model_explainability.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "model_explainability.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "f04341e6e2d04b4e966c8d06f95294d5": {
          "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_a876f66896ce48628e444f89ad71032a",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_9ed0c581deba4683be859cc2e2248c65",
              "IPY_MODEL_770d02a44dc84b66a56b1023009b8dec"
            ]
          }
        },
        "a876f66896ce48628e444f89ad71032a": {
          "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
          }
        },
        "9ed0c581deba4683be859cc2e2248c65": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_61aac437cd4c438ea61bafb43103e4be",
            "_dom_classes": [],
            "description": "100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 68,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 68,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_c45d5dee6ab0454b83fbfb7109a241c3"
          }
        },
        "770d02a44dc84b66a56b1023009b8dec": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_f02059eac4484337ad882aea11f0e624",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 68/68 [00:27&lt;00:00,  2.45ba/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_af6062e190d24af1b13d19aaf884b7c9"
          }
        },
        "61aac437cd4c438ea61bafb43103e4be": {
          "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"
          }
        },
        "c45d5dee6ab0454b83fbfb7109a241c3": {
          "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
          }
        },
        "f02059eac4484337ad882aea11f0e624": {
          "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"
          }
        },
        "af6062e190d24af1b13d19aaf884b7c9": {
          "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
          }
        },
        "0b4ea732c4234b9dbdf264d10d3a3ba9": {
          "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_a5a40100d9764b15891f4e2d505788cb",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_0332405b8fa24c929fdc51a8878b6c95",
              "IPY_MODEL_da60f1fdd941498da00e30b988fd5bbc"
            ]
          }
        },
        "a5a40100d9764b15891f4e2d505788cb": {
          "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
          }
        },
        "0332405b8fa24c929fdc51a8878b6c95": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_d558c2f95cbf40b1a22aa810cb122234",
            "_dom_classes": [],
            "description": "Downloading: 100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 570,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 570,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_16b024cd789f4e81bd55e3fb41801a3f"
          }
        },
        "da60f1fdd941498da00e30b988fd5bbc": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_9821cd996c8643eca613b2dde09cdbf9",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 570/570 [00:02&lt;00:00, 256B/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_54fc95cb87ca4f3293edd9945ec4618d"
          }
        },
        "d558c2f95cbf40b1a22aa810cb122234": {
          "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"
          }
        },
        "16b024cd789f4e81bd55e3fb41801a3f": {
          "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
          }
        },
        "9821cd996c8643eca613b2dde09cdbf9": {
          "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"
          }
        },
        "54fc95cb87ca4f3293edd9945ec4618d": {
          "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
          }
        },
        "91116a7703ca424697cb132c7630031e": {
          "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_c31a73e844424565ab7db7f2ec92fd2b",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_9d2a83764a8642b0ae9b5b0cd992c502",
              "IPY_MODEL_79028584d4bd4917b9a80273655a6730"
            ]
          }
        },
        "c31a73e844424565ab7db7f2ec92fd2b": {
          "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
          }
        },
        "9d2a83764a8642b0ae9b5b0cd992c502": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_47ff1e11fab844f38aeb59b028ec234b",
            "_dom_classes": [],
            "description": "Downloading: 100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 213450,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 213450,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_8553878a4c2a490d8bbdd7fffe0d1c07"
          }
        },
        "79028584d4bd4917b9a80273655a6730": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_4b020cb3b50a4e2c84bc75177a194a4a",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 213k/213k [00:01&lt;00:00, 131kB/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_1bb46b2aaf2e4ffd941db22a50ecdb71"
          }
        },
        "47ff1e11fab844f38aeb59b028ec234b": {
          "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"
          }
        },
        "8553878a4c2a490d8bbdd7fffe0d1c07": {
          "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
          }
        },
        "4b020cb3b50a4e2c84bc75177a194a4a": {
          "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"
          }
        },
        "1bb46b2aaf2e4ffd941db22a50ecdb71": {
          "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
          }
        },
        "be626e6198364cefb10ea556a1c5db0f": {
          "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_3608fa382b0a431dadba14f02ec774b4",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_e855a9d8372247a093b14bd32bde89b1",
              "IPY_MODEL_da3ec3940567421b8ebe35b50b13f911"
            ]
          }
        },
        "3608fa382b0a431dadba14f02ec774b4": {
          "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
          }
        },
        "e855a9d8372247a093b14bd32bde89b1": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_1d7cbe8309864a1d91e1516b6ea4ee33",
            "_dom_classes": [],
            "description": "Downloading: 100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 435797,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 435797,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_14b9ee169c47401d9e86828bfaf64e2a"
          }
        },
        "da3ec3940567421b8ebe35b50b13f911": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_358142a7978547ff8fff2dee6fbc6166",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 436k/436k [00:00&lt;00:00, 461kB/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_9546f5656c6c40d3b333a1ac68c81494"
          }
        },
        "1d7cbe8309864a1d91e1516b6ea4ee33": {
          "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"
          }
        },
        "14b9ee169c47401d9e86828bfaf64e2a": {
          "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
          }
        },
        "358142a7978547ff8fff2dee6fbc6166": {
          "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"
          }
        },
        "9546f5656c6c40d3b333a1ac68c81494": {
          "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
          }
        },
        "408d09ef41a24bc2815d55735aad503e": {
          "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_03d5fe0b754d40cdbebe181cc885ed13",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_6e0a4cec14f246a08aa7a70fec5daac8",
              "IPY_MODEL_371f867193be4783a0a1db0b4a17ed7b"
            ]
          }
        },
        "03d5fe0b754d40cdbebe181cc885ed13": {
          "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
          }
        },
        "6e0a4cec14f246a08aa7a70fec5daac8": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_3807f31fbabb4a0b88e8e22156f18189",
            "_dom_classes": [],
            "description": "Downloading: 100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 29,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 29,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_209a9b321f924d45979fcbbd98a039d3"
          }
        },
        "371f867193be4783a0a1db0b4a17ed7b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_ef2eb51fabe342e3953608b93af5df52",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 29.0/29.0 [00:06&lt;00:00, 4.45B/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_9dd2664852b84689948529115fc9c160"
          }
        },
        "3807f31fbabb4a0b88e8e22156f18189": {
