r\n",
        "    # Choose an interval on which to print progress updates.\r\n",
        "    update_interval = good_update_interval(total_iters=len(text_samples), num_desired_updates=10)\r\n",
        "\r\n",
        "    # For each training example...\r\n",
        "    for text in text_samples:\r\n",
        "        \r\n",
        "        # Report progress.\r\n",
        "        if ((len(full_input_ids) % update_interval) == 0):\r\n",
        "            print('  Tokenized {:,} samples.'.format(len(full_input_ids)))\r\n",
        "\r\n",
        "        # Tokenize the sample.\r\n",
        "        input_ids = tokenizer.encode(text=text,              # Text to encode.\r\n",
        "                                    add_special_tokens=True, # Do add specials.\r\n",
        "                                    max_length=max_input_length,      \r\n",
        "                                    truncation=True,         # Do Truncate!\r\n",
        "                                    padding=False)           # DO NOT pad.\r\n",
        "                                    \r\n",
        "        # Add the tokenized result to our list.\r\n",
        "        full_input_ids.append(input_ids)\r\n",
        "        \r\n",
        "    print('DONE.')\r\n",
        "    print('{:>10,} samples\\n'.format(len(full_input_ids)))\r\n",
        "\r\n",
        "    # =========================\r\n",
        "    #      Select Batches\r\n",
        "    # =========================    \r\n",
        "\r\n",
        "    # Sort the two lists together by the length of the input sequence.\r\n",
        "    samples = sorted(zip(full_input_ids, text_ids), key=lambda x: len(x[0]))\r\n",
        "\r\n",
        "    print('{:>10,} samples after sorting\\n'.format(len(samples)))\r\n",
        "\r\n",
        "    import random\r\n",
        "\r\n",
        "    # List of batches that we'll construct.\r\n",
        "    batch_ordered_sentences = []\r\n",
        "    batch_ordered_ids = []\r\n",
        "\r\n",
        "    print('Creating batches of size {:}...'.format(batch_size))\r\n",
        "\r\n",
        "    # Choose an interval on which to print progress updates.\r\n",
        "    update_interval = good_update_interval(total_iters=len(samples), num_desired_updates=10)\r\n",
        "    \r\n",
        "    # Loop over all of the input samples...    \r\n",
        "    while len(samples) > 0:\r\n",
        "        \r\n",
        "        # Report progress.\r\n",
        "        if ((len(batch_ordered_sentences) % update_interval) == 0 \\\r\n",
        "            and not len(batch_ordered_sentences) == 0):\r\n",
        "            print('  Selected {:,} batches.'.format(len(batch_ordered_sentences)))\r\n",
        "\r\n",
        "        # `to_take` is our actual batch size. It will be `batch_size` until \r\n",
        "        # we get to the last batch, which may be smaller. \r\n",
        "        to_take = min(batch_size, len(samples))\r\n",
        "\r\n",
        "        # Pick a random index in the list of remaining samples to start\r\n",
        "        # our batch at.\r\n",
        "        select = random.randint(0, len(samples) - to_take)\r\n",
        "\r\n",
        "        # Select a contiguous batch of samples starting at `select`.\r\n",
        "        #print(\"Selecting batch from {:} to {:}\".format(select, select+to_take))\r\n",
        "        batch = samples[select:(select + to_take)]\r\n",
        "\r\n",
        "        #print(\"Batch length:\", len(batch))\r\n",
        "\r\n",
        "        # Each sample is a tuple--split them apart to create a separate list of \r\n",
        "        # sequences and a list of labels for this batch.\r\n",
        "        batch_ordered_sentences.append([s[0] for s in batch])\r\n",
        "        batch_ordered_ids.append([s[1] for s in batch])\r\n",
        "\r\n",
        "        # Remove these samples from the list.\r\n",
        "        del samples[select:select + to_take]\r\n",
        "\r\n",
        "    print('\\n  DONE - Selected {:,} batches.\\n'.format(len(batch_ordered_sentences)))\r\n",
        "\r\n",
        "    # =========================\r\n",
        "    #        Add Padding\r\n",
        "    # =========================    \r\n",
        "\r\n",
        "    print('Padding out sequences within each batch...')\r\n",
        "\r\n",
        "    py_inputs = []\r\n",
        "    py_attn_masks = []\r\n",
        "    py_ids = []\r\n",
        "\r\n",
        "    # For each batch...\r\n",
        "    for (batch_inputs, batch_ids) in zip(batch_ordered_sentences, batch_ordered_ids):\r\n",
        "\r\n",
        "        # New version of the batch, this time with padded sequences and now with\r\n",
        "        # attention masks defined.\r\n",
        "        batch_padded_inputs = []\r\n",
        "        batch_attn_masks = []\r\n",
        "        \r\n",
        "        # First, find the longest sample in the batch. \r\n",
        "        # Note that the sequences do currently include the special tokens!\r\n",
        "        max_size = max([len(sen) for sen in batch_inputs])\r\n",
        "\r\n",
        "        # For each input in this batch...\r\n",
        "        for sen in batch_inputs:\r\n",
        "            \r\n",
        "            # How many pad tokens do we need to add?\r\n",
        "            num_pads = max_size - len(sen)\r\n",
        "\r\n",
        "            # Add `num_pads` padding tokens to the end of the sequence.\r\n",
        "            padded_input = sen + [tokenizer.pad_token_id]*num_pads\r\n",
        "\r\n",
        "            # Define the attention mask--it's just a `1` for every real token\r\n",
        "            # and a `0` for every padding token.\r\n",
        "            attn_mask = [1] * len(sen) + [0] * num_pads\r\n",
        "\r\n",
        "            # Add the padded results to the batch.\r\n",
        "            batch_padded_inputs.append(padded_input)\r\n",
        "            batch_attn_masks.append(attn_mask)\r\n",
        "\r\n",
        "        # Our batch has been padded, so we need to save this updated batch.\r\n",
        "        # We also need the inputs to be PyTorch tensors, so we'll do that here.\r\n",
        "        # Todo - Michael's code specified \"dtype=torch.long\"\r\n",
        "        py_inputs.append(torch.tensor(batch_padded_inputs))\r\n",
        "        py_attn_masks.append(torch.tensor(batch_attn_masks))\r\n",
        "        py_ids.append(torch.tensor(batch_ids))\r\n",
        "    \r\n",
        "    print('  DONE.')\r\n",
        "\r\n",
        "    # Return the smart-batched dataset!\r\n",
        "    return (py_inputs, py_attn_masks, py_ids)\r\n"
      ],
      "execution_count": 24,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RfXvAn47Rlv5"
      },
      "source": [
        "### 5) Function for calculating accuracy per class"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v3usMdEvRlv7"
      },
      "source": [
        "def accuracy_per_class(preds, labels , label_dict):\r\n",
        "    \r\n",
        "    preds_flat = np.argmax(preds, axis=1).flatten()\r\n",
        "    labels_flat = labels.flatten()\r\n",
        "\r\n",
        "    for label in np.unique(labels_flat):\r\n",
        "        y_preds = preds_flat[labels_flat==label]\r\n",
        "        y_true = labels_flat[labels_flat==label]\r\n",
        "        print(f'Class: {label_dict[label]}')\r\n",
        "        print(f'Accuracy: {round(100*len(y_preds[y_preds==label])/len(y_true) , 2)}\\n')"
      ],
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aF5EckowRlv9"
      },
      "source": [
        "### 6) Function for plotting training history"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KVPwVOGxRlv-"
      },
      "source": [
        "def plot_training_history(history):\r\n",
        "  fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(18, 6))\r\n",
        "\r\n",
        "  ax1.plot(history['train_loss'], label='train loss')\r\n",
        "  ax1.plot(history['val_loss'], label='validation loss')\r\n",
        "\r\n",
        "  ax1.xaxis.set_major_locator(MaxNLocator(integer=True))\r\n",
        "  ax1.set_ylim([-0.05, 1.05])\r\n",
        "  ax1.legend()\r\n",
        "  ax1.set_ylabel('Loss')\r\n",
        "  ax1.set_xlabel('Epoch')\r\n",
        "\r\n",
        "  ax2.plot(history['train_acc'], label='train accuracy')\r\n",
        "  ax2.plot(history['val_acc'], label='validation accuracy')\r\n",
        "\r\n",
        "  ax2.xaxis.set_major_locator(MaxNLocator(integer=True))\r\n",
        "  ax2.set_ylim([-0.05, 1.05])\r\n",
        "  ax2.legend()\r\n",
        "\r\n",
        "  ax2.set_ylabel('Accuracy')\r\n",
        "  ax2.set_xlabel('Epoch')\r\n",
        "\r\n",
        "  fig.suptitle('Training history')"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EWrNYpQ_Rlv_"
      },
      "source": [
        "### 7) Function to check accuracy"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IudqmF2tRlwA"
      },
      "source": [
        "def check_accuracy(predictions,true_labels):\r\n",
        "  # Combine the results across the batches.\r\n",
        "  predictions = np.concatenate(predictions, axis=0)\r\n",
        "  true_labels = np.concatenate(true_labels, axis=0)\r\n",
        "\r\n",
        "  # Choose the label with the highest score as our prediction.\r\n",
        "  preds = np.argmax(predictions, axis=1).flatten()\r\n",
        "\r\n",
        "  # Calculate simple flat accuracy -- number correct over total number.\r\n",
        "  accuracy = (preds == true_labels).mean()\r\n",
        "\r\n",
        "  return accuracy"
      ],
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z4NcJu-eRlwB"
      },
      "source": [
        "### 8) Function to evaluate model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "znBBqCPlRlwC"
      },
      "source": [
        "def eval_model(model,py_inputs, py_attn_masks, py_labels):\r\n",
        "# Prediction on test set\r\n",
        "  t0 = time.time()\r\n",
        "\r\n",
        "  print('Doing validation on {:,} sentences...'.format(len(py_labels)))\r\n",
        "\r\n",
        "  # Put model in evaluation mode\r\n",
        "  model.eval()\r\n",
        "\r\n",
        "  # Tracking variables \r\n",
        "  predictions , true_labels = [], []\r\n",
        "\r\n",
        "  # Choose an interval on which to print progress updates.\r\n",
        "  update_interval = good_update_interval(total_iters=len(py_inputs), num_desired_updates=10)\r\n",
        "\r\n",
        "  # Measure elapsed time.\r\n",
        "  t0 = time.time()\r\n",
        "  total_val_loss = 0\r\n",
        "\r\n",
        "  # Put model in evaluation mode\r\n",
        "  model.eval()\r\n",
        "\r\n",
        "  # For each batch of training data...\r\n",
        "  for step in range(0, len(py_inputs)):\r\n",
        "\r\n",
        "      # Progress update every 100 batches.\r\n",
        "      if step % update_interval == 0 and not step == 0:\r\n",
        "          # Calculate elapsed time in minutes.\r\n",
        "          elapsed = format_time(time.time() - t0)\r\n",
        "          \r\n",
        "          # Calculate the time remaining based on our progress.\r\n",
        "          steps_per_sec = (time.time() - t0) / step\r\n",
        "          remaining_sec = steps_per_sec * (len(py_inputs) - step)\r\n",
        "          remaining = format_time(remaining_sec)\r\n",
        "\r\n",
        "          # Report progress.\r\n",
        "          print('  Batch {:>7,}  of  {:>7,}.    Elapsed: {:}.  Remaining: {:}'.format(step, len(py_inputs), elapsed, remaining))\r\n",
        "\r\n",
        "      # Copy the batch to the GPU.\r\n",
        "      b_input_ids = py_inputs[step].to(device)\r\n",
        "      b_input_mask = py_attn_masks[step].to(device)\r\n",
        "      b_labels = py_labels[step].to(device)\r\n",
        "    \r\n",
        "      # Telling the model not to compute or store gradients, saving memory and \r\n",
        "      # speeding up prediction\r\n",
        "      with torch.no_grad():\r\n",
        "        output = model(b_input_ids, \r\n",
        "                             token_type_ids=None, \r\n",
        "                             attention_mask=b_input_mask, \r\n",
        "                             labels=b_labels)\r\n",
        "\r\n",
        "\r\n",
        "      logits = output.logits \r\n",
        "      loss = output.loss\r\n",
        "      # Move logits and labels to CPU\r\n",
        "      logits = logits.detach().cpu().numpy()\r\n",
        "      label_ids = b_labels.to('cpu').numpy()\r\n",
        "    \r\n",
        "      # Store predictions and true labels\r\n",
        "      predictions.append(logits)\r\n",
        "      true_labels.append(label_ids)\r\n",
        "\r\n",
        "      total_val_loss += loss.item()\r\n",
        "  \r\n",
        "  val_accuracy = check_accuracy(predictions,true_labels)\r\n",
        "\r\n",
        "  validation_time = format_time(time.time() - t0)\r\n",
        "  \r\n",
        "  # Calculate the average loss over all of the batches.\r\n",
        "  avg_val_loss = total_val_loss / len(py_inputs)     \r\n",
        "    \r\n",
        "\r\n",
        "  # print('Accuracy: {:.3f}'.format(val_accuracy))\r\n",
        "  return (avg_val_loss,val_accuracy,validation_time)\r\n"
      ],
      "execution_count": 28,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kiX8z2qwRlwD"
      },
      "source": [
        "### 9) Function for making predictions on our test dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "15KDyc3xRlwE"
      },
      "source": [
        "def get_predictions(py_inputs, py_attn_masks, py_labels):\r\n",
        "\r\n",
        "  print('Predicting labels for {:,} test batches...'.format(len(py_labels)))\r\n",
        "\r\n",
        "  # Put model in evaluation mode\r\n",
        "  model.eval()\r\n",
        "\r\n",
        "  # Tracking variables \r\n",
