Transformers_example.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Transformers_example.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyPadjT/CXpGse6FapSY/vsz",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU",
    "gpuClass": "standard",
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "a47a6b8bd0974681ac9d39203fa26915": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_8da628d2070d4308ae7ba20a11dfabf8",
              "IPY_MODEL_3ac81a77bcdf4da194c5ac4409770f1b",
              "IPY_MODEL_9fc1c7b64309485eaa411813e8dc99b0"
            ],
            "layout": "IPY_MODEL_8c5a9a47cdff4b6186611459a68c7d78"
          }
        },
        "8da628d2070d4308ae7ba20a11dfabf8": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_9321160bf2bc40edb65286baabbe7dc7",
            "placeholder": "​",
            "style": "IPY_MODEL_0ac3934799ff48cc818343e069227f9b",
            "value": "Dl Completed...: 100%"
          }
        },
        "3ac81a77bcdf4da194c5ac4409770f1b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_5eda2751d7e14e19a6ac8796f0177565",
            "max": 1,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_33df1cb7900b4ed6a6e512c583af22de",
            "value": 1
          }
        },
        "9fc1c7b64309485eaa411813e8dc99b0": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_245a4aece6e04658b5d18ecb1fa36ba3",
            "placeholder": "​",
            "style": "IPY_MODEL_22aabd693ca14aaaba90aef01bc3e1d6",
            "value": " 1/1 [00:16&lt;00:00,  7.72s/ url]"
          }
        },
        "8c5a9a47cdff4b6186611459a68c7d78": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "9321160bf2bc40edb65286baabbe7dc7": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "0ac3934799ff48cc818343e069227f9b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "5eda2751d7e14e19a6ac8796f0177565": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": "20px"
          }
        },
        "33df1cb7900b4ed6a6e512c583af22de": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "245a4aece6e04658b5d18ecb1fa36ba3": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "22aabd693ca14aaaba90aef01bc3e1d6": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "62b35e2df7984aa1bba29248380a2621": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_96832a6965c14547b565a5be398452f3",
              "IPY_MODEL_78c93918ce61423bae61f7d3c1e02536",
              "IPY_MODEL_1a4af1a28e234d0994ca84ae05d3e6ef"
            ],
            "layout": "IPY_MODEL_fb56044c2a7a40d897a55ba2f9653c57"
          }
        },
        "96832a6965c14547b565a5be398452f3": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_992e312ebf7949fe9c2706e511ed0e12",
            "placeholder": "​",
            "style": "IPY_MODEL_13764c7c8405465388abee5e636b5d31",
            "value": "Dl Size...: 100%"
          }
        },
        "78c93918ce61423bae61f7d3c1e02536": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_0e649c0ba0924429ad5e0a8e931fab8d",
            "max": 1,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_17894bdcedbb44769e6cb69adc3c5dcf",
            "value": 1
          }
        },
        "1a4af1a28e234d0994ca84ae05d3e6ef": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_e6b4fec1e6534d129a46cb094d3e7970",
            "placeholder": "​",
            "style": "IPY_MODEL_afdbbb29e9bf45cc9d6b242a811d43f2",
            "value": " 124/124 [00:16&lt;00:00, 14.62 MiB/s]"
          }
        },
        "fb56044c2a7a40d897a55ba2f9653c57": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "992e312ebf7949fe9c2706e511ed0e12": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "13764c7c8405465388abee5e636b5d31": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "0e649c0ba0924429ad5e0a8e931fab8d": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": "20px"
          }
        },
        "17894bdcedbb44769e6cb69adc3c5dcf": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "e6b4fec1e6534d129a46cb094d3e7970": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "afdbbb29e9bf45cc9d6b242a811d43f2": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "9d3628e72de04db2afd55ed758767d67": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_7d6588165503434088be28d5b777bb7a",
              "IPY_MODEL_bc26fb93b3d34c26ab10b94035f81a86",
              "IPY_MODEL_349f278f06fd479187716730e564018a"
            ],
            "layout": "IPY_MODEL_eb8a7ccf3adf46f8991ca05538bf52a6"
          }
        },
        "7d6588165503434088be28d5b777bb7a": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_86038ea1a53d47dba5adafe981c8a3b4",
            "placeholder": "​",
            "style": "IPY_MODEL_a50f78f5c8254bdda191d4ad2172008a",