          "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"
          }
        },
        "209a9b321f924d45979fcbbd98a039d3": {
          "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
          }
        },
        "ef2eb51fabe342e3953608b93af5df52": {
          "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"
          }
        },
        "9dd2664852b84689948529115fc9c160": {
          "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
          }
        },
        "2dbb371423b3466e9becc54b30190d4a": {
          "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_785eb66e7b7441beb61c961b742e0239",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_20cb76b46ea740028180748328ad2a4e",
              "IPY_MODEL_6ffedf1bc38f4b4dbf1276a859830745"
            ]
          }
        },
        "785eb66e7b7441beb61c961b742e0239": {
          "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
          }
        },
        "20cb76b46ea740028180748328ad2a4e": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_76cfc6f1071046a09b04eec07acd52dd",
            "_dom_classes": [],
            "description": "100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 67349,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 67349,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_fd8ea05e5cb743eeaeec09a41a2a5ed0"
          }
        },
        "6ffedf1bc38f4b4dbf1276a859830745": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_fbe0da7b6bbb452eb3b0ca438a65da7b",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 67349/67349 [03:46&lt;00:00, 296.73ex/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_6698ba0edf484d1ba9b50185c300abe1"
          }
        },
        "76cfc6f1071046a09b04eec07acd52dd": {
          "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"
          }
        },
        "fd8ea05e5cb743eeaeec09a41a2a5ed0": {
          "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
          }
        },
        "fbe0da7b6bbb452eb3b0ca438a65da7b": {
          "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"
          }
        },
        "6698ba0edf484d1ba9b50185c300abe1": {
          "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
          }
        },
        "b4e49976611a4067929dd65940b62d07": {
          "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_b4b5991f1dca4a0ea48dea5dc86985ac",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_c93ad7720d8e4504a0ade349143676ce",
              "IPY_MODEL_e4148066e4264fd291c5c471088b62cb"
            ]
          }
        },
        "b4b5991f1dca4a0ea48dea5dc86985ac": {
          "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
          }
        },
        "c93ad7720d8e4504a0ade349143676ce": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_3b580f8cd0d8445a9dffb245a64964d1",
            "_dom_classes": [],
            "description": "100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 872,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 872,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_5b719123bfc34f6dad74c9e8e4dacf2e"
          }
        },
        "e4148066e4264fd291c5c471088b62cb": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_8013cbea9339492f92916e97f5a900f8",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 872/872 [03:32&lt;00:00,  4.11ex/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_519ac51e146b4416b062a6b1462e3129"
          }
        },
        "3b580f8cd0d8445a9dffb245a64964d1": {
          "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"
          }
        },
        "5b719123bfc34f6dad74c9e8e4dacf2e": {
          "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
          }
        },
        "8013cbea9339492f92916e97f5a900f8": {
          "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"
          }
        },
        "519ac51e146b4416b062a6b1462e3129": {
          "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
          }
        },
        "7708b8ecd0ed44a0a1335f116277f4ba": {
          "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_e1bf8aa2e62f4bf284193a5034ef2ca7",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_3ab350ec6a09481aad8e3b0b912703c4",
              "IPY_MODEL_9757a8eac4024ff3b1db99df6aeeb38f"
            ]
          }
        },
        "e1bf8aa2e62f4bf284193a5034ef2ca7": {
          "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
          }
        },
        "3ab350ec6a09481aad8e3b0b912703c4": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_3a4357ee6d9845a5922132108f9d2d9b",
            "_dom_classes": [],
            "description": "100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 1821,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 1821,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_c33c470b41cc46ddb23c7a8af6fd2ed7"
          }
        },
        "9757a8eac4024ff3b1db99df6aeeb38f": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_6d40a93298cd433f9ca02aa752c5cbc6",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 1821/1821 [03:31&lt;00:00,  8.60ex/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_4b1dc8a5fdf940d1ae03e2b5631ce4d7"
          }
        },
        "3a4357ee6d9845a5922132108f9d2d9b": {
          "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"
          }
        },
        "c33c470b41cc46ddb23c7a8af6fd2ed7": {
          "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
          }
        },
        "6d40a93298cd433f9ca02aa752c5cbc6": {
          "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"
          }
        },
        "4b1dc8a5fdf940d1ae03e2b5631ce4d7": {
          "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
          }
        },
        "1e6f9d8c22f24980b365046a91bce910": {
          "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_030beb0fee49412b80fd7ae4deaea85c",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_74c1cd41c17a4096b9620aafab815a88",
              "IPY_MODEL_75ee7bdf5211434c9e1f54dbe42453cb"
            ]
          }
        },
        "030beb0fee49412b80fd7ae4deaea85c": {
          "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
          }
        },
        "74c1cd41c17a4096b9620aafab815a88": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_277f281606ec4602997d1670c40ba923",
            "_dom_classes": [],
            "description": "Downloading: 100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 435779157,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 435779157,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_dd86296fa3f64e519d46e03b083a1807"
          }
        },
        "75ee7bdf5211434c9e1f54dbe42453cb": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_83cb7c3303ad4dfaa1e2429c60e9cedf",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 436M/436M [00:59&lt;00:00, 7.33MB/s]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_92d824d31a544b54a88d91a14228f707"
          }
        },
        "277f281606ec4602997d1670c40ba923": {
          "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"
          }
        },
        "dd86296fa3f64e519d46e03b083a1807": {
          "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
          }
        },
        "83cb7c3303ad4dfaa1e2429c60e9cedf": {
          "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"
          }
        },
        "92d824d31a544b54a88d91a14228f707": {
          "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
          }
        }
      }
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/TurkuNLP/Text_Mining_Course/blob/master/model_explainability.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aV3y7R77TnBF"
      },
      "source": [
        "# Model explainability example\n",
        "\n",
        "* This notebook is based off the BERT text classifier notebook in this same course (https://github.com/TurkuNLP/Text_Mining_Course/blob/master/train_bert_for_text_classification.ipynb)\n",
        "* Remember to watch the explain-ml tutorial https://explainml-tutorial.github.io/neurips20 this is not meant to be a self-contained notebook!\n",
        "* The new code related to model explainability starts half-way through, after the classifier has been trained\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3LPd4TaxwbMR"
      },
      "source": [
        "# Train BERT text classifier using the Transformers library\n",
        "\n",