        "  predictions , true_labels = [], []\r\n",
        "\r\n",
        "  # Choose an interval on which to print progress updates.\r\n",
        "  update_interval = good_update_interval(total_iters=len(py_inputs), num_desired_updates=10)\r\n",
        "\r\n",
        "  # Measure elapsed time.\r\n",
        "  t0 = time.time()\r\n",
        "\r\n",
        "  # Put model in evaluation mode\r\n",
        "  model.eval()\r\n",
        "\r\n",
        "  # For each batch of training data...\r\n",
        "  for step in range(0, len(py_inputs)):\r\n",
        "\r\n",
        "      # Progress update every 100 batches.\r\n",
        "      if step % update_interval == 0 and not step == 0:\r\n",
        "          # Calculate elapsed time in minutes.\r\n",
        "          elapsed = format_time(time.time() - t0)\r\n",
        "          \r\n",
        "          # Calculate the time remaining based on our progress.\r\n",
        "          steps_per_sec = (time.time() - t0) / step\r\n",
        "          remaining_sec = steps_per_sec * (len(py_inputs) - step)\r\n",
        "          remaining = format_time(remaining_sec)\r\n",
        "\r\n",
        "          # Report progress.\r\n",
        "          print('  Batch {:>7,}  of  {:>7,}.    Elapsed: {:}.  Remaining: {:}'.format(step, len(py_inputs), elapsed, remaining))\r\n",
        "\r\n",
        "      # Copy the batch to the GPU.\r\n",
        "      b_input_ids = py_inputs[step].to(device)\r\n",
        "      b_input_mask = py_attn_masks[step].to(device)\r\n",
        "      b_labels = py_labels[step].to(device)\r\n",
        "    \r\n",
        "      # Telling the model not to compute or store gradients, saving memory and \r\n",
        "      # speeding up prediction\r\n",
        "      with torch.no_grad():\r\n",
        "          # Forward pass, calculate logit predictions\r\n",
        "          outputs = model(b_input_ids, b_input_mask)\r\n",
        "\r\n",
        "      logits = outputs.logits \r\n",
        "\r\n",
        "      # Move logits and labels to CPU\r\n",
        "      logits = logits.detach().cpu().numpy()\r\n",
        "      label_ids = b_labels.to('cpu').numpy()\r\n",
        "    \r\n",
        "      # Store predictions and true labels\r\n",
        "      predictions.append(logits)\r\n",
        "      true_labels.append(label_ids)\r\n",
        "      # Combine the results across the batches.\r\n",
        "  predictions = np.concatenate(predictions, axis=0)\r\n",
        "  true_labels = np.concatenate(true_labels, axis=0)\r\n",
        "\r\n",
        "      # Choose the label with the highest score as our prediction.\r\n",
        "  preds = np.argmax(predictions, axis=1).flatten()\r\n",
        "\r\n",
        "  return (preds,true_labels)"
      ],
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aJEMLZiZRlwF"
      },
      "source": [
        "### 10) Function for making predictions on test dataset(without labels)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PSmN7uOaRlwG"
      },
      "source": [
        "def get_predictions_test(py_inputs, py_attn_masks,py_ids):\r\n",
        "\r\n",
        "  print('Predicting labels for {:,} test batches...'.format(len(py_inputs)))\r\n",
        "\r\n",
        "  # Put model in evaluation mode\r\n",
        "  model.eval()\r\n",
        "\r\n",
        "  # Tracking variables \r\n",
        "  predictions , true_labels ,ids = [], [] , []\r\n",
        "\r\n",
        "  # Choose an interval on which to print progress updates.\r\n",
        "  update_interval = good_update_interval(total_iters=len(py_inputs), num_desired_updates=10)\r\n",
        "\r\n",
        "  # Measure elapsed time.\r\n",
        "  t0 = time.time()\r\n",
        "\r\n",
        "  # Put model in evaluation mode\r\n",
        "  model.eval()\r\n",
        "\r\n",
        "  # For each batch of training data...\r\n",
        "  for step in range(0, len(py_inputs)):\r\n",
        "\r\n",
        "      # Progress update every 100 batches.\r\n",
        "      if step % update_interval == 0 and not step == 0:\r\n",
        "          # Calculate elapsed time in minutes.\r\n",
        "          elapsed = format_time(time.time() - t0)\r\n",
        "          \r\n",
        "          # Calculate the time remaining based on our progress.\r\n",
        "          steps_per_sec = (time.time() - t0) / step\r\n",
        "          remaining_sec = steps_per_sec * (len(py_inputs) - step)\r\n",
        "          remaining = format_time(remaining_sec)\r\n",
        "\r\n",
        "          # Report progress.\r\n",
        "          print('  Batch {:>7,}  of  {:>7,}.    Elapsed: {:}.  Remaining: {:}'.format(step, len(py_inputs), elapsed, remaining))\r\n",
        "\r\n",
        "      # Copy the batch to the GPU.\r\n",
        "      b_input_ids = py_inputs[step].to(device)\r\n",
        "      b_input_mask = py_attn_masks[step].to(device)\r\n",
        "      b_ids = py_ids[step].to(device)\r\n",
        "    \r\n",
        "      # Telling the model not to compute or store gradients, saving memory and \r\n",
        "      # speeding up prediction\r\n",
        "      with torch.no_grad():\r\n",
        "          # Forward pass, calculate logit predictions\r\n",
        "          outputs = model(b_input_ids, b_input_mask)\r\n",
        "\r\n",
        "      logits = outputs.logits \r\n",
        "\r\n",
        "      # Move logits and labels to CPU\r\n",
        "      logits = logits.detach().cpu().numpy()\r\n",
        "      b_ids = b_ids.detach().cpu().numpy()\r\n",
        "      # Store predictions\r\n",
        "      predictions.append(logits)\r\n",
        "      ids.append(b_ids)\r\n",
        "      # Combine the results across the batches.\r\n",
        "  predictions = np.concatenate(predictions, axis=0)\r\n",
        "  ids = np.concatenate(ids,axis=0)\r\n",
        "  # Choose the label with the highest score as our prediction.\r\n",
        "  preds = np.argmax(predictions, axis=1).flatten()\r\n",
        "\r\n",
        "  return ids,preds"
      ],
      "execution_count": 30,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BG6CAheNRlwK"
      },
      "source": [
        "### 11) Confusion Matrix Helper Function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VDFOg-jKRlwL"
      },
      "source": [
        "def show_confusion_matrix(confusion_matrix, class_names):\r\n",
        "\r\n",
        "  cm = confusion_matrix.copy()\r\n",
        "\r\n",
        "  cell_counts = cm.flatten()\r\n",
        "\r\n",
        "  cm_row_norm = cm / cm.sum(axis=1)[:, np.newaxis]\r\n",
        "\r\n",
        "  row_percentages = [\"{0:.2f}\".format(value) for value in cm_row_norm.flatten()]\r\n",
        "\r\n",
        "  cell_labels = [f\"{cnt}\\n{per}\" for cnt, per in zip(cell_counts, row_percentages)]\r\n",
        "  cell_labels = np.asarray(cell_labels).reshape(cm.shape[0], cm.shape[1])\r\n",
        "\r\n",
        "  df_cm = pd.DataFrame(cm_row_norm, index=class_names, columns=class_names)\r\n",
        "\r\n",
        "  hmap = sns.heatmap(df_cm, annot=cell_labels, fmt=\"\", cmap=\"Blues\")\r\n",
        "  hmap.yaxis.set_ticklabels(hmap.yaxis.get_ticklabels(), rotation=0, ha='right')\r\n",
        "  hmap.xaxis.set_ticklabels(hmap.xaxis.get_ticklabels(), rotation=30, ha='right')\r\n",
        "  plt.ylabel('True Sign')\r\n",
        "  plt.xlabel('Predicted Sign');"
      ],
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MBxHY4rMojd-"
      },
      "source": [
        "## Loading the tokenizer using AutoTokenizer"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 180,
          "referenced_widgets": [
            "fa9fc5d63de741b79b5a1e551a33f8ec",
            "27996e6c85474da4b82bb07d95152be4",
            "0a8891f0d23b403d953e48255296e209",
            "274e163afdb2477d9e2fd806347591e9",
            "9a4a8053c28e41769b7912b5c2b40b0e",
            "ecf66b7de19e496c8c055ef846a6d5f9",
            "0e379111c44e4fa9a994f6917dbec311",
            "98ea15bc5fc64ed683b9a90fa1a077c1",
            "b0162a1dc13e4a03b9abdc95d3821e5b",
            "516716eaf84448928c0a1af754560d4b",
            "c8a30b358cec4057b27e023eb5c38814",
            "3fcf4328921c47a892ec171420e0c14e",
            "bfd6c70f296843deb56abf8a1aa33d94",
            "fb475406e0154be3903a6e475b8d0d13",
            "64ddcf2ec31a4f959282b4e788a3386c",
            "d22b6683ba684fa090cd31b20456f8c9",
            "b211934af71b4c9f96bc335f15c89a3f",
            "cab29ff05da3498e8c7997e1eb5f3167",
            "d1dc2dc61a754cf68078a8803fe05ef9",
            "e63cdfdb93904aed9ccc71868f7e473c",
            "4266bd9ccf944b43b8dcaddf738c054b",
            "89bfa715ad4f4599a07f59dd57fd348c",
            "113904683b65459ba4ef274563a12051",
            "0c3b3575d6354568ada3f6ed39a15fce"
          ]
        },
        "id": "fIlwEvcMb4pb",
        "outputId": "611e2eb0-b2dd-49b6-fabe-f07393b22983"
      },
      "source": [
        "model_name = 'bert-base-uncased'\r\n",
        "# Load the BERT tokenizer.\r\n",
        "print(f'Loading {model_name} tokenizer...')\r\n",
        "tokenizer = AutoTokenizer.from_pretrained(model_name, do_lower_case=True)"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Loading bert-base-uncased tokenizer...\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "fa9fc5d63de741b79b5a1e551a33f8ec",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=433.0, style=ProgressStyle(description_…"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "b0162a1dc13e4a03b9abdc95d3821e5b",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=231508.0, style=ProgressStyle(descripti…"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "b211934af71b4c9f96bc335f15c89a3f",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=466062.0, style=ProgressStyle(descripti…"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OVs3Jb3Vb89i"
      },
      "source": [
        "## Smart Batching of the training data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cWNt4eOLofmF"
      },
      "source": [
        "lengths = []\r\n",
        "for text in train_sentences:\r\n",
        "  lengths.append(len(text))       "
      ],
      "execution_count": 33,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 427
        },
        "id": "hbdGRA3Bo02O",
        "outputId": "cc0af162-1a8d-4e2a-f864-07c10a917a1c"
      },
      "source": [
        "\r\n",
        "# Use plot styling from seaborn.\r\n",
        "sns.set(style='darkgrid')\r\n",
        "\r\n",
        "# Increase the plot size and font size.\r\n",
        "sns.set(font_scale=1.5)\r\n",
        "plt.rcParams[\"figure.figsize\"] = (12,6)\r\n",
        "\r\n",
        "plt.scatter(range(0, len(lengths)), lengths, marker=\"|\")\r\n",
        "\r\n",
        "plt.xlabel('Sample Number')\r\n",
        "plt.ylabel('Sequence Length')\r\n",
        "plt.title('Samples BEFORE Tokenizing')\r\n",
        "\r\n",
        "plt.show()"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9rMhLNhPcZjU"
      },
      "source": [
        "### Tokenizing the sequences"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t7JsLi3EdEje"
      },
      "source": [
        "max_input_length = 400"
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "J968o8RJcUEd",
        "outputId": "3d2bfb85-c311-4c07-8780-74e006f41460"
      },
      "source": [
        "full_input_ids = []\r\n",
        "labels = []\r\n",
        "\r\n",
        "# Tokenize all training examples\r\n",
        "print('Tokenizing {:,} training samples...'.format(len(train_sentences)))\r\n",
        "\r\n",
        "# Choose an interval on which to print progress updates.\r\n",
        "update_interval = good_update_interval(total_iters=len(train_sentences), num_desired_updates=10)\r\n",
        "\r\n",
        "# For each training example...\r\n",
        "for text in train_sentences:\r\n",
        "    \r\n",
        "    # Report progress.\r\n",
        "    if ((len(full_input_ids) % update_interval) == 0):\r\n",
        "        print('  Tokenized {:,} samples.'.format(len(full_input_ids)))\r\n",
        "\r\n",
        "    # Tokenize the sentence.\r\n",
        "    input_ids = tokenizer.encode(text=text,           \r\n",
        "                                 add_special_tokens=True, \r\n",
        "                                 max_length=max_input_length,  \r\n",
        "                                 truncation=True,     \r\n",
        "                                 padding=False)       \r\n",
        "                                 \r\n",
        "    # Add the tokenized result to our list.\r\n",
        "    full_input_ids.append(input_ids)\r\n",
        "    \r\n",
        "print('DONE.')\r\n",
        "print('{:>10,} samples'.format(len(full_input_ids)))"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Tokenizing 22,762 training samples...\n",
            "  Tokenized 0 samples.\n",
            "  Tokenized 2,000 samples.\n",
            "  Tokenized 4,000 samples.\n",
            "  Tokenized 6,000 samples.\n",
            "  Tokenized 8,000 samples.\n",
            "  Tokenized 10,000 samples.\n",
            "  Tokenized 12,000 samples.\n",
            "  Tokenized 14,000 samples.\n",
            "  Tokenized 16,000 samples.\n",
            "  Tokenized 18,000 samples.\n",
            "  Tokenized 20,000 samples.\n",
            "  Tokenized 22,000 samples.\n",
            "DONE.\n",
            "    22,762 samples\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rZzhO2tZeXur"
      },
      "source": [
        "# Get all of the lengths.\r\n",
        "unsorted_lengths = [len(x) for x in full_input_ids]"
      ],
      "execution_count": 37,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 427
        },
        "id": "x7c_mYuVed_H",
        "outputId": "6bcb7be9-1bb2-4433-82a7-5c6af064d2a7"
      },
      "source": [
        "\r\n",
        "# Use plot styling from seaborn.\r\n",
        "sns.set(style='darkgrid')\r\n",
        "\r\n",
        "# Increase the plot size and font size.\r\n",