            "value": "Extraction completed...: 100%"
          }
        },
        "bc26fb93b3d34c26ab10b94035f81a86": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_9edc499924a1459cbdbf5a6d2ec6f27c",
            "max": 1,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_5a20956c30e04a95aadc3adfac751939",
            "value": 1
          }
        },
        "349f278f06fd479187716730e564018a": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_39fcad7c783b4aab93462904d07e7e85",
            "placeholder": "​",
            "style": "IPY_MODEL_ad09641fab04493d8c0ede403fb51a72",
            "value": " 1/1 [00:16&lt;00:00, 16.40s/ file]"
          }
        },
        "eb8a7ccf3adf46f8991ca05538bf52a6": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "86038ea1a53d47dba5adafe981c8a3b4": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "a50f78f5c8254bdda191d4ad2172008a": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "9edc499924a1459cbdbf5a6d2ec6f27c": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": "20px"
          }
        },
        "5a20956c30e04a95aadc3adfac751939": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "39fcad7c783b4aab93462904d07e7e85": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "ad09641fab04493d8c0ede403fb51a72": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "5b118b411dd942c28a76b683ed9f5be4": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_76ef8e5413144e339cb6771ad8a456d8",
              "IPY_MODEL_e5313446e0734de199d5832bdb72b97a",
              "IPY_MODEL_88b8da8515894553adc75bed9d89099a"
            ],
            "layout": "IPY_MODEL_60c40ef411164b01801427e5b5b2d331"
          }
        },
        "76ef8e5413144e339cb6771ad8a456d8": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_a8bc7ed49ec545e6a41450e7ef256048",
            "placeholder": "​",
            "style": "IPY_MODEL_2d6ef9ca68784e5ca691c61888f507e2",
            "value": ""
          }
        },
        "e5313446e0734de199d5832bdb72b97a": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "info",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_6a9577849a2a4b33bc9d02e0c59647c4",
            "max": 1,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_5dbcb775249a4e5f9e256b667ae6bd27",
            "value": 1
          }
        },
        "88b8da8515894553adc75bed9d89099a": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_3081ecfea6d54a24934c40b612e35869",
            "placeholder": "​",
            "style": "IPY_MODEL_9e22673d0463498999a6cd587a6dd84e",
            "value": " 51239/0 [00:13&lt;00:00, 7319.21 examples/s]"
          }
        },
        "60c40ef411164b01801427e5b5b2d331": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "a8bc7ed49ec545e6a41450e7ef256048": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "2d6ef9ca68784e5ca691c61888f507e2": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "6a9577849a2a4b33bc9d02e0c59647c4": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": "20px"
          }
        },
        "5dbcb775249a4e5f9e256b667ae6bd27": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "3081ecfea6d54a24934c40b612e35869": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "9e22673d0463498999a6cd587a6dd84e": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "549857ca63b1406d9007f463edc36480": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_8e3664cc2b9f400487ebf1aaf85ddaf4",
              "IPY_MODEL_6741f6e4389a44b4a4b367d2259f5174",
              "IPY_MODEL_bc732c8af9ee41b98cb6e0b4ff315f94"
            ],
            "layout": "IPY_MODEL_03166b1c7dfa4ec5b09f4014c8cfd8f5"
          }
        },
        "8e3664cc2b9f400487ebf1aaf85ddaf4": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_dc16367ad4e54c909125227273225cb9",
            "placeholder": "​",
            "style": "IPY_MODEL_3f7ff04a06c642f3b62fef2c3e7a009c",
            "value": "100%"
          }
        },
        "6741f6e4389a44b4a4b367d2259f5174": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "danger",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_237561ec12b4470cbf487c5429b2419f",
            "max": 51785,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_a1a65f2a42d54328ba329c2f11516516",
            "value": 51784
          }
        },
        "bc732c8af9ee41b98cb6e0b4ff315f94": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_03d33320ff314c6b9bda5c3151623b47",
            "placeholder": "​",
            "style": "IPY_MODEL_4f7ffd5dcff84e488f6f0e4454c4e2b2",
            "value": " 51784/51785 [00:00&lt;00:00, 286759.88 examples/s]"
          }
        },