        "This notebook shows how to train a simple text classifier by fine-tuning a pre-trained BERT model using the Hugging Face [Transformers](https://huggingface.co/transformers/) library.\n",
        "\n",
        "This notebook is based in part on the [Text classification on GLUE](https://colab.research.google.com/github/huggingface/notebooks/blob/master/examples/text_classification.ipynb) notebook.\n",
        "\n",
        "**NOTE**: it's recommended to run this using a runtime with a GPU. Select \"Runtime\" -> \"Change runtime type\" from the top menu in Colab and set \"Hardware accelerator\" to \"GPU\" when starting"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "u3uu9vMewmM6"
      },
      "source": [
        "## Install libraries\n",
        "\n",
        "First, we'll use [`pip`](https://pypi.org/project/pip/) to install two Python libraries that are used in this notebook: [`transformers`](https://huggingface.co/transformers/) and [`datasets`](https://huggingface.co/docs/datasets/)."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "C1XGi09l-KxP",
        "outputId": "5b9f78be-1fc2-4b61-d0e4-af382501bf83"
      },
      "source": [
        "!pip --quiet install transformers\n",
        "!pip --quiet install datasets"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\u001b[K     |████████████████████████████████| 2.1MB 7.9MB/s \n",
            "\u001b[K     |████████████████████████████████| 3.3MB 54.0MB/s \n",
            "\u001b[K     |████████████████████████████████| 901kB 51.8MB/s \n",
            "\u001b[K     |████████████████████████████████| 204kB 8.1MB/s \n",
            "\u001b[K     |████████████████████████████████| 245kB 13.9MB/s \n",
            "\u001b[K     |████████████████████████████████| 112kB 16.0MB/s \n",
            "\u001b[?25h"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ETfE2d5Ryisz"
      },
      "source": [
        "## Import libraries\n",
        "\n",
        "Next, let's import the classes and functions we'll be using (click links for documentation):\n",
        "\n",
        "* [AutoTokenizer](https://huggingface.co/transformers/model_doc/auto.html#transformers.AutoTokenizer): tokenizer class with support for the tokenization conventions of various pre-trained models\n",
        "* [AutoModelForSequenceClassification](https://huggingface.co/transformers/model_doc/auto.html#automodelforsequenceclassification): an [AutoModel](https://huggingface.co/transformers/model_doc/auto.html#transformers.AutoModel) class that can load various pre-trained models and supports fine-tuning for text (sequence) classification\n",
        "* [TrainingArguments](https://huggingface.co/transformers/main_classes/trainer.html#transformers.TrainingArguments): simple class for holding hyperparameters and other settings for model training\n",
        "* [Trainer](https://huggingface.co/transformers/main_classes/trainer.html): class supporting various forms of training transformer models\n",
        "* [load_dataset](https://huggingface.co/docs/datasets/package_reference/loading_methods.html#datasets.load_dataset): function for loading datasets e.g. from the [datasets](https://huggingface.co/datasets) collection"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "q21P5qTsyyzO"
      },
      "source": [
        "from transformers import AutoTokenizer\n",
        "from transformers import AutoModelForSequenceClassification\n",
        "from transformers import TrainingArguments\n",
        "from transformers import Trainer\n",
        "from datasets import load_dataset"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HOM98a-B1gT2"
      },
      "source": [
        "## Set model, dataset and hyperparameters\n",
        "\n",
        "Let's then set some global variables such as the name of the pre-trained model and the hyperparameters we'll use for fine-tuning it.\n",
        "\n",
        "* `MODEL_NAME`: the name of a pretrained model included in the [model repository](https://huggingface.co/models)\n",
        "* `DATASET`: the path and name of a dataset included in the [dataset repository](https://huggingface.co/datasets)\n",
        "* `LEARNING_RATE`, `BATCH_SIZE`, and `TRAIN_EPOCHS`: hyperparameters to use for fine-turning the model. (Try different values here!)\n",
        "\n",
        "Here, we'll use the [Stanford Sentiment Treebank](https://huggingface.co/datasets/sst) dataset as prepared for the [GLUE](https://huggingface.co/datasets/glue) collection. This is a binary text classification task where the objective is to determine if sentences express a positive or negative sentiment."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7ciPcqKNksJr"
      },
      "source": [
        "MODEL_NAME = 'bert-base-cased'\n",
        "DATASET = ('glue', 'sst2')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DB9k8bU_6ZFK"
      },
      "source": [
        "## Load dataset\n",
        "\n",
        "We'll first load the dataset with [`load_dataset`](https://huggingface.co/docs/datasets/package_reference/loading_methods.html#datasets.load_dataset) and determine the number of distinct labels in the (training) data."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pUO0grs56oDF",
        "outputId": "866c5cf8-9fc8-4a23-e635-7b8f99ecbb6e"
      },
      "source": [
        "dataset = load_dataset(*DATASET)\n",
        "num_labels = len(set(dataset['train']['label']))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Reusing dataset glue (/root/.cache/huggingface/datasets/glue/sst2/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad)\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nu17UF9D7EFB"
      },
      "source": [
        "The loaded dataset is a simple dictionary-like container for distinct `DataSet` objects for training, development (validation), and test data:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gVEvezq37PqH",
        "outputId": "f8f2e6fe-d4dc-4aa1-83ce-1a012fdda2c4"
      },
      "source": [
        "print(dataset)\n",
        "print(f'number of distinct labels: {num_labels}')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "DatasetDict({\n",
            "    train: Dataset({\n",
            "        features: ['sentence', 'label', 'idx'],\n",
            "        num_rows: 67349\n",
            "    })\n",
            "    validation: Dataset({\n",
            "        features: ['sentence', 'label', 'idx'],\n",
            "        num_rows: 872\n",
            "    })\n",
            "    test: Dataset({\n",
            "        features: ['sentence', 'label', 'idx'],\n",
            "        num_rows: 1821\n",
            "    })\n",
            "})\n",
            "number of distinct labels: 2\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6yVSGu_C8o8D"
      },
      "source": [
        "To reduce training time a bit, let's just take every 10th item in the `train` subset of the dataset."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 66,
          "referenced_widgets": [
            "f04341e6e2d04b4e966c8d06f95294d5",
            "a876f66896ce48628e444f89ad71032a",
            "9ed0c581deba4683be859cc2e2248c65",
            "770d02a44dc84b66a56b1023009b8dec",
            "61aac437cd4c438ea61bafb43103e4be",
            "c45d5dee6ab0454b83fbfb7109a241c3",
            "f02059eac4484337ad882aea11f0e624",
            "af6062e190d24af1b13d19aaf884b7c9"
          ]
        },
        "id": "K-zoQFxy8xq6",
        "outputId": "4b554e1a-ee87-493f-9332-b1bb96055781"
      },
      "source": [
        "# I leave this undone to get the best possible model later for the explainability part\n",
        "dataset['train'] = dataset['train'].filter(lambda example, idx: idx % 10 == 0, with_indices=True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "f04341e6e2d04b4e966c8d06f95294d5",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, max=68.0), HTML(value='')))"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Yf_R3aYC38bM"
      },
      "source": [
        "## Load tokenizer and tokenize data\n",
        "\n",
        "We'll then load an appropriate tokenizer for the pre-trained model we'll be using with [`AutoTokenizer.from_pretrained`](https://huggingface.co/transformers/model_doc/auto.html#transformers.AutoTokenizer.from_pretrained)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 213,
          "referenced_widgets": [
            "0b4ea732c4234b9dbdf264d10d3a3ba9",
            "a5a40100d9764b15891f4e2d505788cb",
            "0332405b8fa24c929fdc51a8878b6c95",