        "sns.set(font_scale=1.5)\r\n",
        "plt.rcParams[\"figure.figsize\"] = (12,6)\r\n",
        "\r\n",
        "plt.scatter(range(0, len(unsorted_lengths)), unsorted_lengths, marker=\"|\")\r\n",
        "\r\n",
        "plt.xlabel('Sample Number')\r\n",
        "plt.ylabel('Sequence Length')\r\n",
        "plt.title('Samples BEFORE Sorting')\r\n",
        "\r\n",
        "plt.show()"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7PaeohUVegmv"
      },
      "source": [
        "# Sort the two lists together by the length of the input sequence.\r\n",
        "train_samples = sorted(zip(full_input_ids, train_labels), key=lambda x: len(x[0]))"
      ],
      "execution_count": 39,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_9nyx-BtepE4",
        "outputId": "345ab8e7-0d92-43fe-b513-854561631408"
      },
      "source": [
        "train_samples[0:5]"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[([101, 4067, 2017, 102], 1),\n",
              " ([101, 2033, 2205, 102], 1),\n",
              " ([101, 3835, 4357, 102], 1),\n",
              " ([101, 3835, 4357, 102], 1),\n",
              " ([101, 2026, 7780, 102], 0)]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 40
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rBxJhEieetFh",
        "outputId": "393c0476-3f40-487c-bdea-f5893da3aee4"
      },
      "source": [
        "print('Shortest sample:', len(train_samples[0][0]))\r\n",
        "print('Longest sample:', len(train_samples[-1][0]))"
      ],
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Shortest sample: 4\n",
            "Longest sample: 249\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "P44bFOI-e7hW"
      },
      "source": [
        "# Get the new list of lengths after sorting.\r\n",
        "sorted_lengths = [len(s[0]) for s in train_samples]"
      ],
      "execution_count": 42,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 427
        },
        "id": "9P5ZX99le_G1",
        "outputId": "e389f75a-aa6d-468d-b8a5-6cad19d8b0e3"
      },
      "source": [
        "\r\n",
        "plt.plot(range(0, len(sorted_lengths)), sorted_lengths)\r\n",
        "\r\n",
        "plt.xlabel('Sample Number')\r\n",
        "plt.ylabel('Sequence Length')\r\n",
        "plt.title('Samples after Sorting')\r\n",
        "\r\n",
        "plt.show()"
      ],
      "execution_count": 43,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Eh1z0xugfF2t"
      },
      "source": [
        "### Random Batch Selections"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gPrYkFMcgm5H",
        "outputId": "65b9b187-4ac7-47cb-d412-73bd7590d57c"
      },
      "source": [
        "train_samples[0]"
      ],
      "execution_count": 44,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "([101, 4067, 2017, 102], 1)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 44
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "isMxwgezfAvO"
      },
      "source": [
        "batch_size = 16"
      ],
      "execution_count": 45,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "s0tjD8pIfJ0S",
        "outputId": "26790d98-b057-4a34-e87b-abd2f9f77446"
      },
      "source": [
        "# List of batches that we'll construct.\r\n",
        "batch_ordered_sentences = []\r\n",
        "batch_ordered_labels = []\r\n",
        "\r\n",
        "print('Creating training batches of size {:}'.format(batch_size))\r\n",
        "\r\n",
        "# Loop over all of the input samples...    \r\n",
        "while len(train_samples) > 0:\r\n",
        "    \r\n",
        "    # Report progress.\r\n",
        "    if ((len(batch_ordered_sentences) % 100) == 0):\r\n",
        "        print('  Selected {:,} batches.'.format(len(batch_ordered_sentences)))\r\n",
        "\r\n",
        "    # `to_take` is our actual batch size. It will be `batch_size` until \r\n",
        "    # we get to the last batch, which may be smaller. \r\n",
        "    to_take = min(batch_size, len(train_samples))\r\n",
        "\r\n",
        "    # Pick a random index in the list of remaining samples to start\r\n",
        "    # our batch at.\r\n",
        "    select = random.randint(0, len(train_samples) - to_take)\r\n",
        "\r\n",
        "    # Select a contiguous batch of samples starting at `select`.\r\n",
        "    batch = train_samples[select:(select + to_take)]\r\n",
        "\r\n",
        "    # Each sample is a tuple--split them apart to create a separate list of \r\n",
        "    # sequences and a list of labels for this batch.\r\n",
        "    batch_ordered_sentences.append([s[0] for s in batch])\r\n",
        "    batch_ordered_labels.append([s[1] for s in batch])\r\n",
        "\r\n",
        "    # Remove these samples from the list.\r\n",
        "    del train_samples[select:select + to_take]\r\n",
        "\r\n",
        "print('\\n  DONE - {:,} batches.'.format(len(batch_ordered_sentences)))"
      ],
      "execution_count": 46,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Creating training batches of size 16\n",
            "  Selected 0 batches.\n",
            "  Selected 100 batches.\n",
            "  Selected 200 batches.\n",
            "  Selected 300 batches.\n",
            "  Selected 400 batches.\n",
            "  Selected 500 batches.\n",
            "  Selected 600 batches.\n",
            "  Selected 700 batches.\n",
            "  Selected 800 batches.\n",
            "  Selected 900 batches.\n",
            "  Selected 1,000 batches.\n",
            "  Selected 1,100 batches.\n",
            "  Selected 1,200 batches.\n",
            "  Selected 1,300 batches.\n",
            "  Selected 1,400 batches.\n",
            "\n",
            "  DONE - 1,423 batches.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rf91Dd3LhBaP",
        "outputId": "b50cdcf0-becc-4764-cc60-c3138bafc8db"
      },
      "source": [
        "print(batch_ordered_sentences[0])"
      ],
      "execution_count": 47,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[101, 1045, 2215, 2689, 1012, 17012, 1012, 2182, 1005, 1055, 2054, 1045, 1005, 2222, 2079, 1012, 1045, 1005, 2222, 24405, 2026, 2126, 2083, 2025, 4911, 4277, 1050, 6595, 2185, 2013, 4390, 1050, 2065, 10047, 2428, 1037, 2158, 1997, 2026, 2616, 5665, 2393, 2554, 1050, 2025, 2707, 8769, 102], [101, 1030, 7632, 1053, 13777, 3723, 7987, 27225, 1057, 2156, 2008, 2016, 2404, 2041, 8866, 1998, 1057, 19776, 2009, 2011, 3038, 2821, 10453, 1997, 2111, 5223, 2014, 2061, 2016, 1005, 1055, 4688, 1998, 18411, 2860, 2054, 3350, 2079, 1057, 2031, 2000, 2360, 2008, 2016, 1005, 1055, 4688, 102], [101, 2064, 1045, 2022, 7481, 1029, 8909, 3654, 2309, 1042, 2054, 2500, 2228, 2055, 2033, 1998, 2572, 2074, 4083, 2008, 2045, 4995, 1005, 1056, 2116, 2308, 1999, 1996, 3330, 2492, 1012, 2009, 2987, 1005, 1056, 2428, 4390, 2033, 2028, 2978, 1012, 2026, 3125, 2003, 2000, 2817, 2009, 102], [101, 2111, 2040, 2360, 2035, 3268, 3043, 2024, 2725, 2009, 2004, 1037, 3433, 2000, 2304, 3268, 3043, 2138, 2027, 2024, 2502, 27428, 1998, 2024, 2046, 2373, 1012, 2027, 2024, 2635, 2304, 3268, 3043, 2041, 1997, 6123, 1012, 2304, 3268, 3043, 2003, 2025, 1037, 3049, 2114, 2317, 3268, 102], [101, 2027, 2074, 2812, 1012, 1012, 2304, 3268, 3043, 2205, 1012, 1012, 2138, 2058, 2246, 2632, 4140, 1999, 2627, 1012, 1012, 2012, 2335, 1012, 1012, 2008, 2015, 2035, 2027, 2812, 1012, 1012, 4241, 23644, 23644, 2232, 1012, 1997, 2607, 3071, 4282, 2035, 3268, 3043, 2130, 4176, 1012, 102], [101, 9945, 1997, 4495, 1998, 9560, 2024, 2048, 3584, 2477, 1012, 2308, 1999, 1996, 2225, 2031, 2296, 4495, 2000, 2131, 2019, 2495, 1998, 2022, 2054, 2027, 2215, 21461, 2015, 999, 999, 2061, 2644, 10768, 5244, 24759, 8113, 2075, 4897, 2039, 1996, 15114, 1998, 2131, 2000, 2147, 1012, 102], [101, 1030, 13433, 13433, 2087, 2111, 2024, 2025, 21591, 1012, 2115, 3038, 2087, 2111, 2024, 1012, 26539, 2378, 2035, 3268, 3043, 1012, 2025, 2074, 2304, 1012, 22064, 2063, 2024, 1037, 5020, 1012, 2017, 2024, 2025, 2682, 4496, 2917, 1012, 2175, 3275, 1998, 4982, 2039, 1012, 19852, 1012, 102], [101, 2304, 3268, 3043, 2138, 2023, 2003, 2073, 1996, 3689, 2318, 2144, 1996, 3689, 2318, 3071, 2215, 2000, 2022, 2443, 1999, 3005, 2166, 3043, 3071, 1005, 1055, 2166, 2515, 3043, 1999, 4507, 2021, 2023, 2003, 1037, 2304, 3277, 8046, 2667, 2000, 2735, 2009, 2317, 2030, 2505, 2842, 102], [101, 1045, 5993, 2021, 2008, 1005, 1055, 2069, 2070, 1998, 1045, 2228, 2087, 2812, 2000, 2224, 1999, 1996, 1996, 3574, 1998, 3793, 1045, 4541, 2009, 1005, 1055, 2025, 2467, 1996, 2028, 2017, 4541, 1998, 2008, 2064, 2022, 1037, 24216, 2043, 2111, 7868, 2008, 2008, 2001, 2037, 11174, 102], [101, 1045, 2001, 4427, 2004, 1037, 2308, 2000, 2903, 1999, 2870, 1998, 2022, 1037, 2844, 2981, 2308, 1012, 1045, 2001, 2992, 2007, 2053, 15846, 1998, 2000, 2022, 25192, 1012, 2016, 5749, 100, 18168, 2290, 3053, 6639, 2000, 2014, 7615, 2012, 1996, 2203, 1520, 2196, 2507, 2039, 1521, 102], [101, 2000, 2122, 8801, 1998, 1996, 1996, 2502, 5085, 1998, 11578, 2008, 2067, 2068, 2053, 3268, 3043, 1012, 2009, 2003, 6389, 1996, 2304, 2679, 1999, 5517, 2008, 2069, 2304, 3268, 3043, 2024, 2437, 14398, 4788, 1012, 1999, 1996, 2159, 1997, 2643, 1996, 2269, 2035, 3268, 3043, 1012, 102], [101, 1030, 6187, 3540, 23689, 2483, 7842, 27457, 2213, 2527, 11867, 25667, 2072, 18168, 2290, 2339, 2079, 1057, 2031, 2000, 2467, 2360, 2049, 10303, 2000, 6449, 2000, 2115, 12051, 2043, 24471, 1996, 13534, 2028, 2004, 13337, 2024, 4288, 1999, 10085, 13013, 2111, 2074, 2000, 2265, 2020, 5020, 102], [101, 1006, 2304, 3268, 3043, 1027, 2304, 3268, 2036, 3043, 1998, 2057, 5807, 1005, 1056, 8568, 2008, 2755, 1012, 1007, 1006, 2035, 3268, 3043, 1027, 2045, 2003, 2053, 13661, 5325, 1999, 1996, 2304, 2451, 2013, 2610, 24083, 1998, 1996, 3563, 3291, 2003, 2025, 2008, 7882, 1012, 1007, 102], [101, 1030, 3389, 6243, 1045, 1005, 2310, 3191, 2115, 7615, 8529, 13876, 12129, 2335, 1998, 1045, 1005, 1049, 3904, 1996, 7968, 2099, 1012, 1045, 2442, 16519, 2009, 1005, 1055, 2074, 1037, 7215, 1997, 2616, 2013, 1996, 9714, 24241, 1005, 1055, 16105, 8663, 1012, 2190, 4756, 1997, 2035, 102], [101, 2831, 2831, 2831, 1998, 2027, 2123, 1005, 1056, 2079, 2505, 2000, 10938, 2304, 4279, 1012, 2085, 2027, 2180, 1005, 1056, 3499, 8398, 2079, 2009, 2593, 1012, 23961, 24158, 2001, 1018, 29100, 2015, 3283, 1998, 2027, 2024, 2067, 2153, 1012, 2304, 3268, 2069, 3043, 2076, 3864, 1012, 102], [101, 1030, 5199, 1055, 1999, 3088, 2009, 6433, 1999, 2162, 7950, 12917, 2087, 3060, 3032, 2031, 2069, 2042, 2109, 2005, 8864, 2011, 13481, 2057, 2097, 3613, 2000, 6033, 16215, 22573, 11342, 3251, 1057, 2066, 2009, 2030, 2025, 1998, 2129, 2515, 9846, 2028, 6231, 10083, 1996, 2878, 2103, 102]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qmkQK_gVgiTu",
        "outputId": "206d26d3-1bf0-458b-a4f4-dcf86ed0f180"
      },
      "source": [
        "batch_ordered_labels[0]"
      ],
      "execution_count": 48,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[1, 1, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 48
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xZ0TFA8Mfn6m"
      },
      "source": [
        "### Padding"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "c9rdswcpfTip"
      },
      "source": [
        "py_inputs = []\r\n",
        "py_attn_masks = []\r\n",
        "py_labels = []\r\n",
        "\r\n",
        "# For each batch...\r\n",
        "for (batch_inputs, batch_labels) in zip(batch_ordered_sentences, batch_ordered_labels):\r\n",
        "\r\n",
        "    # New version of the batch, this time with padded sequences and now with\r\n",
        "    # attention masks defined.\r\n",
        "    batch_padded_inputs = []\r\n",
        "    batch_attn_masks = []\r\n",
        "    \r\n",
        "    # First, find the longest sample in the batch. \r\n",
        "    # Note that the sequences do currently include the special tokens!\r\n",
        "    max_size = max([len(sen) for sen in batch_inputs])\r\n",
        "\r\n",
        "    #print('Max size:', max_size)\r\n",
        "\r\n",
        "    # For each input in this batch...\r\n",
        "    for sen in batch_inputs:\r\n",
        "        \r\n",
        "        # How many pad tokens do we need to add?\r\n",
        "        num_pads = max_size - len(sen)\r\n",
        "\r\n",
        "        # Add `num_pads` padding tokens to the end of the sequence.\r\n",
        "        padded_input = sen + [tokenizer.pad_token_id]*num_pads\r\n",
        "\r\n",
        "        # Define the attention mask--it's just a `1` for every real token\r\n",
        "        # and a `0` for every padding token.\r\n",
        "        attn_mask = [1] * len(sen) + [0] * num_pads\r\n",
        "\r\n",
        "        # Add the padded results to the batch.\r\n",
        "        batch_padded_inputs.append(padded_input)\r\n",
        "        batch_attn_masks.append(attn_mask)\r\n",
        "\r\n",
        "    # Our batch has been padded, so we need to save this updated batch.\r\n",
        "    # We also need the inputs to be PyTorch tensors, so we'll do that here.\r\n",
        "    py_inputs.append(torch.tensor(batch_padded_inputs))\r\n",
        "    py_attn_masks.append(torch.tensor(batch_attn_masks))\r\n",
        "    py_labels.append(torch.tensor(batch_labels))\r\n"
      ],
      "execution_count": 49,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "roS5lLWdX-5J"
      },
      "source": [
        "### Check the number of token reductions because of smart batching"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wlh7cteVfroV",
        "outputId": "f8e6f6a9-455f-4b8e-b478-21f5ca2a1b44"
      },
      "source": [
        "# Get the new list of lengths after sorting.\r\n",
        "\r\n",
        "padded_lengths = []\r\n",
        "\r\n",
        "# For each batch...\r\n",
        "for batch in py_inputs:\r\n",
        "    \r\n",
        "    # For each sample...\r\n",
        "    for s in batch:\r\n",
        "    \r\n",
        "        # Record its length.\r\n",
        "        padded_lengths.append(len(s))\r\n",
        "\r\n",
        "# Sum up the lengths to the get the total number of tokens after smart batching.\r\n",