        "03166b1c7dfa4ec5b09f4014c8cfd8f5": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "dc16367ad4e54c909125227273225cb9": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "3f7ff04a06c642f3b62fef2c3e7a009c": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "237561ec12b4470cbf487c5429b2419f": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "a1a65f2a42d54328ba329c2f11516516": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "03d33320ff314c6b9bda5c3151623b47": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "4f7ffd5dcff84e488f6f0e4454c4e2b2": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "810d447752de45e7b37c738d4ede3b1f": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_7b81e6acbaf44b09b107dc0d26736d6d",
              "IPY_MODEL_1f231fa4fa1340eb86d8b583b255d9e5",
              "IPY_MODEL_47369549ed1a47c594d488775804729c"
            ],
            "layout": "IPY_MODEL_a60f9653c2784753aec0fc5009b8afc9"
          }
        },
        "7b81e6acbaf44b09b107dc0d26736d6d": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_c52603e271064c54b05bf57c89008922",
            "placeholder": "​",
            "style": "IPY_MODEL_26ae9a3453064c15a11d48b37af22a45",
            "value": ""
          }
        },
        "1f231fa4fa1340eb86d8b583b255d9e5": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "info",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_c480746b3bf34dd8ba1c95953f589fa5",
            "max": 1,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_9406360cf2db4989984238904ab559b7",
            "value": 1
          }
        },
        "47369549ed1a47c594d488775804729c": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_64e5965695674bfbbe4674a4ac005ec4",
            "placeholder": "​",
            "style": "IPY_MODEL_a37e20c274064bb5a10c552df35f5be3",
            "value": " 960/0 [00:00&lt;00:00, 5242.85 examples/s]"
          }
        },
        "a60f9653c2784753aec0fc5009b8afc9": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "c52603e271064c54b05bf57c89008922": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "26ae9a3453064c15a11d48b37af22a45": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "c480746b3bf34dd8ba1c95953f589fa5": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": "20px"
          }
        },
        "9406360cf2db4989984238904ab559b7": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "64e5965695674bfbbe4674a4ac005ec4": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "a37e20c274064bb5a10c552df35f5be3": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "6994f7d8daa343ddbaf5851a9ced3df1": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_98bd9736d6394da6b7e46c4f55909bd8",
              "IPY_MODEL_e36086046d5a4e219bdd425bce54ff29",
              "IPY_MODEL_51c3a907593c4aa49662f182c1b4bf52"
            ],
            "layout": "IPY_MODEL_e7ca07e0313f4406997ea5bc18f3973e"
          }
        },
        "98bd9736d6394da6b7e46c4f55909bd8": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_5f0c73c3d5af4019b6b472bb2982d84b",
            "placeholder": "​",
            "style": "IPY_MODEL_e676b208c45e411d8a16e2bc02e8ffad",
            "value": "100%"
          }
        },
        "e36086046d5a4e219bdd425bce54ff29": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "danger",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_6b64cdee5a5f47608b05cbc02122cb52",
            "max": 1193,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_60996775c77946349b1166409c919d0a",
            "value": 1192
          }
        },
        "51c3a907593c4aa49662f182c1b4bf52": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_9aa782d2025d48fcb021c9ace1a03c63",
            "placeholder": "​",
            "style": "IPY_MODEL_b89ca59ad55143c3969a67831bd185df",
            "value": " 1192/1193 [00:00&lt;00:00, 29636.80 examples/s]"
          }
        },
        "e7ca07e0313f4406997ea5bc18f3973e": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "5f0c73c3d5af4019b6b472bb2982d84b": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "e676b208c45e411d8a16e2bc02e8ffad": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "6b64cdee5a5f47608b05cbc02122cb52": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "60996775c77946349b1166409c919d0a": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "9aa782d2025d48fcb021c9ace1a03c63": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "b89ca59ad55143c3969a67831bd185df": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "7bef511296fa410aa0e6a07245f25ed6": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_b6a94e11ba2144f38debd9770cfadde8",
              "IPY_MODEL_1f7f6676fb3b4b54bfbe9cc8538bee70",