            "da60f1fdd941498da00e30b988fd5bbc",
            "d558c2f95cbf40b1a22aa810cb122234",
            "16b024cd789f4e81bd55e3fb41801a3f",
            "9821cd996c8643eca613b2dde09cdbf9",
            "54fc95cb87ca4f3293edd9945ec4618d",
            "91116a7703ca424697cb132c7630031e",
            "c31a73e844424565ab7db7f2ec92fd2b",
            "9d2a83764a8642b0ae9b5b0cd992c502",
            "79028584d4bd4917b9a80273655a6730",
            "47ff1e11fab844f38aeb59b028ec234b",
            "8553878a4c2a490d8bbdd7fffe0d1c07",
            "4b020cb3b50a4e2c84bc75177a194a4a",
            "1bb46b2aaf2e4ffd941db22a50ecdb71",
            "be626e6198364cefb10ea556a1c5db0f",
            "3608fa382b0a431dadba14f02ec774b4",
            "e855a9d8372247a093b14bd32bde89b1",
            "da3ec3940567421b8ebe35b50b13f911",
            "1d7cbe8309864a1d91e1516b6ea4ee33",
            "14b9ee169c47401d9e86828bfaf64e2a",
            "358142a7978547ff8fff2dee6fbc6166",
            "9546f5656c6c40d3b333a1ac68c81494",
            "408d09ef41a24bc2815d55735aad503e",
            "03d5fe0b754d40cdbebe181cc885ed13",
            "6e0a4cec14f246a08aa7a70fec5daac8",
            "371f867193be4783a0a1db0b4a17ed7b",
            "3807f31fbabb4a0b88e8e22156f18189",
            "209a9b321f924d45979fcbbd98a039d3",
            "ef2eb51fabe342e3953608b93af5df52",
            "9dd2664852b84689948529115fc9c160"
          ]
        },
        "id": "_zsBiAyj4q4i",
        "outputId": "27cffbcb-d5dc-4426-8196-fab3deb2cd42"
      },
      "source": [
        "tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "0b4ea732c4234b9dbdf264d10d3a3ba9",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=570.0, 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": "91116a7703ca424697cb132c7630031e",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=213450.0, 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": "be626e6198364cefb10ea556a1c5db0f",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=435797.0, 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": "408d09ef41a24bc2815d55735aad503e",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=29.0, style=ProgressStyle(description_w…"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mvzXyW504yaR"
      },
      "source": [
        "The tokenizer can take a string (or e.g. a list of strings) and performs all the necessary preprocessing steps to prepare its input for use by the model, including splitting input texts into tokens, adding padding and special characters, and mapping those to integer IDs (`input_ids`). The tokenizer also creates attention weights (`attention_mask`) for the model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WTj0eaev5MN-",
        "outputId": "4f6ba607-e2bd-4dad-e246-f2a9ad1710ab"
      },
      "source": [
        "tokenizer('Hello, BERT!')    # Demonstrate tokenizer output"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'input_ids': [101, 8667, 117, 139, 9637, 1942, 106, 102], 'token_type_ids': [0, 0, 0, 0, 0, 0, 0, 0], 'attention_mask': [1, 1, 1, 1, 1, 1, 1, 1]}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ufKlKoqsBpuq"
      },
      "source": [
        "We'll then define an encoding function applying the tokenizer to the text data  (key `\"sentence\"`) of a [`Dataset`](https://huggingface.co/docs/datasets/package_reference/main_classes.html?highlight=dataset#datasets.Dataset) object, and use the `map` function of the `DatasetDict` to tokenize the train, development, and test datasets."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 164,
          "referenced_widgets": [
            "2dbb371423b3466e9becc54b30190d4a",
            "785eb66e7b7441beb61c961b742e0239",
            "20cb76b46ea740028180748328ad2a4e",
            "6ffedf1bc38f4b4dbf1276a859830745",
            "76cfc6f1071046a09b04eec07acd52dd",
            "fd8ea05e5cb743eeaeec09a41a2a5ed0",
            "fbe0da7b6bbb452eb3b0ca438a65da7b",
            "6698ba0edf484d1ba9b50185c300abe1",
            "b4e49976611a4067929dd65940b62d07",
            "b4b5991f1dca4a0ea48dea5dc86985ac",
            "c93ad7720d8e4504a0ade349143676ce",
            "e4148066e4264fd291c5c471088b62cb",
            "3b580f8cd0d8445a9dffb245a64964d1",
            "5b719123bfc34f6dad74c9e8e4dacf2e",
            "8013cbea9339492f92916e97f5a900f8",
            "519ac51e146b4416b062a6b1462e3129",
            "7708b8ecd0ed44a0a1335f116277f4ba",
            "e1bf8aa2e62f4bf284193a5034ef2ca7",
            "3ab350ec6a09481aad8e3b0b912703c4",
            "9757a8eac4024ff3b1db99df6aeeb38f",
            "3a4357ee6d9845a5922132108f9d2d9b",
            "c33c470b41cc46ddb23c7a8af6fd2ed7",
            "6d40a93298cd433f9ca02aa752c5cbc6",
            "4b1dc8a5fdf940d1ae03e2b5631ce4d7"
          ]
        },
        "id": "m-9TlSVDBx0E",
        "outputId": "9ef792cb-10f0-4176-b3b4-80bfdad8445f"
      },
      "source": [
        "def encode_dataset(d):\n",
        "  return tokenizer(d['sentence'])\n",
        "\n",
        "encoded_dataset = dataset.map(encode_dataset)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "2dbb371423b3466e9becc54b30190d4a",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, max=67349.0), HTML(value='')))"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "b4e49976611a4067929dd65940b62d07",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, max=872.0), HTML(value='')))"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "7708b8ecd0ed44a0a1335f116277f4ba",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, max=1821.0), HTML(value='')))"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aN2CBQnp5pEF"
      },
      "source": [
        "## Load pre-trained model\n",
        "\n",
        "Next, we'll load the pre-trained model with support for text classification output using [`AutoModelForSequenceClassification.from_prertained`](https://huggingface.co/transformers/model_doc/auto.html#transformers.AutoModelForSequenceClassification.from_pretrained).\n",
        "\n",
        "Note that we need to provide the number of labels in the data when loading the model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 171,
          "referenced_widgets": [
            "1e6f9d8c22f24980b365046a91bce910",
            "030beb0fee49412b80fd7ae4deaea85c",
            "74c1cd41c17a4096b9620aafab815a88",
            "75ee7bdf5211434c9e1f54dbe42453cb",
            "277f281606ec4602997d1670c40ba923",
            "dd86296fa3f64e519d46e03b083a1807",
            "83cb7c3303ad4dfaa1e2429c60e9cedf",
            "92d824d31a544b54a88d91a14228f707"
          ]
        },
        "id": "6Pq-BuIm86YD",
        "outputId": "36bcf96a-1f52-4168-e4a2-f1d3f17d4210"
      },
      "source": [
        "model = AutoModelForSequenceClassification.from_pretrained(MODEL_NAME, num_labels=num_labels)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "1e6f9d8c22f24980b365046a91bce910",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, description='Downloading', max=435779157.0, 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-cased 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-cased 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": "markdown",
      "metadata": {
        "id": "qCcf-EZF9x0j"
      },
      "source": [
        "## Training parameters and metrics\n",
        "We're almost ready to train. We'll next create a [`TrainingArguments`](https://huggingface.co/transformers/main_classes/trainer.html#transformers.TrainingArguments) object to hold the hyperparameters and other settings that are appropriate for training on Colab:\n",
        "\n",
        "* `save_strategy`: set to `\"epoch\"` so that model checkpoints are saved every epoch, this I can do because I save to own Google Drive\n",
        "* `evaluation_strategy` and `logging_strategy` set to `\"epoch\"` so that evaluation and logging are performed once per epoch\n",
        "* The hyperparameters `LEARNING_RATE`, `BATCH_SIZE` and `TRAIN_EPOCHS` set above are passed to the training process through this object"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vV_hOr7y_xaM"
      },
      "source": [
        "Finally, we'll define a simple accuracy metric measuring how many predictions match their correct values."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6Wd3_VaQ_94D"
      },
      "source": [
        "def compute_accuracy(pred):\n",
        "    y_pred = pred.predictions.argmax(axis=1)\n",
        "    y_true = pred.label_ids\n",