        "smart_token_count = np.sum(padded_lengths)\r\n",
        "\r\n",
        "# To get the total number of tokens in the dataset using fixed padding, it's\r\n",
        "# as simple as the number of samples times our `max_len` parameter (that we\r\n",
        "# would pad everything to).\r\n",
        "fixed_token_count = len(train_sentences) * max_input_length\r\n",
        "\r\n",
        "# Calculate the percentage reduction.\r\n",
        "prcnt_reduced = (fixed_token_count - smart_token_count) / float(fixed_token_count) \r\n",
        "\r\n",
        "print('Total tokens:')\r\n",
        "print('   Fixed Padding: {:,}'.format(fixed_token_count))\r\n",
        "print('  Smart Batching: {:,}  ({:.1%} less)'.format(smart_token_count, prcnt_reduced))\r\n"
      ],
      "execution_count": 50,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Total tokens:\n",
            "   Fixed Padding: 9,104,800\n",
            "  Smart Batching: 527,638  (94.2% less)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8wgdhsI-YIAL"
      },
      "source": [
        "## Load the model configuration from the transformers library using AutoConfig"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9er7EHvdfudf",
        "outputId": "4427c1e3-77a3-49b8-f07f-c4a30475c903"
      },
      "source": [
        "# Load the Config object, with an output configured for classification.\r\n",
        "config = AutoConfig.from_pretrained(pretrained_model_name_or_path=model_name,\r\n",
        "                                    num_labels=3)\r\n",
        "\r\n",
        "print('Config type:', str(type(config)), '\\n')\r\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Config type: <class 'transformers.models.bert.configuration_bert.BertConfig'> \n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 170,
          "referenced_widgets": [
            "18c4291e51e140d9a85ea1fe8d50ecc3",
            "c3454a2687be43a1aa053816d60ba180",
            "300a8f8d8b814fb798a020dffa57d955",
            "f4912fb547764567ab9728e0b22eb3db",
            "28906f1fd0f140f88a93c0ecf74749e9",
            "7a9432d5bd8b4739879b5eafe10614c9",
            "8b5800c073bc4d379ccb18a369ab3482",
            "05863112e5b746bd80b5b037c7417610"
          ]
        },
        "id": "bWvc9Ikv4HtD",
        "outputId": "6cff544f-8a3b-429c-b0c6-0e604b37a19a"
      },
      "source": [
        "model = AutoModelForSequenceClassification.from_pretrained(pretrained_model_name_or_path=model_name,config = config)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "18c4291e51e140d9a85ea1fe8d50ecc3",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=440473133.0, style=ProgressStyle(descri…"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "Some weights of the model checkpoint at bert-base-uncased were not used when initializing BertForSequenceClassification: ['cls.predictions.bias', 'cls.predictions.transform.dense.weight', 'cls.predictions.transform.dense.bias', 'cls.predictions.decoder.weight', 'cls.seq_relationship.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.predictions.transform.LayerNorm.bias']\n",
            "- This IS expected if you are initializing BertForSequenceClassification 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 BertForSequenceClassification 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 BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased and are newly initialized: ['classifier.weight', 'classifier.bias']\n",
            "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Xouf5M2B55ei",
        "outputId": "5e133a91-4b30-4757-e8e0-29e381bc9116"
      },
      "source": [
        "model.cuda()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "BertForSequenceClassification(\n",
              "  (bert): BertModel(\n",
              "    (embeddings): BertEmbeddings(\n",
              "      (word_embeddings): Embedding(30522, 768, padding_idx=0)\n",
              "      (position_embeddings): Embedding(512, 768)\n",
              "      (token_type_embeddings): Embedding(2, 768)\n",
              "      (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "      (dropout): Dropout(p=0.1, inplace=False)\n",
              "    )\n",
              "    (encoder): BertEncoder(\n",
              "      (layer): ModuleList(\n",
              "        (0): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (1): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (2): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (3): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (4): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (5): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (6): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (7): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (8): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (9): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (10): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (11): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "      )\n",
              "    )\n",
              "    (pooler): BertPooler(\n",
              "      (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "      (activation): Tanh()\n",
              "    )\n",
              "  )\n",
              "  (dropout): Dropout(p=0.1, inplace=False)\n",
              "  (classifier): Linear(in_features=768, out_features=3, bias=True)\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 55
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SXJHBVlxYR-l"
      },
      "source": [
        "## Load the model from the transformers library using AutoModel"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "n8J4r6KyhveR",
        "outputId": "2de247e1-f6c2-4e1b-fe34-92edcd98efe5"
      },
      "source": [
        "# # Load the pre-trained model for classification, passing in the `config` from\r\n",
        "# # above.\r\n",
        "# bert = AutoModel.from_pretrained(\r\n",
        "#     pretrained_model_name_or_path=model_name,\r\n",
        "#     config=config)\r\n",
        "\r\n",
        "# print('\\nModel type:', str(type(bert)))\r\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "Model type: <class 'transformers.models.bert.modeling_bert.BertModel'>\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RcFMgf5EuPAF"
      },
      "source": [
        "# # freeze all the parameters\r\n",
        "# for param in bert.parameters():\r\n",
        "#     param.requires_grad = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NS_-0rBDYeY5"
      },
      "source": [
        "## Creating a custom BERT model "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bmGtZwykJ39i"
      },
      "source": [
        "\r\n",
        "# class BERT_Model(nn.Module):\r\n",
        "#     ## defining the constructor for the class\r\n",
        "#     def __init__(self, bert,num_labels):\r\n",
        "#       ## calling the super class constructor\r\n",
        "#       super(BERT_Model, self).__init__()\r\n",
        "#       ## adding the bert layer to the model\r\n",
        "#       self.bert = bert        \r\n",
        "#       # relu activation function\r\n",
        "#       self.relu =  nn.ReLU()\r\n",
        "#       # adding a dense layer to our custom model\r\n",
        "#       self.fc1 = nn.Linear(768,512)\r\n",
        "#       # adding another dense layer to our custom model ,i.e., the Output layer\r\n",
        "#       self.fc2 = nn.Linear(512,num_labels)\r\n",
        "#       # adding a softmax  activation function for our custom model's output layer\r\n",
        "#       self.softmax = nn.LogSoftmax(dim=1)\r\n",
        "\r\n",
        "#     #defining the forward pass\r\n",
        "#     def forward(self, input_id, mask):\r\n",
        "#       #pass the inputs to the model  \r\n",
        "#       outputs = self.bert(input_id,mask)\r\n",
        "  \r\n",
        "#       last_hidden_state = outputs.last_hidden_state       ## last hidden state from the model\r\n",
        "#       pooler_output = outputs.pooler_output               ## pooler output from the model\r\n",
        "#       ## adding a fully connected layer to the BERT model\r\n",
        "#       x = self.fc1(pooler_output)\r\n",
        "#       ## applying relu activation function\r\n",
        "#       x = self.relu(x)\r\n",
        "#       # the final output layer\r\n",
        "#       x = self.fc2(x)\r\n",
        "#       # apply softmax activation to our output layer\r\n",
        "#       x = self.softmax(x)\r\n",
        "\r\n",
        "#       return x"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jvOs0G5sh02S"
      },
      "source": [
        "\r\n",
        "\r\n",
        "# print('\\nLoading model ...')\r\n",
        "# # pass the pre-trained BERT to our define architecture\r\n",
        "# model = BERT_Model(bert,num_labels=3)\r\n",
        "\r\n",
        "# model.cuda()"
      ],
      "execution_count": 51,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LjQSUqCWJTbI"
      },
      "source": [
        "## Custom Loss function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iOHVl1RRJSUB"
      },
      "source": [
        "# # convert class weights to tensor\r\n",
        "# weights= torch.tensor(class_wts,dtype=torch.float)\r\n",
        "# weights = weights.to(device)\r\n",
        "\r\n",
        "# # loss function\r\n",
        "# cross_entropy  = nn.NLLLoss(weight=weights) \r\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "evYV0yrQ2u_b"
      },
      "source": [
        "## Loading Optimizer"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8tAijXBEh7t8"
      },
      "source": [
        "# Note: AdamW is a class from the huggingface library (as opposed to pytorch) \r\n",
        "# I believe the 'W' stands for 'Weight Decay fix\"\r\n",
        "optimizer = AdamW(model.parameters(),\r\n",
        "                  lr = 2e-5, # This is the value Michael used.\r\n",
        "                  eps = 1e-8 # args.adam_epsilon  - default is 1e-8.\r\n",
        "                )\r\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "s9n3_5jI2ywo"
      },
      "source": [
        "## Loading lr scheduler"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "agidP6D6iBFI"
      },
      "source": [
        "# Number of training epochs. I chose to train for 1 simply because the training\r\n",
        "# time is long. More epochs may improve the model's accuracy.\r\n",
        "epochs = 4\r\n",
        "\r\n",
        "# Total number of training steps is [number of batches] x [number of epochs]. \r\n",
        "# Note that it's the number of *batches*, not *samples*!\r\n",
        "total_steps = len(py_inputs) * epochs\r\n",
        "\r\n",
        "# Create the learning rate scheduler.\r\n",
        "scheduler = get_linear_schedule_with_warmup(optimizer, \r\n",
        "                                            num_warmup_steps = 0, # Default value in run_glue.py\r\n",
        "                                            num_training_steps = total_steps)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MDDp05Cm245e"
      },
      "source": [
        "## Training Loop"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "H4ERR8fBiEJn",
        "outputId": "8a6ae39b-592c-43c0-fcbc-2857a0781145"
      },
      "source": [
        "# We'll store a number of quantities such as training and validation loss, \r\n",
        "# validation accuracy, and timings.\r\n",
        "training_stats = {\r\n",
        "    'epoch':[],\r\n",
        "    'train_loss':[],\r\n",
        "    'Training Time':[],\r\n",
        "    'val_loss':[],\r\n",
        "    'Validation Time':[],\r\n",
        "    'train_acc':[],\r\n",
        "    'val_acc':[]\r\n",
        "}\r\n",
        "\r\n",
        "# Update every `update_interval` batches.\r\n",
        "update_interval = good_update_interval(total_iters=len(py_inputs), num_desired_updates=10)\r\n",
        "\r\n",
        "# Measure the total training time for the whole run.\r\n",
        "total_t0 = time.time()\r\n",
        "\r\n",
        "# For each epoch...\r\n",
        "for epoch_i in range(0, epochs):\r\n",
        "\r\n",
        "    predictions = []\r\n",
        "    true_labels = []\r\n",
        "    \r\n",
        "    # ========================================\r\n",
        "    #               Training\r\n",
        "    # ========================================\r\n",
        "    \r\n",
        "    # Perform one full pass over the training set.\r\n",
        "\r\n",
        "    print(\"\")\r\n",
        "    print('======== Epoch {:} / {:} ========'.format(epoch_i + 1, epochs))\r\n",
        "    \r\n",
        "    # At the start of each epoch (except for the first) we need to re-randomize\r\n",
        "    # our training data.\r\n",
        "    if epoch_i > 0:\r\n",
        "        # Use our `make_smart_batches` function (from 6.1.) to re-shuffle the \r\n",
        "        # dataset into new batches.\r\n",
        "        (py_inputs, py_attn_masks, py_labels) = make_smart_batches(train_sentences, train_labels, batch_size,tokenizer,max_input_length)\r\n",
        "    \r\n",
        "    print('Training on {:,} batches...'.format(len(py_inputs)))\r\n",
        "\r\n",
        "    # Measure how long the training epoch takes.\r\n",
        "    t0 = time.time()\r\n",
        "\r\n",
        "    # Reset the total loss for this epoch.\r\n",
        "    total_train_loss = 0\r\n",
        "\r\n",
        "    # Put the model into training mode. Don't be mislead--the call to \r\n",
        "    # `train` just changes the *mode*, it doesn't *perform* the training.\r\n",
        "    # `dropout` and `batchnorm` layers behave differently during training\r\n",
        "    # vs. test (source: https://stackoverflow.com/questions/51433378/what-does-model-train-do-in-pytorch)\r\n",
        "    model.train()\r\n",
        "\r\n",
        "    # For each batch of training data...\r\n",
        "    for step in range(0, len(py_inputs)):\r\n",
        "\r\n",