              "IPY_MODEL_ac905feea82145de86f0e597be501d66"
            ],
            "layout": "IPY_MODEL_d5a3a1bc852345c08e3474e99562e757"
          }
        },
        "b6a94e11ba2144f38debd9770cfadde8": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_19652f29a18c454981642410a2c2b5d4",
            "placeholder": "​",
            "style": "IPY_MODEL_af008c0d4d10460cbfe0d4ed7bcccfee",
            "value": ""
          }
        },
        "1f7f6676fb3b4b54bfbe9cc8538bee70": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "info",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_72ea9c25c2464a629126d5af5dd0a8b5",
            "max": 1,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_c32d85b0e0ab4c55a8a5a3460f4bf1a9",
            "value": 1
          }
        },
        "ac905feea82145de86f0e597be501d66": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_fd6ff43cf8d0470ea24c4a857c08ec95",
            "placeholder": "​",
            "style": "IPY_MODEL_c1caa53280f54bdea896b2f4f18506fb",
            "value": " 1510/0 [00:00&lt;00:00, 5595.61 examples/s]"
          }
        },
        "d5a3a1bc852345c08e3474e99562e757": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "19652f29a18c454981642410a2c2b5d4": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "af008c0d4d10460cbfe0d4ed7bcccfee": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "72ea9c25c2464a629126d5af5dd0a8b5": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": "20px"
          }
        },
        "c32d85b0e0ab4c55a8a5a3460f4bf1a9": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "fd6ff43cf8d0470ea24c4a857c08ec95": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "c1caa53280f54bdea896b2f4f18506fb": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "93aff741c413426388746f1a257f7a29": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_1709a3eed7ac4aef8f16e418e1777556",
              "IPY_MODEL_30900a51eca546c7b7a8ca1df0b0a58d",
              "IPY_MODEL_d84d59141bbd4df19c3f4f3f839a61f9"
            ],
            "layout": "IPY_MODEL_735c0805ef3745fe945cfff508bcccdc"
          }
        },
        "1709a3eed7ac4aef8f16e418e1777556": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_458dd41e9a3143bf900775cda4249d3b",
            "placeholder": "​",
            "style": "IPY_MODEL_e18ac44e51a04616afb8d7a9184e7672",
            "value": "100%"
          }
        },
        "30900a51eca546c7b7a8ca1df0b0a58d": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "danger",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_1d3e01f99b824ab8829218168b38582d",
            "max": 1803,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_d0493d32818f441c9370372cd907c7e8",
            "value": 1802
          }
        },
        "d84d59141bbd4df19c3f4f3f839a61f9": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_6a76b2af5bdf492594c626bcb4720ee6",
            "placeholder": "​",
            "style": "IPY_MODEL_d05c5cd7177f4853a9cf2b297a325b6d",
            "value": " 1802/1803 [00:00&lt;00:00, 45226.88 examples/s]"
          }
        },
        "735c0805ef3745fe945cfff508bcccdc": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "458dd41e9a3143bf900775cda4249d3b": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "e18ac44e51a04616afb8d7a9184e7672": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "1d3e01f99b824ab8829218168b38582d": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "d0493d32818f441c9370372cd907c7e8": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "6a76b2af5bdf492594c626bcb4720ee6": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "d05c5cd7177f4853a9cf2b297a325b6d": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        }
      }
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/mathias-driesse/climatehack/blob/main/Transformers_example.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "863DJQOCBXlA",
        "outputId": "b6e741c9-8891-4dae-b464-3758e53e6b3b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: tensorflow_datasets in /usr/local/lib/python3.7/dist-packages (4.0.1)\n",
            "Requirement already satisfied: dill in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (0.3.5.1)\n",
            "Requirement already satisfied: absl-py in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (1.1.0)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (1.15.0)\n",
            "Requirement already satisfied: importlib-resources in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (5.7.1)\n",
            "Requirement already satisfied: promise in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (2.3)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (4.64.0)\n",