        "    return { 'accuracy': sum(y_pred == y_true) / len(y_true) }"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nyRihTPA_SOU"
      },
      "source": [
        "# Training (fine-tuning)\n",
        "\n",
        "For fine-tuning the pre-trained model, we'll create a [`Trainer`](https://huggingface.co/transformers/main_classes/trainer.html) object, providing it with the pre-trained model, settings, training and development (validation) data, and the evaluation metric created above."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cH-oojr_1sOq"
      },
      "source": [
        "\n",
        "\n",
        "# Hyperparameters\n",
        "LEARNING_RATE=1e-4\n",
        "BATCH_SIZE=128\n",
        "TRAIN_EPOCHS=4"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CqCC-WEWnJYz",
        "outputId": "537684fa-096c-47d0-a91b-1a229a18e960"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9p0qlQrl-NSp"
      },
      "source": [
        "train_args = TrainingArguments(\n",
        "    '/content/drive/MyDrive/WorkStuff/ssd2-sentiment-model',    # output directory for checkpoints and predictions\n",
        "    load_best_model_at_end=True,\n",
        "    evaluation_strategy='epoch',\n",
        "    logging_strategy='epoch',\n",
        "    learning_rate=LEARNING_RATE,\n",
        "    per_device_train_batch_size=BATCH_SIZE,\n",
        "    num_train_epochs=TRAIN_EPOCHS,\n",
        ")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3uwIGX6Y_fP8"
      },
      "source": [
        "trainer = Trainer(\n",
        "      model,\n",
        "      train_args,\n",
        "      train_dataset=encoded_dataset['train'],\n",
        "      eval_dataset=encoded_dataset['validation'],\n",
        "      tokenizer=tokenizer,\n",
        "      compute_metrics=compute_accuracy\n",
        ")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VsEpkLgOAQ9c"
      },
      "source": [
        "Training is then performed simply by calling the `train` function of the `Trainer` object."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 236
        },
        "id": "Zk0pspNEAeWr",
        "outputId": "402f9867-fefc-4332-aa08-b54344771d59"
      },
      "source": [
        "trainer.train()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "        </style>\n",
              "      \n",
              "      <progress value='2108' max='2108' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      [2108/2108 40:50, Epoch 4/4]\n",
              "    </div>\n",
              "    <table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>Epoch</th>\n",
              "      <th>Training Loss</th>\n",
              "      <th>Validation Loss</th>\n",
              "      <th>Accuracy</th>\n",
              "      <th>Runtime</th>\n",
              "      <th>Samples Per Second</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>0.224500</td>\n",
              "      <td>0.251691</td>\n",
              "      <td>0.901376</td>\n",
              "      <td>3.350500</td>\n",
              "      <td>260.260000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>0.112600</td>\n",
              "      <td>0.265011</td>\n",
              "      <td>0.903670</td>\n",
              "      <td>3.331200</td>\n",
              "      <td>261.767000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>0.064200</td>\n",
              "      <td>0.299133</td>\n",
              "      <td>0.905963</td>\n",
              "      <td>3.359100</td>\n",
              "      <td>259.593000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>0.031400</td>\n",
              "      <td>0.303404</td>\n",
              "      <td>0.915138</td>\n",
              "      <td>3.323100</td>\n",
              "      <td>262.409000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table><p>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "TrainOutput(global_step=2108, training_loss=0.10818358209599127, metrics={'train_runtime': 2452.1031, 'train_samples_per_second': 0.86, 'total_flos': 9023432196007320.0, 'epoch': 4.0, 'init_mem_cpu_alloc_delta': 0, 'init_mem_gpu_alloc_delta': 433776128, 'init_mem_cpu_peaked_delta': 0, 'init_mem_gpu_peaked_delta': 0, 'train_mem_cpu_alloc_delta': 5898240, 'train_mem_gpu_alloc_delta': 1741112320, 'train_mem_cpu_peaked_delta': 0, 'train_mem_gpu_peaked_delta': 6627404800})"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 31
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ERPvLNK9EcN8"
      },
      "source": [
        "## Evaluation and predictions for user input\n",
        "\n",
        "We can use `trainer.evaluate()` to evaluate the trained model on the `eval_dataset` given to the trainer:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        },
        "id": "-f-XXjvkE0aL",
        "outputId": "d0a9aa03-5b4c-4d81-a1a1-ec288199198b"
      },
      "source": [
        "results = trainer.evaluate()\n",
        "print(f'Accuracy: {results[\"eval_accuracy\"]}')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "        </style>\n",
              "      \n",
              "      <progress value='109' max='109' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      [109/109 00:03]\n",
              "    </div>\n",
              "    "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Accuracy: 0.9162844036697247\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RazZoJR2U3Lo"
      },
      "source": [
        "# Model save\n",
        "\n",
        "* Now that the model is trained, we may want to save it to Google Drive\n",
        "* That way we don't need to train it every time"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9mu-2AHanki8"
      },
      "source": [
        "import torch\n",
        "#make sure you make this directory\n",
        "torch.save(trainer.model,\"/content/drive/MyDrive/WorkStuff/sent_model.pt\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7mcnYVMdYDf_"
      },
      "source": [
        "# Model load\n",
        "\n",
        "* Now we load the model and repeat some of the imports so it is possible to run the notebook from this point onwards"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "99hMGMUra14q"
      },
      "source": [
        "# relevant stuff repeated from above so you can run it from here onwards if you happen to have the model trained\n",
        "\n",
        "from transformers import AutoTokenizer\n",
        "from transformers import AutoModelForSequenceClassification\n",
        "from transformers import TrainingArguments\n",
        "from transformers import Trainer\n",
        "from datasets import load_dataset\n",
        "\n",
        "MODEL_NAME = 'bert-base-cased'\n",
        "DATASET = ('glue', 'sst2')\n",
        "\n",
        "tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "vmTiUqOEcfrH",
        "outputId": "395ae31b-f43d-47f0-9bf5-a9b28edc95af"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ocPeVRnPWF_m"
      },
      "source": [
        "import torch\n",
        "model = torch.load(\"/content/drive/MyDrive/WorkStuff/sent_model.pt\")\n",
        "model.to('cpu')    # simplifies input placement\n",
        "\n",
        "sentiment_label = [\n",
        "    'negative',\n",
        "    'positive'\n",
        "]\n",
        "\n",
        "def predict_sentiment(string):\n",
        "    tokenized = tokenizer(string, return_tensors='pt')\n",
        "    pred = model(**tokenized)\n",
        "    pred_idx = pred.logits.detach().numpy().argmax()\n",
        "    return sentiment_label[pred_idx]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "G-up4Ua7iMMO"
      },
      "source": [
        "try that out"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "a_QcEfttZgn9",
        "outputId": "a6e163ab-0d49-4634-c8ff-3c1f39a30c9b"
      },
      "source": [
        "example_sentences = [\n",
        "    'I think this is very good, and I am in support.',\n",
        "    'This is bad news.',\n",
        "    'This could not have been better.',\n",
        "    'This is not good news at all.',\n",
        "    'Donald Trump lost the election and will no longer be every day on CNN.',\n",
        "    'I think your service fulfilled my expectations, I could not really expect you to be any better'\n",
        "]\n",
        "\n",
        "for e in example_sentences:\n",
        "    print(e, '->', predict_sentiment(e))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "I think this is very good, and I am in support. -> positive\n",