        "        # Progress update every, e.g., 100 batches.\r\n",
        "        if step % update_interval == 0 and not step == 0:\r\n",
        "            # Calculate elapsed time in minutes.\r\n",
        "            elapsed = format_time(time.time() - t0)\r\n",
        "            \r\n",
        "            # Calculate the time remaining based on our progress.\r\n",
        "            steps_per_sec = (time.time() - t0) / step\r\n",
        "            remaining_sec = steps_per_sec * (len(py_inputs) - step)\r\n",
        "            remaining = format_time(remaining_sec)\r\n",
        "\r\n",
        "            # Report progress.\r\n",
        "            print('  Batch {:>7,}  of  {:>7,}.    Elapsed: {:}.  Remaining: {:}'.format(step, len(py_inputs), elapsed, remaining))\r\n",
        "\r\n",
        "        # Copy the current training batch to the GPU using the `to` method.\r\n",
        "        b_input_ids = py_inputs[step].to(device)\r\n",
        "        b_input_mask = py_attn_masks[step].to(device)\r\n",
        "        b_labels = py_labels[step].to(device)\r\n",
        "\r\n",
        "        # Always clear any previously calculated gradients before performing a\r\n",
        "        # backward pass.\r\n",
        "        model.zero_grad()        \r\n",
        "\r\n",
        "        # Perform a forward pass (evaluate the model on this training batch).\r\n",
        "        # The call returns the loss (because we provided labels) and the \r\n",
        "        # \"logits\"--the model outputs prior to activation.\r\n",
        "        output = model(b_input_ids, \r\n",
        "                             token_type_ids=None, \r\n",
        "                             attention_mask=b_input_mask, \r\n",
        "                             labels=b_labels)\r\n",
        "\r\n",
        "        # Accumulate the training loss over all of the batches so that we can\r\n",
        "        # calculate the average loss at the end. `loss` is a Tensor containing a\r\n",
        "        # single value; the `.item()` function just returns the Python value \r\n",
        "        # from the tensor.\r\n",
        "        loss = output.loss \r\n",
        "        logits = output.logits\r\n",
        "\r\n",
        "        # Move logits and labels to CPU\r\n",
        "        logits = logits.detach().cpu().numpy()\r\n",
        "        label_ids = b_labels.to('cpu').numpy()\r\n",
        "      \r\n",
        "        # Store predictions and true labels\r\n",
        "        predictions.append(logits)\r\n",
        "        true_labels.append(label_ids)\r\n",
        "        \r\n",
        "        # Accumulate the training loss over all of the batches so that we can\r\n",
        "        # calculate the average loss at the end. `loss` is a Tensor containing a\r\n",
        "        # single value; the `.item()` function just returns the Python value \r\n",
        "        # from the tensor.\r\n",
        "        total_train_loss += loss.item()\r\n",
        "\r\n",
        "        # Perform a backward pass to calculate the gradients.\r\n",
        "        loss.backward()\r\n",
        "\r\n",
        "        # Clip the norm of the gradients to 1.0.\r\n",
        "        # This is to help prevent the \"exploding gradients\" problem.\r\n",
        "        torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)\r\n",
        "\r\n",
        "        # Update parameters and take a step using the computed gradient.\r\n",
        "        # The optimizer dictates the \"update rule\"--how the parameters are\r\n",
        "        # modified based on their gradients, the learning rate, etc.\r\n",
        "        optimizer.step()\r\n",
        "\r\n",
        "        # Update the learning rate.\r\n",
        "        scheduler.step()\r\n",
        "\r\n",
        "    # Calculate the average loss over all of the batches.\r\n",
        "    avg_train_loss = total_train_loss / len(py_inputs)     \r\n",
        "       \r\n",
        "    training_accuracy = check_accuracy(predictions,true_labels)       \r\n",
        "    \r\n",
        "    # Measure how long this epoch took.\r\n",
        "    training_time = format_time(time.time() - t0)\r\n",
        "\r\n",
        "    print(\"\")\r\n",
        "    print(\"  Average training loss: {0:.2f}\".format(avg_train_loss))\r\n",
        "    print(\"  Training Accuracy: {0:.2f}\".format(training_accuracy))\r\n",
        "    print(\"  Training epoch took: {:}\".format(training_time))\r\n",
        "        \r\n",
        "    (py_inputs, py_attn_masks, py_labels) = make_smart_batches(val_sentences, val_labels, batch_size ,tokenizer,max_input_length)\r\n",
        "    val_loss,val_accuracy,validation_time = eval_model(model,py_inputs, py_attn_masks, py_labels)  \r\n",
        "    # Record all statistics from this epoch.\r\n",
        "\r\n",
        "    print(\"\")\r\n",
        "    print(\"  Average validation loss: {0:.2f}\".format(val_loss))\r\n",
        "    print(\"  Validation Accuracy: {0:.2f}\".format(val_accuracy))\r\n",
        "    print(\"  Validation epoch took: {:}\".format(validation_time))\r\n",
        "    \r\n",
        "    \r\n",
        "    training_stats['epoch'].append(epoch_i + 1)\r\n",
        "    training_stats['train_loss'].append(avg_train_loss)\r\n",
        "    training_stats['Training Time'].append(training_time)\r\n",
        "    training_stats['val_loss'].append(val_loss)\r\n",
        "    training_stats['Validation Time'].append(validation_time)\r\n",
        "    training_stats['train_acc'].append(training_accuracy)\r\n",
        "    training_stats['val_acc'].append(val_accuracy)\r\n",
        "\r\n",
        "print(\"\")\r\n",
        "print(\"Training complete!\")\r\n",
        "\r\n",
        "print(\"Total training took {:} (h:mm:ss)\".format(format_time(time.time()-total_t0)))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "======== Epoch 1 / 4 ========\n",
            "Training on 1,423 batches...\n",
            "  Batch     100  of    1,423.    Elapsed: 0:00:11.  Remaining: 0:02:31\n",
            "  Batch     200  of    1,423.    Elapsed: 0:00:23.  Remaining: 0:02:21\n",
            "  Batch     300  of    1,423.    Elapsed: 0:00:34.  Remaining: 0:02:09\n",
            "  Batch     400  of    1,423.    Elapsed: 0:00:44.  Remaining: 0:01:53\n",
            "  Batch     500  of    1,423.    Elapsed: 0:00:55.  Remaining: 0:01:41\n",
            "  Batch     600  of    1,423.    Elapsed: 0:01:06.  Remaining: 0:01:30\n",
            "  Batch     700  of    1,423.    Elapsed: 0:01:16.  Remaining: 0:01:19\n",
            "  Batch     800  of    1,423.    Elapsed: 0:01:27.  Remaining: 0:01:08\n",
            "  Batch     900  of    1,423.    Elapsed: 0:01:37.  Remaining: 0:00:57\n",
            "  Batch   1,000  of    1,423.    Elapsed: 0:01:48.  Remaining: 0:00:46\n",
            "  Batch   1,100  of    1,423.    Elapsed: 0:01:58.  Remaining: 0:00:35\n",
            "  Batch   1,200  of    1,423.    Elapsed: 0:02:08.  Remaining: 0:00:24\n",
            "  Batch   1,300  of    1,423.    Elapsed: 0:02:18.  Remaining: 0:00:13\n",
            "  Batch   1,400  of    1,423.    Elapsed: 0:02:29.  Remaining: 0:00:02\n",
            "\n",
            "  Average training loss: 0.23\n",
            "  Training Accuracy: 0.92\n",
            "  Training epoch took: 0:02:32\n",
            "Creating Smart Batches from 2,843 examples with batch size 16...\n",
            "\n",
            "Tokenizing 2,843 samples...\n",
            "  Tokenized 0 samples.\n",
            "  Tokenized 300 samples.\n",
            "  Tokenized 600 samples.\n",
            "  Tokenized 900 samples.\n",
            "  Tokenized 1,200 samples.\n",
            "  Tokenized 1,500 samples.\n",
            "  Tokenized 1,800 samples.\n",
            "  Tokenized 2,100 samples.\n",
            "  Tokenized 2,400 samples.\n",
            "  Tokenized 2,700 samples.\n",
            "DONE.\n",
            "     2,843 samples\n",
            "\n",
            "     2,843 samples after sorting\n",
            "\n",
            "Creating batches of size 16...\n",
            "\n",
            "  DONE - Selected 178 batches.\n",
            "\n",
            "Padding out sequences within each batch...\n",
            "  DONE.\n",
            "Doing validation on 178 sentences...\n",
            "  Batch      20  of      178.    Elapsed: 0:00:00.  Remaining: 0:00:04\n",
            "  Batch      40  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:03\n",
            "  Batch      60  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:02\n",
            "  Batch      80  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:02\n",
            "  Batch     100  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:02\n",
            "  Batch     120  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:01\n",
            "  Batch     140  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:01\n",
            "  Batch     160  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:00\n",
            "\n",
            "  Average validation loss: 0.19\n",
            "  Validation Accuracy: 0.93\n",
            "  Validation epoch took: 0:00:04\n",
            "\n",
            "======== Epoch 2 / 4 ========\n",
            "Creating Smart Batches from 22,762 examples with batch size 16...\n",
            "\n",
            "Tokenizing 22,762 samples...\n",
            "  Tokenized 0 samples.\n",
            "  Tokenized 2,000 samples.\n",
            "  Tokenized 4,000 samples.\n",
            "  Tokenized 6,000 samples.\n",
            "  Tokenized 8,000 samples.\n",
            "  Tokenized 10,000 samples.\n",
            "  Tokenized 12,000 samples.\n",
            "  Tokenized 14,000 samples.\n",
            "  Tokenized 16,000 samples.\n",
            "  Tokenized 18,000 samples.\n",
            "  Tokenized 20,000 samples.\n",
            "  Tokenized 22,000 samples.\n",
            "DONE.\n",
            "    22,762 samples\n",
            "\n",
            "    22,762 samples after sorting\n",
            "\n",
            "Creating batches of size 16...\n",
            "\n",
            "  DONE - Selected 1,423 batches.\n",
            "\n",
            "Padding out sequences within each batch...\n",
            "  DONE.\n",
            "Training on 1,423 batches...\n",
            "  Batch     100  of    1,423.    Elapsed: 0:00:10.  Remaining: 0:02:18\n",
            "  Batch     200  of    1,423.    Elapsed: 0:00:21.  Remaining: 0:02:11\n",
            "  Batch     300  of    1,423.    Elapsed: 0:00:32.  Remaining: 0:01:59\n",
            "  Batch     400  of    1,423.    Elapsed: 0:00:42.  Remaining: 0:01:48\n",
            "  Batch     500  of    1,423.    Elapsed: 0:00:52.  Remaining: 0:01:37\n",
            "  Batch     600  of    1,423.    Elapsed: 0:01:03.  Remaining: 0:01:27\n",
            "  Batch     700  of    1,423.    Elapsed: 0:01:13.  Remaining: 0:01:16\n",
            "  Batch     800  of    1,423.    Elapsed: 0:01:24.  Remaining: 0:01:05\n",
            "  Batch     900  of    1,423.    Elapsed: 0:01:35.  Remaining: 0:00:55\n",
            "  Batch   1,000  of    1,423.    Elapsed: 0:01:46.  Remaining: 0:00:45\n",
            "  Batch   1,100  of    1,423.    Elapsed: 0:01:56.  Remaining: 0:00:34\n",
            "  Batch   1,200  of    1,423.    Elapsed: 0:02:07.  Remaining: 0:00:24\n",
            "  Batch   1,300  of    1,423.    Elapsed: 0:02:18.  Remaining: 0:00:13\n",
            "  Batch   1,400  of    1,423.    Elapsed: 0:02:28.  Remaining: 0:00:02\n",
            "\n",
            "  Average training loss: 0.15\n",
            "  Training Accuracy: 0.95\n",
            "  Training epoch took: 0:02:31\n",
            "Creating Smart Batches from 2,843 examples with batch size 16...\n",
            "\n",
            "Tokenizing 2,843 samples...\n",
            "  Tokenized 0 samples.\n",
            "  Tokenized 300 samples.\n",
            "  Tokenized 600 samples.\n",
            "  Tokenized 900 samples.\n",
            "  Tokenized 1,200 samples.\n",
            "  Tokenized 1,500 samples.\n",
            "  Tokenized 1,800 samples.\n",
            "  Tokenized 2,100 samples.\n",
            "  Tokenized 2,400 samples.\n",
            "  Tokenized 2,700 samples.\n",
            "DONE.\n",
            "     2,843 samples\n",
            "\n",
            "     2,843 samples after sorting\n",
            "\n",
            "Creating batches of size 16...\n",
            "\n",
            "  DONE - Selected 178 batches.\n",
            "\n",
            "Padding out sequences within each batch...\n",
            "  DONE.\n",
            "Doing validation on 178 sentences...\n",
            "  Batch      20  of      178.    Elapsed: 0:00:00.  Remaining: 0:00:03\n",
            "  Batch      40  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:03\n",
            "  Batch      60  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:02\n",
            "  Batch      80  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:02\n",
            "  Batch     100  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:02\n",
            "  Batch     120  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:01\n",
            "  Batch     140  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:01\n",
            "  Batch     160  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:00\n",
            "\n",
            "  Average validation loss: 0.25\n",
            "  Validation Accuracy: 0.93\n",
            "  Validation epoch took: 0:00:04\n",
            "\n",
            "======== Epoch 3 / 4 ========\n",
            "Creating Smart Batches from 22,762 examples with batch size 16...\n",
            "\n",
            "Tokenizing 22,762 samples...\n",
            "  Tokenized 0 samples.\n",
            "  Tokenized 2,000 samples.\n",
            "  Tokenized 4,000 samples.\n",
            "  Tokenized 6,000 samples.\n",
            "  Tokenized 8,000 samples.\n",