            "Requirement already satisfied: termcolor in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (1.1.0)\n",
            "Requirement already satisfied: attrs>=18.1.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (21.4.0)\n",
            "Requirement already satisfied: dm-tree in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (0.1.7)\n",
            "Requirement already satisfied: requests>=2.19.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (2.23.0)\n",
            "Requirement already satisfied: future in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (0.16.0)\n",
            "Requirement already satisfied: protobuf>=3.6.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (3.17.3)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (1.21.6)\n",
            "Requirement already satisfied: tensorflow-metadata in /usr/local/lib/python3.7/dist-packages (from tensorflow_datasets) (1.8.0)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->tensorflow_datasets) (2.10)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->tensorflow_datasets) (3.0.4)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->tensorflow_datasets) (2022.6.15)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->tensorflow_datasets) (1.24.3)\n",
            "Requirement already satisfied: zipp>=3.1.0 in /usr/local/lib/python3.7/dist-packages (from importlib-resources->tensorflow_datasets) (3.8.0)\n",
            "Requirement already satisfied: googleapis-common-protos<2,>=1.52.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow-metadata->tensorflow_datasets) (1.56.2)\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting tensorflow-text==2.8.*\n",
            "  Downloading tensorflow_text-2.8.2-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (4.9 MB)\n",
            "\u001b[K     |████████████████████████████████| 4.9 MB 7.2 MB/s \n",
            "\u001b[?25hRequirement already satisfied: tensorflow-hub>=0.8.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow-text==2.8.*) (0.12.0)\n",
            "Requirement already satisfied: tensorflow<2.9,>=2.8.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow-text==2.8.*) (2.8.2+zzzcolab20220527125636)\n",
            "Requirement already satisfied: numpy>=1.20 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.21.6)\n",
            "Requirement already satisfied: libclang>=9.0.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (14.0.1)\n",
            "Requirement already satisfied: absl-py>=0.4.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.1.0)\n",
            "Requirement already satisfied: keras-preprocessing>=1.1.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.1.2)\n",
            "Requirement already satisfied: tensorboard<2.9,>=2.8 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (2.8.0)\n",
            "Requirement already satisfied: grpcio<2.0,>=1.24.3 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.46.3)\n",
            "Requirement already satisfied: tensorflow-estimator<2.9,>=2.8 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (2.8.0)\n",
            "Requirement already satisfied: google-pasta>=0.1.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (0.2.0)\n",
            "Requirement already satisfied: tensorflow-io-gcs-filesystem>=0.23.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (0.26.0)\n",
            "Requirement already satisfied: opt-einsum>=2.3.2 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (3.3.0)\n",
            "Requirement already satisfied: keras<2.9,>=2.8.0rc0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (2.8.0)\n",
            "Requirement already satisfied: six>=1.12.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.15.0)\n",
            "Requirement already satisfied: termcolor>=1.1.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.1.0)\n",
            "Requirement already satisfied: protobuf<3.20,>=3.9.2 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (3.17.3)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (57.4.0)\n",
            "Requirement already satisfied: h5py>=2.9.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (3.1.0)\n",
            "Requirement already satisfied: flatbuffers>=1.12 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (2.0)\n",
            "Requirement already satisfied: typing-extensions>=3.6.6 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (4.1.1)\n",
            "Requirement already satisfied: wrapt>=1.11.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.14.1)\n",
            "Requirement already satisfied: gast>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (0.5.3)\n",