            "This is bad news. -> negative\n",
            "This could not have been better. -> positive\n",
            "This is not good news at all. -> negative\n",
            "Donald Trump lost the election and will no longer be every day on CNN. -> negative\n",
            "I think your service fulfilled my expectations, I could not really expect you to be any better -> positive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "I2oGRF0ssbYh"
      },
      "source": [
        "# Model explainability with the captum library\n",
        "\n",
        "* captum.ai is a nifty library with many of the explainability algorithms implemented\n",
        "* we will use it here to try the techniques\n",
        "* this is not the easiest of code, btw, many of the libraries are bleeding-edge, so you cut yourself here and there (the blood on the edge is yours)\n",
        "* the code below is (quite substantial) adaptation of https://captum.ai/tutorials/Bert_SQUAD_Interpret"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wz69yC5EbkFA",
        "outputId": "4d99db63-1134-40f3-d747-c80494678cb3"
      },
      "source": [
        "!pip install captum pandas matplotlib seaborn"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting captum\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/13/b0/8fa3ab89e2e37c960cdd09595fa911fbb8d6da216c8bc98e18c858a0128d/captum-0.3.1-py3-none-any.whl (4.4MB)\n",
            "\u001b[K     |████████████████████████████████| 4.4MB 8.3MB/s \n",
            "\u001b[?25hRequirement already satisfied: pandas in /usr/local/lib/python3.7/dist-packages (1.1.5)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (3.2.2)\n",
            "Requirement already satisfied: seaborn in /usr/local/lib/python3.7/dist-packages (0.11.1)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from captum) (1.19.5)\n",
            "Requirement already satisfied: torch>=1.2 in /usr/local/lib/python3.7/dist-packages (from captum) (1.8.1+cu101)\n",
            "Requirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.7/dist-packages (from pandas) (2018.9)\n",
            "Requirement already satisfied: python-dateutil>=2.7.3 in /usr/local/lib/python3.7/dist-packages (from pandas) (2.8.1)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib) (1.3.1)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib) (0.10.0)\n",
            "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib) (2.4.7)\n",
            "Requirement already satisfied: scipy>=1.0 in /usr/local/lib/python3.7/dist-packages (from seaborn) (1.4.1)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from torch>=1.2->captum) (3.7.4.3)\n",
            "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.7.3->pandas) (1.15.0)\n",
            "Installing collected packages: captum\n",
            "Successfully installed captum-0.3.1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cRDMMKWtb1DU"
      },
      "source": [
        "from captum.attr import visualization as viz\n",
        "from captum.attr import IntegratedGradients, LayerConductance, LayerIntegratedGradients\n",
        "from captum.attr import configure_interpretable_embedding_layer, remove_interpretable_embedding_layer"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G7ieBNPfcJF8"
      },
      "source": [
        "#Tells the model that it is in evaluation mode, and zeroes out the gradients\n",
        "model.eval()\n",
        "model.zero_grad()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JV52we4xGJK2"
      },
      "source": [
        "# Helper functions"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BmHmhdRAeMW6"
      },
      "source": [
        "# Forward on the model -> data in, prediction out, nothing fancy really\n",
        "def predict(inputs, token_type_ids=None, position_ids=None, attention_mask=None):\n",
        "    pred=model(inputs, token_type_ids=token_type_ids,\n",
        "                 position_ids=position_ids, attention_mask=attention_mask)\n",
        "    return pred.logits #return the output of the classification layer\n",
        "\n",
        "ref_token_id = tokenizer.pad_token_id # A token used for generating token reference\n",
        "sep_token_id = tokenizer.sep_token_id # A token used as a separator between question and text and it is also added to the end of the text.\n",
        "cls_token_id = tokenizer.cls_token_id # A token used for prepending to the concatenated question-text word sequence\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6rQx10XHpRmq"
      },
      "source": [
        "# Given input text, construct a pair of (text input, blank reference input as long as the text itself)\n",
        "# ...token indices:\n",
        "def construct_input_ref_pair(text, ref_token_id, sep_token_id, cls_token_id, device):\n",
        "    text_ids = tokenizer.encode(text, add_special_tokens=False)\n",
        "\n",
        "    # construct input token ids\n",
        "    input_ids = [cls_token_id] + text_ids + [sep_token_id] #the standard way of feeding the input in\n",
        "\n",
        "    # construct reference token ids \n",
        "    ref_input_ids = [cls_token_id] + [ref_token_id] * len(text_ids) + [sep_token_id]  #basically [CLS] [PAD] [PAD] [PAD] ... [SEP] ... blank\n",
        "    return torch.tensor([input_ids], device=device), torch.tensor([ref_input_ids], device=device)\n",
        "\n",
        "# ...token types - since we only have one sentence, these are always 0\n",
        "def construct_input_ref_token_type_pair(input_ids, device):\n",
        "    seq_len = input_ids.size(1)\n",
        "    token_type_ids = torch.zeros((1,seq_len), dtype=torch.long, device=device)\n",
        "    ref_token_type_ids = torch.zeros_like(token_type_ids, device=device)\n",
        "    return token_type_ids, ref_token_type_ids\n",
        "\n",
        "# ...token positions\n",
        "def construct_input_ref_pos_id_pair(input_ids, device):\n",
        "    seq_length = input_ids.size(1)\n",
        "    position_ids = torch.arange(seq_length, dtype=torch.long, device=device)\n",
        "    # we could potentially also use random permutation with `torch.randperm(seq_length, device=device)`\n",
        "    ref_position_ids = torch.zeros(seq_length, dtype=torch.long, device=device)\n",
        "\n",
        "    #make sure shapes match\n",
        "    position_ids = position_ids.unsqueeze(0).expand_as(input_ids)\n",
        "    ref_position_ids = ref_position_ids.unsqueeze(0).expand_as(input_ids)\n",
        "    return position_ids, ref_position_ids\n",
        "\n",
        "# ...attention mask, that is the same for both input and reference and basically all ones\n",
        "def construct_attention_mask(input_ids,device):\n",
        "    return torch.ones_like(input_ids,device=device)\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QamJbfoFuurn",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "70f147e8-0011-40e2-a56b-5823e0da37cb"
      },
      "source": [
        "# Let's try it!\n",
        "\n",
        "device=model.device\n",
        "\n",
        "text=\"This piece of news is great, even if you think it is rather unfortunate\"\n",
        "\n",
        "#input:\n",
        "input_ids, ref_input_ids = construct_input_ref_pair(text, ref_token_id, sep_token_id, cls_token_id, device)\n",
        "#token type:\n",
        "token_type_ids, ref_token_type_ids = construct_input_ref_token_type_pair(input_ids, device)\n",
        "#position ids:\n",
        "position_ids, ref_position_ids = construct_input_ref_pos_id_pair(input_ids, device)\n",
        "#attention mask:\n",
        "attention_mask = construct_attention_mask(input_ids, device)\n",
        "\n",
        "all_tokens = tokenizer.convert_ids_to_tokens(input_ids[0])\n",
        "print(all_tokens)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "['[CLS]', 'This', 'piece', 'of', 'news', 'is', 'great', ',', 'even', 'if', 'you', 'think', 'it', 'is', 'rather', 'unfortunate', '[SEP]']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FfUPcEF_vx9z",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "622ff508-85d5-462c-e671-3b9b054ff2e8"
      },
      "source": [
        "p=predict(input_ids,token_type_ids=token_type_ids,position_ids=position_ids,attention_mask=attention_mask)\n",
        "print(\"p=\",p)\n",
        "print(\"p.shape\",p.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "p= tensor([[-2.3281,  2.5443]], grad_fn=<AddmmBackward>)\n",