            "  Tokenized 10,000 samples.\n",
            "  Tokenized 12,000 samples.\n",
            "  Tokenized 14,000 samples.\n",
            "  Tokenized 16,000 samples.\n",
            "  Tokenized 18,000 samples.\n",
            "  Tokenized 20,000 samples.\n",
            "  Tokenized 22,000 samples.\n",
            "DONE.\n",
            "    22,762 samples\n",
            "\n",
            "    22,762 samples after sorting\n",
            "\n",
            "Creating batches of size 16...\n",
            "\n",
            "  DONE - Selected 1,423 batches.\n",
            "\n",
            "Padding out sequences within each batch...\n",
            "  DONE.\n",
            "Training on 1,423 batches...\n",
            "  Batch     100  of    1,423.    Elapsed: 0:00:10.  Remaining: 0:02:15\n",
            "  Batch     200  of    1,423.    Elapsed: 0:00:21.  Remaining: 0:02:06\n",
            "  Batch     300  of    1,423.    Elapsed: 0:00:31.  Remaining: 0:01:55\n",
            "  Batch     400  of    1,423.    Elapsed: 0:00:42.  Remaining: 0:01:47\n",
            "  Batch     500  of    1,423.    Elapsed: 0:00:52.  Remaining: 0:01:36\n",
            "  Batch     600  of    1,423.    Elapsed: 0:01:02.  Remaining: 0:01:25\n",
            "  Batch     700  of    1,423.    Elapsed: 0:01:12.  Remaining: 0:01:14\n",
            "  Batch     800  of    1,423.    Elapsed: 0:01:22.  Remaining: 0:01:04\n",
            "  Batch     900  of    1,423.    Elapsed: 0:01:32.  Remaining: 0:00:53\n",
            "  Batch   1,000  of    1,423.    Elapsed: 0:01:42.  Remaining: 0:00:43\n",
            "  Batch   1,100  of    1,423.    Elapsed: 0:01:53.  Remaining: 0:00:33\n",
            "  Batch   1,200  of    1,423.    Elapsed: 0:02:03.  Remaining: 0:00:23\n",
            "  Batch   1,300  of    1,423.    Elapsed: 0:02:13.  Remaining: 0:00:13\n",
            "  Batch   1,400  of    1,423.    Elapsed: 0:02:23.  Remaining: 0:00:02\n",
            "\n",
            "  Average training loss: 0.09\n",
            "  Training Accuracy: 0.97\n",
            "  Training epoch took: 0:02:27\n",
            "Creating Smart Batches from 2,843 examples with batch size 16...\n",
            "\n",
            "Tokenizing 2,843 samples...\n",
            "  Tokenized 0 samples.\n",
            "  Tokenized 300 samples.\n",
            "  Tokenized 600 samples.\n",
            "  Tokenized 900 samples.\n",
            "  Tokenized 1,200 samples.\n",
            "  Tokenized 1,500 samples.\n",
            "  Tokenized 1,800 samples.\n",
            "  Tokenized 2,100 samples.\n",
            "  Tokenized 2,400 samples.\n",
            "  Tokenized 2,700 samples.\n",
            "DONE.\n",
            "     2,843 samples\n",
            "\n",
            "     2,843 samples after sorting\n",
            "\n",
            "Creating batches of size 16...\n",
            "\n",
            "  DONE - Selected 178 batches.\n",
            "\n",
            "Padding out sequences within each batch...\n",
            "  DONE.\n",
            "Doing validation on 178 sentences...\n",
            "  Batch      20  of      178.    Elapsed: 0:00:00.  Remaining: 0:00:03\n",
            "  Batch      40  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:02\n",
            "  Batch      60  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:02\n",
            "  Batch      80  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:02\n",
            "  Batch     100  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:01\n",
            "  Batch     120  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:01\n",
            "  Batch     140  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:01\n",
            "  Batch     160  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:00\n",
            "\n",
            "  Average validation loss: 0.30\n",
            "  Validation Accuracy: 0.93\n",
            "  Validation epoch took: 0:00:04\n",
            "\n",
            "======== Epoch 4 / 4 ========\n",
            "Creating Smart Batches from 22,762 examples with batch size 16...\n",
            "\n",
            "Tokenizing 22,762 samples...\n",
            "  Tokenized 0 samples.\n",
            "  Tokenized 2,000 samples.\n",
            "  Tokenized 4,000 samples.\n",
            "  Tokenized 6,000 samples.\n",
            "  Tokenized 8,000 samples.\n",
            "  Tokenized 10,000 samples.\n",
            "  Tokenized 12,000 samples.\n",
            "  Tokenized 14,000 samples.\n",
            "  Tokenized 16,000 samples.\n",
            "  Tokenized 18,000 samples.\n",
            "  Tokenized 20,000 samples.\n",
            "  Tokenized 22,000 samples.\n",
            "DONE.\n",
            "    22,762 samples\n",
            "\n",
            "    22,762 samples after sorting\n",
            "\n",
            "Creating batches of size 16...\n",
            "\n",
            "  DONE - Selected 1,423 batches.\n",
            "\n",
            "Padding out sequences within each batch...\n",
            "  DONE.\n",
            "Training on 1,423 batches...\n",
            "  Batch     100  of    1,423.    Elapsed: 0:00:11.  Remaining: 0:02:21\n",
            "  Batch     200  of    1,423.    Elapsed: 0:00:21.  Remaining: 0:02:07\n",
            "  Batch     300  of    1,423.    Elapsed: 0:00:31.  Remaining: 0:01:55\n",
            "  Batch     400  of    1,423.    Elapsed: 0:00:41.  Remaining: 0:01:44\n",
            "  Batch     500  of    1,423.    Elapsed: 0:00:51.  Remaining: 0:01:34\n",
            "  Batch     600  of    1,423.    Elapsed: 0:01:01.  Remaining: 0:01:23\n",
            "  Batch     700  of    1,423.    Elapsed: 0:01:11.  Remaining: 0:01:13\n",
            "  Batch     800  of    1,423.    Elapsed: 0:01:21.  Remaining: 0:01:03\n",
            "  Batch     900  of    1,423.    Elapsed: 0:01:31.  Remaining: 0:00:53\n",
            "  Batch   1,000  of    1,423.    Elapsed: 0:01:41.  Remaining: 0:00:43\n",
            "  Batch   1,100  of    1,423.    Elapsed: 0:01:51.  Remaining: 0:00:33\n",
            "  Batch   1,200  of    1,423.    Elapsed: 0:02:01.  Remaining: 0:00:23\n",
            "  Batch   1,300  of    1,423.    Elapsed: 0:02:12.  Remaining: 0:00:12\n",
            "  Batch   1,400  of    1,423.    Elapsed: 0:02:22.  Remaining: 0:00:02\n",
            "\n",
            "  Average training loss: 0.05\n",
            "  Training Accuracy: 0.99\n",
            "  Training epoch took: 0:02:25\n",
            "Creating Smart Batches from 2,843 examples with batch size 16...\n",
            "\n",
            "Tokenizing 2,843 samples...\n",
            "  Tokenized 0 samples.\n",
            "  Tokenized 300 samples.\n",
            "  Tokenized 600 samples.\n",
            "  Tokenized 900 samples.\n",
            "  Tokenized 1,200 samples.\n",
            "  Tokenized 1,500 samples.\n",
            "  Tokenized 1,800 samples.\n",
            "  Tokenized 2,100 samples.\n",
            "  Tokenized 2,400 samples.\n",
            "  Tokenized 2,700 samples.\n",
            "DONE.\n",
            "     2,843 samples\n",
            "\n",
            "     2,843 samples after sorting\n",
            "\n",
            "Creating batches of size 16...\n",
            "\n",
            "  DONE - Selected 178 batches.\n",
            "\n",
            "Padding out sequences within each batch...\n",
            "  DONE.\n",
            "Doing validation on 178 sentences...\n",
            "  Batch      20  of      178.    Elapsed: 0:00:00.  Remaining: 0:00:03\n",
            "  Batch      40  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:03\n",
            "  Batch      60  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:02\n",
            "  Batch      80  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:02\n",
            "  Batch     100  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:02\n",
            "  Batch     120  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:01\n",
            "  Batch     140  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:01\n",
            "  Batch     160  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:00\n",
            "\n",
            "  Average validation loss: 0.35\n",
            "  Validation Accuracy: 0.92\n",
            "  Validation epoch took: 0:00:04\n",
            "\n",
            "Training complete!\n",
            "Total training took 0:10:21 (h:mm:ss)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 450
        },
        "id": "GOmua_S1eXNM",
        "outputId": "c49ad932-af4d-4afa-a7d6-143bf8d228d0"
      },
      "source": [
        "plot_training_history(training_stats)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1296x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3c-6bOIxVyrM"
      },
      "source": [
        "## Evaluating Performance Over Training Set"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2QmKPNAeWHkZ",
        "outputId": "ddebd2b6-bb5e-4702-e596-56a129938e18"
      },
      "source": [
        "(py_inputs, py_attn_masks, py_labels) = make_smart_batches(train_sentences, train_labels, batch_size ,tokenizer,max_input_length)\r\n",
        "\r\n",
        "y_pred , y_true = get_predictions(py_inputs, py_attn_masks, py_labels)\r\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Creating Smart Batches from 22,762 examples with batch size 16...\n",
            "\n",
            "Tokenizing 22,762 samples...\n",
            "  Tokenized 0 samples.\n",
            "  Tokenized 2,000 samples.\n",
            "  Tokenized 4,000 samples.\n",
            "  Tokenized 6,000 samples.\n",
            "  Tokenized 8,000 samples.\n",
            "  Tokenized 10,000 samples.\n",
            "  Tokenized 12,000 samples.\n",
            "  Tokenized 14,000 samples.\n",
            "  Tokenized 16,000 samples.\n",
            "  Tokenized 18,000 samples.\n",
            "  Tokenized 20,000 samples.\n",
            "  Tokenized 22,000 samples.\n",
            "DONE.\n",
            "    22,762 samples\n",
            "\n",
            "    22,762 samples after sorting\n",
            "\n",
            "Creating batches of size 16...\n",
            "\n",
            "  DONE - Selected 1,423 batches.\n",
            "\n",
            "Padding out sequences within each batch...\n",
            "  DONE.\n",
            "Predicting labels for 1,423 test batches...\n",
            "  Batch     100  of    1,423.    Elapsed: 0:00:02.  Remaining: 0:00:28\n",
            "  Batch     200  of    1,423.    Elapsed: 0:00:04.  Remaining: 0:00:25\n",
            "  Batch     300  of    1,423.    Elapsed: 0:00:06.  Remaining: 0:00:22\n",
            "  Batch     400  of    1,423.    Elapsed: 0:00:08.  Remaining: 0:00:20\n",
            "  Batch     500  of    1,423.    Elapsed: 0:00:10.  Remaining: 0:00:18\n",
            "  Batch     600  of    1,423.    Elapsed: 0:00:12.  Remaining: 0:00:16\n",
            "  Batch     700  of    1,423.    Elapsed: 0:00:14.  Remaining: 0:00:14\n",
            "  Batch     800  of    1,423.    Elapsed: 0:00:16.  Remaining: 0:00:12\n",
            "  Batch     900  of    1,423.    Elapsed: 0:00:18.  Remaining: 0:00:10\n",
            "  Batch   1,000  of    1,423.    Elapsed: 0:00:20.  Remaining: 0:00:08\n",
            "  Batch   1,100  of    1,423.    Elapsed: 0:00:22.  Remaining: 0:00:06\n",
            "  Batch   1,200  of    1,423.    Elapsed: 0:00:23.  Remaining: 0:00:04\n",
            "  Batch   1,300  of    1,423.    Elapsed: 0:00:25.  Remaining: 0:00:02\n",
            "  Batch   1,400  of    1,423.    Elapsed: 0:00:27.  Remaining: 0:00:00\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uXsVn0wngZhv",
        "outputId": "c3222e67-20c3-48e2-8e9d-dab003a9f86e"
      },
      "source": [
        "print(classification_report(y_true, y_pred, target_names=class_names))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "                 precision    recall  f1-score   support\n",
            "\n",
            "    Hope_speech       0.96      0.95      0.96      1962\n",
            "Non-hope-speech       0.99      1.00      1.00     20778\n",
            "    not-English       1.00      0.18      0.31        22\n",
            "\n",
            "       accuracy                           0.99     22762\n",
            "      macro avg       0.99      0.71      0.75     22762\n",
            "   weighted avg       0.99      0.99      0.99     22762\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 476
        },
        "id": "guClUdRhYPql",
        "outputId": "af149f48-c63a-476c-8b39-ba87b5042ef9"
      },
      "source": [
        "cm = confusion_matrix(y_true, y_pred)\r\n",
        "show_confusion_matrix(cm, class_names)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VM2N-X1tV3il"
      },
      "source": [
        "## Evaluating Performance Over Validation Set"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BnadNvtxWIYw",
        "outputId": "15d63cd7-9ba4-4b21-a2ba-30270c9d1f65"
      },
      "source": [
        "(py_inputs, py_attn_masks, py_labels) = make_smart_batches(val_sentences, val_labels, batch_size ,tokenizer,max_input_length)\r\n",
        "\r\n",
        "y_pred , y_true = get_predictions(py_inputs, py_attn_masks, py_labels)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Creating Smart Batches from 2,843 examples with batch size 16...\n",
            "\n",
            "Tokenizing 2,843 samples...\n",
            "  Tokenized 0 samples.\n",
            "  Tokenized 300 samples.\n",
            "  Tokenized 600 samples.\n",
            "  Tokenized 900 samples.\n",
            "  Tokenized 1,200 samples.\n",
            "  Tokenized 1,500 samples.\n",
            "  Tokenized 1,800 samples.\n",
            "  Tokenized 2,100 samples.\n",
            "  Tokenized 2,400 samples.\n",
            "  Tokenized 2,700 samples.\n",
            "DONE.\n",
            "     2,843 samples\n",
            "\n",
            "     2,843 samples after sorting\n",
            "\n",
            "Creating batches of size 16...\n",
            "\n",
            "  DONE - Selected 178 batches.\n",
            "\n",
            "Padding out sequences within each batch...\n",
            "  DONE.\n",
            "Predicting labels for 178 test batches...\n",
            "  Batch      20  of      178.    Elapsed: 0:00:00.  Remaining: 0:00:04\n",
            "  Batch      40  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:03\n",
            "  Batch      60  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:02\n",
            "  Batch      80  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:02\n",
            "  Batch     100  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:02\n",
            "  Batch     120  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:01\n",
            "  Batch     140  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:01\n",
            "  Batch     160  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:00\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wAjlATBPgh7c",
        "outputId": "0ab76a6c-45fb-4170-da7f-5a581b1a4b36"
      },
      "source": [
        "print(classification_report(y_true, y_pred, target_names=class_names))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "                 precision    recall  f1-score   support\n",
            "\n",
            "    Hope_speech       0.62      0.57      0.59       272\n",
            "Non-hope-speech       0.95      0.96      0.96      2569\n",
            "    not-English       0.00      0.00      0.00         2\n",
            "\n",
            "       accuracy                           0.92      2843\n",
            "      macro avg       0.52      0.51      0.52      2843\n",
            "   weighted avg       0.92      0.92      0.92      2843\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/sklearn/metrics/_classification.py:1272: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 483
        },
        "id": "7nELY57NYshD",
        "outputId": "fd4b898c-cf23-4148-c7f5-78ea01e0b860"
      },
      "source": [
        "cm = confusion_matrix(y_true, y_pred)\r\n",
        "show_confusion_matrix(cm, class_names)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J8vxEcoQSZng"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eV1nPYflSeo_"
      },
      "source": [
        "## Saving the model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hsm1xnwtSepE"
      },
      "source": [
        "torch.save(\r\n",
        "    model,\r\n",
        "    \"./drive/MyDrive/Codalab/HopeSpeech Detection For Equality,DiversityAnd Inclusion/Models/BERTForSequenceClassification_English_v2\"\r\n",
        "    )"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "a9M_91z1yj0F"
      },
      "source": [
        "## Loading the model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "42ObCUU1QcsJ"
      },
      "source": [
        "model = torch.load(\r\n",
        "    './drive/MyDrive/HopeSpeech Detection For Equality,DiversityAnd Inclusion/Models/BERTForSequenceClassification_English_v2',\r\n",
        "                   map_location=device\r\n",
        "                   )"
      ],
      "execution_count": 52,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sAaHpZdIzsnT",
        "outputId": "a9a59bc5-72ef-4547-ee59-067485be1162"
      },
      "source": [
        "model.cuda()"
      ],
      "execution_count": 53,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "BertForSequenceClassification(\n",
              "  (bert): BertModel(\n",
              "    (embeddings): BertEmbeddings(\n",
              "      (word_embeddings): Embedding(30522, 768, padding_idx=0)\n",
              "      (position_embeddings): Embedding(512, 768)\n",
              "      (token_type_embeddings): Embedding(2, 768)\n",
              "      (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "      (dropout): Dropout(p=0.1, inplace=False)\n",
              "    )\n",
              "    (encoder): BertEncoder(\n",
              "      (layer): ModuleList(\n",
              "        (0): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (1): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (2): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (3): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (4): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (5): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (6): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (7): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (8): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (9): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (10): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (11): BertLayer(\n",
              "          (attention): BertAttention(\n",
              "            (self): BertSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): BertSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): BertIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): BertOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "      )\n",
              "    )\n",
              "    (pooler): BertPooler(\n",
              "      (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "      (activation): Tanh()\n",
              "    )\n",
              "  )\n",
              "  (dropout): Dropout(p=0.1, inplace=False)\n",
              "  (classifier): Linear(in_features=768, out_features=3, bias=True)\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 53
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jAwCfk8gSepL"
      },
      "source": [
        "## Making Predictions on Test Set"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EHoA-D_xSepL"
      },
      "source": [
        "test_df_path = join(dataset_dir,'english_hope_test.csv')\r\n",
        "test_df = pd.read_csv(test_df_path)"
      ],
      "execution_count": 54,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 402
        },
        "id": "bqdeFPaqSepM",
        "outputId": "5639bd58-46e2-4e98-ad07-373258a8eae4"
      },
      "source": [
        "test_df"
      ],
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<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>text</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>What do you mean by the word sniped?</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>I love this video!! I’m bisexual and it’s just...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>ya the irony but then i don't want to come off...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>A PERSON'S CHARACTER MATTERS. PERIOD!!</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>@Blaster of Gasters</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2841</th>\n",
              "      <td>+Ashrenneemakeup I think it's all a deliberate...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2842</th>\n",
              "      <td>Sheriff David Clarke. This guy is amazing.</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2843</th>\n",
              "      <td>Abandorn Hope Situation</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2844</th>\n",
              "      <td>Sheriff Clarke you are a person of such strong...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2845</th>\n",
              "      <td>Sanders has no room to talk. If there's one pe...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>2846 rows × 1 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "                                                   text\n",
              "0                  What do you mean by the word sniped?\n",
              "1     I love this video!! I’m bisexual and it’s just...\n",
              "2     ya the irony but then i don't want to come off...\n",
              "3                A PERSON'S CHARACTER MATTERS. PERIOD!!\n",
              "4                                   @Blaster of Gasters\n",
              "...                                                 ...\n",
              "2841  +Ashrenneemakeup I think it's all a deliberate...\n",
              "2842         Sheriff David Clarke. This guy is amazing.\n",
              "2843                            Abandorn Hope Situation\n",
              "2844  Sheriff Clarke you are a person of such strong...\n",
              "2845  Sanders has no room to talk. If there's one pe...\n",
              "\n",
              "[2846 rows x 1 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 55
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7jLPpljYQcsQ"
      },
      "source": [
        "## Loading `test_sentences`"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QEVoAhgIQcsR"
      },
      "source": [
        "test_sentences = test_df['text'].values \r\n",
        "test_ids = test_df.index.values"
      ],
      "execution_count": 56,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "slovGuCRThW1",
        "outputId": "208ed450-338d-4d79-ea99-6628c89c5036"
      },
      "source": [
        "test_ids"
      ],
      "execution_count": 57,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([   0,    1,    2, ..., 2843, 2844, 2845])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 57
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gL7aCG2HTjv6",
        "outputId": "d8b382d3-30c8-4816-db1c-f474091a25cd"
      },
      "source": [
        "test_sentences"
      ],
      "execution_count": 58,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array(['What do you mean by the word sniped?',\n",
              "       'I love this video!! I’m bisexual and it’s just really nice to hear that someone has gone through similar things like me and that I’m not alone. My story is nothing like his and my heart goes out to him to slowly heal from that',\n",
              "       \"ya the irony but then i don't want to come off as a bigot myself\",\n",
              "       ..., 'Abandorn Hope Situation',\n",
              "       'Sheriff Clarke you are a person of such strong character and wisdom.nI want to see more of you as a public figure and much less of the celebs who I admire for their acting but not their biased views.',\n",
              "       \"Sanders has no room to talk. If there's one person who has white privilege....\"],\n",
              "      dtype=object)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 58
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "b9dgO1m3QcsU",
        "outputId": "cedd5dd6-e148-49a9-f4da-0e0077df751c"
      },
      "source": [
        "(py_inputs, py_attn_masks,py_ids) = make_smart_batches_on_test(test_sentences,test_ids, 16,tokenizer,max_input_length = 400)"
      ],
      "execution_count": 59,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Creating Smart Batches from 2,846 examples with batch size 16...\n",
            "\n",
            "Tokenizing 2,846 samples...\n",
            "  Tokenized 0 samples.\n",
            "  Tokenized 300 samples.\n",
            "  Tokenized 600 samples.\n",
            "  Tokenized 900 samples.\n",
            "  Tokenized 1,200 samples.\n",
            "  Tokenized 1,500 samples.\n",
            "  Tokenized 1,800 samples.\n",
            "  Tokenized 2,100 samples.\n",
            "  Tokenized 2,400 samples.\n",
            "  Tokenized 2,700 samples.\n",
            "DONE.\n",
            "     2,846 samples\n",
            "\n",
            "     2,846 samples after sorting\n",
            "\n",
            "Creating batches of size 16...\n",
            "\n",
            "  DONE - Selected 178 batches.\n",
            "\n",
            "Padding out sequences within each batch...\n",
            "  DONE.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vjsjqQpRQcsV"
      },
      "source": [
        "## Evaluating accuracy over test set"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rwkx-30AQcsV",
        "outputId": "825f5220-bdb1-49ec-b910-efc755c087a3"
      },
      "source": [
        "y_ids,y_preds  = get_predictions_test(py_inputs, py_attn_masks,py_ids)\r\n"
      ],
      "execution_count": 60,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Predicting labels for 178 test batches...\n",
            "  Batch      20  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:05\n",
            "  Batch      40  of      178.    Elapsed: 0:00:01.  Remaining: 0:00:04\n",
            "  Batch      60  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:03\n",
            "  Batch      80  of      178.    Elapsed: 0:00:02.  Remaining: 0:00:02\n",
            "  Batch     100  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:02\n",
            "  Batch     120  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:01\n",
            "  Batch     140  of      178.    Elapsed: 0:00:03.  Remaining: 0:00:01\n",
            "  Batch     160  of      178.    Elapsed: 0:00:04.  Remaining: 0:00:00\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pDM-cnDHQcsW",
        "outputId": "806f7b50-3dbc-4f3e-b49d-7e6c144bd076"
      },
      "source": [
        "print(y_preds)"
      ],
      "execution_count": 61,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[1 0 1 ... 1 1 1]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kWWWjz7eUPc7",
        "outputId": "949286ed-3899-4493-d679-31404dce5793"
      },
      "source": [
        "print(y_ids)"
      ],
      "execution_count": 62,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[1390 1396 1397 ... 2648   42  105]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "u2ybbsAoQcsY",
        "outputId": "8074a4df-f0df-4749-8302-f94b1308c935"
      },
      "source": [
        "le.inverse_transform(y_preds)"
      ],
      "execution_count": 63,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array(['Non_hope_speech', 'Hope_speech', 'Non_hope_speech', ...,\n",
              "       'Non_hope_speech', 'Non_hope_speech', 'Non_hope_speech'],\n",
              "      dtype=object)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 63
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 424
        },
        "id": "kSMb6sCLQcsZ",