            "p.shape torch.Size([1, 2])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3yEjjc7_xcy1"
      },
      "source": [
        "# Yay, now we finally made it to the attribution part\n",
        "lig = LayerIntegratedGradients(predict, model.bert.embeddings) #attribute the output wrt to embeddings"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eDeDDO2Lx1JG",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "418bc462-dd33-4868-fdc6-c2d1c90e319b"
      },
      "source": [
        "# inputs: inputs\n",
        "# baselines: the blank baseline\n",
        "# target: which of the two classes in the output (pos/neg) to run the prediction against?\n",
        "attrs, delta = lig.attribute(inputs=(input_ids,token_type_ids,position_ids,attention_mask),\n",
        "                                  baselines=(ref_input_ids,ref_token_type_ids,ref_position_ids,attention_mask),\n",
        "                                  return_convergence_delta=True,target=1)\n",
        "print(\"attrs shape\",attrs.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "attrs shape torch.Size([1, 17, 768])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Dgn5rRRZyZgV",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f7dcb0ed-0db7-47d5-ead8-6ad209fb59fe"
      },
      "source": [
        "def summarize_attributions(attributions):\n",
        "    attributions = attributions.sum(dim=-1).squeeze(0)\n",
        "    attributions = attributions / torch.norm(attributions)\n",
        "    return attributions\n",
        "\n",
        "attrs_sum = summarize_attributions(attrs)\n",
        "print(\"attrs_sum shape\",attrs_sum.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "attrs_sum shape torch.Size([17])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LmbI0pCIymt_",
        "outputId": "3eb96b19-e2a3-47eb-f9aa-77a29e90d722"
      },
      "source": [
        "print(attrs_sum)\n",
        "print(tokenizer.convert_ids_to_tokens(input_ids[0]))\n",
        "\n",
        "for a,t in zip(attrs_sum,tokenizer.convert_ids_to_tokens(input_ids[0])):\n",
        "    print(float(a),t)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tensor([ 0.0000, -0.0561, -0.0919, -0.0823, -0.0769, -0.0770,  0.8852, -0.0458,\n",
            "         0.0270, -0.1094, -0.0727, -0.1350, -0.0979, -0.1441, -0.1881, -0.2807,\n",
            "         0.0575], dtype=torch.float64)\n",
            "['[CLS]', 'This', 'piece', 'of', 'news', 'is', 'great', ',', 'even', 'if', 'you', 'think', 'it', 'is', 'rather', 'unfortunate', '[SEP]']\n",
            "0.0 [CLS]\n",
            "-0.05606936110561358 This\n",
            "-0.0918954617673179 piece\n",
            "-0.082304149964058 of\n",
            "-0.07693875808018998 news\n",
            "-0.07697800735507292 is\n",
            "0.8852405197762391 great\n",
            "-0.0457627638431411 ,\n",
            "0.027032084672280212 even\n",
            "-0.10943371691491163 if\n",
            "-0.07266291499865889 you\n",
            "-0.1349669216789558 think\n",
            "-0.09788820942518903 it\n",
            "-0.14408741002558845 is\n",
            "-0.18808658280174176 rather\n",
            "-0.28074980323111726 unfortunate\n",
            "0.05748516461738326 [SEP]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "238tqqadDPbq"
      },
      "source": [
        "Damn, that seems to work!"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 41
        },
        "id": "mjvnyJLaOhWK",
        "outputId": "3098e75d-1101-43c5-89ef-70d09a33e265"
      },
      "source": [
        "import captum\n",
        "from IPython.core.display import HTML, display\n",
        "x=captum.attr.visualization.format_word_importances(tokenizer.convert_ids_to_tokens(input_ids[0]),attrs_sum)\n",
        "HTML(x)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<td><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> [CLS]                    </font></mark><mark style=\"background-color: hsl(0, 75%, 98%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> This                    </font></mark><mark style=\"background-color: hsl(0, 75%, 97%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> piece                    </font></mark><mark style=\"background-color: hsl(0, 75%, 97%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> of                    </font></mark><mark style=\"background-color: hsl(0, 75%, 97%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> news                    </font></mark><mark style=\"background-color: hsl(0, 75%, 97%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> is                    </font></mark><mark style=\"background-color: hsl(120, 75%, 56%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> great                    </font></mark><mark style=\"background-color: hsl(0, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> ,                    </font></mark><mark style=\"background-color: hsl(120, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> even                    </font></mark><mark style=\"background-color: hsl(0, 75%, 96%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> if                    </font></mark><mark style=\"background-color: hsl(0, 75%, 98%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> you                    </font></mark><mark style=\"background-color: hsl(0, 75%, 95%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> think                    </font></mark><mark style=\"background-color: hsl(0, 75%, 97%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> it                    </font></mark><mark style=\"background-color: hsl(0, 75%, 95%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> is                    </font></mark><mark style=\"background-color: hsl(0, 75%, 93%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> rather                    </font></mark><mark style=\"background-color: hsl(0, 75%, 89%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> unfortunate                    </font></mark><mark style=\"background-color: hsl(120, 75%, 98%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> [SEP]                    </font></mark></td>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 226
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "q1po59EFXwYo"
      },
      "source": [
        "# Almost there...\n",
        "\n",
        "* Let's wrap this all into a function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9js3D6h0Xj__"
      },
      "source": [
        "\n",
        "\n",
        "def predict_and_explain(model,text):\n",
        "    model.zero_grad() #to be safe perhaps it's not needed\n",
        "    device=model.device\n",
        "    #input:\n",
        "    input_ids, ref_input_ids = construct_input_ref_pair(text, ref_token_id, sep_token_id, cls_token_id, device)\n",
        "    #token type:\n",
        "    token_type_ids, ref_token_type_ids = construct_input_ref_token_type_pair(input_ids, device)\n",
        "    #position ids:\n",
        "    position_ids, ref_position_ids = construct_input_ref_pos_id_pair(input_ids, device)\n",
        "    #attention mask:\n",
        "    attention_mask = construct_attention_mask(input_ids, device)\n",
        "\n",
        "    all_tokens = tokenizer.convert_ids_to_tokens(input_ids[0])\n",
        "\n",
        "    lig = LayerIntegratedGradients(predict, model.bert.embeddings)\n",
        "    prediction=predict(input_ids,token_type_ids,position_ids,attention_mask)[0]\n",
        "    prediction_cls=int(torch.argmax(prediction))\n",
        "    print(\"Prediction:\", (\"negative\",\"positive\")[prediction_cls],\"Weights:\",prediction.tolist())\n",
        "    for target,classname in enumerate((\"negative\",\"positive\")):\n",
        "        \n",
        "        attrs, delta = lig.attribute(inputs=(input_ids,token_type_ids,position_ids,attention_mask),\n",
        "                                  baselines=(ref_input_ids,ref_token_type_ids,ref_position_ids,attention_mask),\n",
        "                                  return_convergence_delta=True,target=target)\n",
        "        attrs_sum = summarize_attributions(attrs)\n",
        "\n",
        "        x=captum.attr.visualization.format_word_importances(all_tokens,attrs_sum)\n",
        "        print(\"ATTRIBUTION WITH RESPECT TO\",classname)\n",
        "        display(HTML(x))\n",