        "outputId": "2148c64f-4ef4-420f-8757-a8ddd4ec6b30"
      },
      "source": [
        "sns.countplot(y =le.inverse_transform(y_preds))"
      ],
      "execution_count": 64,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7f5a8b124080>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 64
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4KXOXwH9U2MT",
        "outputId": "6e085a59-0f0f-42f9-f873-c5702ace8755"
      },
      "source": [
        "len(y_ids)"
      ],
      "execution_count": 65,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "2846"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 65
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9-fkBgOxU4bX",
        "outputId": "665adfd7-c523-42c4-cb90-276d91ae7bf5"
      },
      "source": [
        "len(le.inverse_transform(y_preds))"
      ],
      "execution_count": 66,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "2846"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 66
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sWPDqogJrjIZ"
      },
      "source": [
        "## Creating Submission DataFrame"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "o4Hb7RD1yHol"
      },
      "source": [
        "df_new = pd.DataFrame({\r\n",
        "    'id':y_ids,\r\n",
        "    'label':le.inverse_transform(y_preds)\r\n",
        "})"
      ],
      "execution_count": 67,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "twQAkVzcn4Lg"
      },
      "source": [
        "test_df['id'] = test_df.index"
      ],
      "execution_count": 68,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 402
        },
        "id": "qov36eWmqawv",
        "outputId": "df0b6f5a-00ca-4be2-f9b6-7eac4b5e3ba9"
      },
      "source": [
        "df_new"
      ],
      "execution_count": 69,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<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>label</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1390</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>1396</td>\n",
              "      <td>Hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>1397</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>1424</td>\n",
              "      <td>Hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>1447</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2841</th>\n",
              "      <td>1068</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2842</th>\n",
              "      <td>1684</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2843</th>\n",
              "      <td>2648</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2844</th>\n",
              "      <td>42</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2845</th>\n",
              "      <td>105</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>2846 rows × 2 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "        id            label\n",
              "0     1390  Non_hope_speech\n",
              "1     1396      Hope_speech\n",
              "2     1397  Non_hope_speech\n",
              "3     1424      Hope_speech\n",
              "4     1447  Non_hope_speech\n",
              "...    ...              ...\n",
              "2841  1068  Non_hope_speech\n",
              "2842  1684  Non_hope_speech\n",
              "2843  2648  Non_hope_speech\n",
              "2844    42  Non_hope_speech\n",
              "2845   105  Non_hope_speech\n",
              "\n",
              "[2846 rows x 2 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 69
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 402
        },
        "id": "z0CEApHmqdLr",
        "outputId": "8371f3cb-79ab-4a43-e0f1-b49256ea4552"
      },
      "source": [
        "test_df"
      ],
      "execution_count": 70,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<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>text</th>\n",
              "      <th>id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>What do you mean by the word sniped?</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>I love this video!! I’m bisexual and it’s just...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>ya the irony but then i don't want to come off...</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>A PERSON'S CHARACTER MATTERS. PERIOD!!</td>\n",
              "      <td>3</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>@Blaster of Gasters</td>\n",
              "      <td>4</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2841</th>\n",
              "      <td>+Ashrenneemakeup I think it's all a deliberate...</td>\n",
              "      <td>2841</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2842</th>\n",
              "      <td>Sheriff David Clarke. This guy is amazing.</td>\n",
              "      <td>2842</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2843</th>\n",
              "      <td>Abandorn Hope Situation</td>\n",
              "      <td>2843</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2844</th>\n",
              "      <td>Sheriff Clarke you are a person of such strong...</td>\n",
              "      <td>2844</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2845</th>\n",
              "      <td>Sanders has no room to talk. If there's one pe...</td>\n",
              "      <td>2845</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>2846 rows × 2 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "                                                   text    id\n",
              "0                  What do you mean by the word sniped?     0\n",
              "1     I love this video!! I’m bisexual and it’s just...     1\n",
              "2     ya the irony but then i don't want to come off...     2\n",
              "3                A PERSON'S CHARACTER MATTERS. PERIOD!!     3\n",
              "4                                   @Blaster of Gasters     4\n",
              "...                                                 ...   ...\n",
              "2841  +Ashrenneemakeup I think it's all a deliberate...  2841\n",
              "2842         Sheriff David Clarke. This guy is amazing.  2842\n",
              "2843                            Abandorn Hope Situation  2843\n",
              "2844  Sheriff Clarke you are a person of such strong...  2844\n",
              "2845  Sanders has no room to talk. If there's one pe...  2845\n",
              "\n",
              "[2846 rows x 2 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 70
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 402
        },
        "id": "vxReWWmxn7Hj",
        "outputId": "2d99b39c-45f5-4dfa-aa88-c12aa1c464d5"
      },
      "source": [
        "df_f = pd.merge(test_df,df_new,on = 'id')\r\n",
        "df_f"
      ],
      "execution_count": 71,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<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>text</th>\n",
              "      <th>id</th>\n",
              "      <th>label</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>What do you mean by the word sniped?</td>\n",
              "      <td>0</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>I love this video!! I’m bisexual and it’s just...</td>\n",
              "      <td>1</td>\n",
              "      <td>Hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>ya the irony but then i don't want to come off...</td>\n",
              "      <td>2</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>A PERSON'S CHARACTER MATTERS. PERIOD!!</td>\n",
              "      <td>3</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>@Blaster of Gasters</td>\n",
              "      <td>4</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2841</th>\n",
              "      <td>+Ashrenneemakeup I think it's all a deliberate...</td>\n",
              "      <td>2841</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2842</th>\n",
              "      <td>Sheriff David Clarke. This guy is amazing.</td>\n",
              "      <td>2842</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2843</th>\n",
              "      <td>Abandorn Hope Situation</td>\n",
              "      <td>2843</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2844</th>\n",
              "      <td>Sheriff Clarke you are a person of such strong...</td>\n",
              "      <td>2844</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2845</th>\n",
              "      <td>Sanders has no room to talk. If there's one pe...</td>\n",
              "      <td>2845</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>2846 rows × 3 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "                                                   text    id            label\n",
              "0                  What do you mean by the word sniped?     0  Non_hope_speech\n",
              "1     I love this video!! I’m bisexual and it’s just...     1      Hope_speech\n",
              "2     ya the irony but then i don't want to come off...     2  Non_hope_speech\n",
              "3                A PERSON'S CHARACTER MATTERS. PERIOD!!     3  Non_hope_speech\n",
              "4                                   @Blaster of Gasters     4  Non_hope_speech\n",
              "...                                                 ...   ...              ...\n",
              "2841  +Ashrenneemakeup I think it's all a deliberate...  2841  Non_hope_speech\n",
              "2842         Sheriff David Clarke. This guy is amazing.  2842  Non_hope_speech\n",
              "2843                            Abandorn Hope Situation  2843  Non_hope_speech\n",
              "2844  Sheriff Clarke you are a person of such strong...  2844  Non_hope_speech\n",
              "2845  Sanders has no room to talk. If there's one pe...  2845  Non_hope_speech\n",
              "\n",
              "[2846 rows x 3 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 71
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 402
        },
        "id": "1kuASymYn9aV",
        "outputId": "1af23bd2-b5a9-49e1-97d5-0404dde68b8a"
      },
      "source": [
        "df_f = df_f[['id','text','label']]\r\n",
        "df_f"
      ],
      "execution_count": 72,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<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>text</th>\n",
              "      <th>label</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0</td>\n",
              "      <td>What do you mean by the word sniped?</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>1</td>\n",
              "      <td>I love this video!! I’m bisexual and it’s just...</td>\n",
              "      <td>Hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>2</td>\n",
              "      <td>ya the irony but then i don't want to come off...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>3</td>\n",
              "      <td>A PERSON'S CHARACTER MATTERS. PERIOD!!</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>4</td>\n",
              "      <td>@Blaster of Gasters</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2841</th>\n",
              "      <td>2841</td>\n",
              "      <td>+Ashrenneemakeup I think it's all a deliberate...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2842</th>\n",
              "      <td>2842</td>\n",
              "      <td>Sheriff David Clarke. This guy is amazing.</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2843</th>\n",
              "      <td>2843</td>\n",
              "      <td>Abandorn Hope Situation</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2844</th>\n",
              "      <td>2844</td>\n",
              "      <td>Sheriff Clarke you are a person of such strong...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2845</th>\n",
              "      <td>2845</td>\n",
              "      <td>Sanders has no room to talk. If there's one pe...</td>\n",
              "      <td>Non_hope_speech</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>2846 rows × 3 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "        id                                               text            label\n",
              "0        0               What do you mean by the word sniped?  Non_hope_speech\n",
              "1        1  I love this video!! I’m bisexual and it’s just...      Hope_speech\n",
              "2        2  ya the irony but then i don't want to come off...  Non_hope_speech\n",
              "3        3             A PERSON'S CHARACTER MATTERS. PERIOD!!  Non_hope_speech\n",
              "4        4                                @Blaster of Gasters  Non_hope_speech\n",
              "...    ...                                                ...              ...\n",
              "2841  2841  +Ashrenneemakeup I think it's all a deliberate...  Non_hope_speech\n",
              "2842  2842         Sheriff David Clarke. This guy is amazing.  Non_hope_speech\n",
              "2843  2843                            Abandorn Hope Situation  Non_hope_speech\n",
              "2844  2844  Sheriff Clarke you are a person of such strong...  Non_hope_speech\n",
              "2845  2845  Sanders has no room to talk. If there's one pe...  Non_hope_speech\n",
              "\n",
              "[2846 rows x 3 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 72
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "p3FtBVTVroJi"
      },
      "source": [
        "## Saving Submission DataFrame"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SzoT0p9rqrO-"
      },
      "source": [
        "path = './drive/MyDrive/Codalab/HopeSpeech Detection For Equality,DiversityAnd Inclusion/SubmissionFiles'"
      ],
      "execution_count": 73,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LZmgZFM4qi7F"
      },
      "source": [
        "df_f.to_csv(join(path,'bert_base_uncased_english.csv'),index=False)"
      ],
      "execution_count": 74,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jW9g_qgMv7mq"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}