        "        print()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 150
        },
        "id": "k7TGcIAWYryu",
        "outputId": "3ff1e857-2dfa-4049-c342-2cb175e4f987"
      },
      "source": [
        "predict_and_explain(model,\"I like Filip, who used to hang out in the Turku office but now is back to the university! Hell, *everyone* likes him.\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Prediction: positive Weights: [-1.37552011013031, 1.6477326154708862]\n",
            "ATTRIBUTION WITH RESPECT TO negative\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<td><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> [CLS]                    </font></mark><mark style=\"background-color: hsl(0, 75%, 96%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> I                    </font></mark><mark style=\"background-color: hsl(0, 75%, 89%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> like                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> Fi                    </font></mark><mark style=\"background-color: hsl(0, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> ##lip                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> ,                    </font></mark><mark style=\"background-color: hsl(0, 75%, 97%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> who                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> used                    </font></mark><mark style=\"background-color: hsl(120, 75%, 98%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> to                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> hang                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> out                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> in                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> the                    </font></mark><mark style=\"background-color: hsl(0, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> Tu                    </font></mark><mark style=\"background-color: hsl(0, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> ##rk                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> ##u                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> office                    </font></mark><mark style=\"background-color: hsl(120, 75%, 97%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> but                    </font></mark><mark style=\"background-color: hsl(0, 75%, 97%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> now                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> is                    </font></mark><mark style=\"background-color: hsl(120, 75%, 97%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> back                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> to                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> the                    </font></mark><mark style=\"background-color: hsl(0, 75%, 98%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> university                    </font></mark><mark style=\"background-color: hsl(0, 75%, 95%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> !                    </font></mark><mark style=\"background-color: hsl(120, 75%, 61%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> Hell                    </font></mark><mark style=\"background-color: hsl(120, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> ,                    </font></mark><mark style=\"background-color: hsl(0, 75%, 95%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> *                    </font></mark><mark style=\"background-color: hsl(120, 75%, 97%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> everyone                    </font></mark><mark style=\"background-color: hsl(0, 75%, 97%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> *                    </font></mark><mark style=\"background-color: hsl(0, 75%, 84%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> likes                    </font></mark><mark style=\"background-color: hsl(0, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> him                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> .                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> [SEP]                    </font></mark></td>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "ATTRIBUTION WITH RESPECT TO positive\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<td><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> [CLS]                    </font></mark><mark style=\"background-color: hsl(120, 75%, 94%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> I                    </font></mark><mark style=\"background-color: hsl(120, 75%, 84%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> like                    </font></mark><mark style=\"background-color: hsl(120, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> Fi                    </font></mark><mark style=\"background-color: hsl(120, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> ##lip                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> ,                    </font></mark><mark style=\"background-color: hsl(120, 75%, 96%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> who                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> used                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> to                    </font></mark><mark style=\"background-color: hsl(120, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> hang                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> out                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> in                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> the                    </font></mark><mark style=\"background-color: hsl(120, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> Tu                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> ##rk                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> ##u                    </font></mark><mark style=\"background-color: hsl(120, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> office                    </font></mark><mark style=\"background-color: hsl(0, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> but                    </font></mark><mark style=\"background-color: hsl(120, 75%, 96%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> now                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> is                    </font></mark><mark style=\"background-color: hsl(0, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> back                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> to                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> the                    </font></mark><mark style=\"background-color: hsl(120, 75%, 98%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> university                    </font></mark><mark style=\"background-color: hsl(120, 75%, 94%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> !                    </font></mark><mark style=\"background-color: hsl(0, 75%, 69%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> Hell                    </font></mark><mark style=\"background-color: hsl(0, 75%, 99%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> ,                    </font></mark><mark style=\"background-color: hsl(120, 75%, 94%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> *                    </font></mark><mark style=\"background-color: hsl(0, 75%, 98%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> everyone                    </font></mark><mark style=\"background-color: hsl(120, 75%, 95%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> *                    </font></mark><mark style=\"background-color: hsl(120, 75%, 80%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> likes                    </font></mark><mark style=\"background-color: hsl(120, 75%, 98%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> him                    </font></mark><mark style=\"background-color: hsl(0, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> .                    </font></mark><mark style=\"background-color: hsl(120, 75%, 100%); opacity:1.0;                     line-height:1.75\"><font color=\"black\"> [SEP]                    </font></mark></td>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}