            "Requirement already satisfied: astunparse>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.6.3)\n",
            "Requirement already satisfied: wheel<1.0,>=0.23.0 in /usr/local/lib/python3.7/dist-packages (from astunparse>=1.6.0->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (0.37.1)\n",
            "Requirement already satisfied: cached-property in /usr/local/lib/python3.7/dist-packages (from h5py>=2.9.0->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.5.2)\n",
            "Requirement already satisfied: google-auth<3,>=1.6.3 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.35.0)\n",
            "Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (0.6.1)\n",
            "Requirement already satisfied: werkzeug>=0.11.15 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.0.1)\n",
            "Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.8.1)\n",
            "Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (3.3.7)\n",
            "Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (0.4.6)\n",
            "Requirement already satisfied: requests<3,>=2.21.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (2.23.0)\n",
            "Requirement already satisfied: rsa<5,>=3.1.4 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (4.8)\n",
            "Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (0.2.8)\n",
            "Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (4.2.4)\n",
            "Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.7/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.3.1)\n",
            "Requirement already satisfied: importlib-metadata>=4.4 in /usr/local/lib/python3.7/dist-packages (from markdown>=2.6.8->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (4.11.4)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata>=4.4->markdown>=2.6.8->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (3.8.0)\n",
            "Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /usr/local/lib/python3.7/dist-packages (from pyasn1-modules>=0.2.1->google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (0.4.8)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (3.0.4)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (1.24.3)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (2.10)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (2022.6.15)\n",
            "Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.9,>=2.8->tensorflow<2.9,>=2.8.0->tensorflow-text==2.8.*) (3.2.0)\n",
            "Installing collected packages: tensorflow-text\n",
            "Successfully installed tensorflow-text-2.8.2\n"
          ]
        }
      ],
      "source": [
        "!pip install tensorflow_datasets\n",
        "!pip install -U 'tensorflow-text==2.8.*'"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import logging\n",
        "import time\n",
        "\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "import tensorflow_datasets as tfds\n",
        "import tensorflow as tf\n",
        "\n",
        "import tensorflow_text"
      ],
      "metadata": {
        "id": "RXno0MrjBo61"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "examples, metadata = tfds.load('ted_hrlr_translate/pt_to_en', with_info = True,\n",
        "                               as_supervised = True)\n",
        "\n",
        "train_examples, val_examples = examples['train'], examples['validation']"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 368,
          "referenced_widgets": [
            "a47a6b8bd0974681ac9d39203fa26915",
            "8da628d2070d4308ae7ba20a11dfabf8",
            "3ac81a77bcdf4da194c5ac4409770f1b",
            "9fc1c7b64309485eaa411813e8dc99b0",
            "8c5a9a47cdff4b6186611459a68c7d78",
            "9321160bf2bc40edb65286baabbe7dc7",
            "0ac3934799ff48cc818343e069227f9b",
            "5eda2751d7e14e19a6ac8796f0177565",
            "33df1cb7900b4ed6a6e512c583af22de",
            "245a4aece6e04658b5d18ecb1fa36ba3",
            "22aabd693ca14aaaba90aef01bc3e1d6",
            "62b35e2df7984aa1bba29248380a2621",
            "96832a6965c14547b565a5be398452f3",
            "78c93918ce61423bae61f7d3c1e02536",
            "1a4af1a28e234d0994ca84ae05d3e6ef",
            "fb56044c2a7a40d897a55ba2f9653c57",
            "992e312ebf7949fe9c2706e511ed0e12",
            "13764c7c8405465388abee5e636b5d31",
            "0e649c0ba0924429ad5e0a8e931fab8d",
            "17894bdcedbb44769e6cb69adc3c5dcf",
            "e6b4fec1e6534d129a46cb094d3e7970",
            "afdbbb29e9bf45cc9d6b242a811d43f2",
            "9d3628e72de04db2afd55ed758767d67",
            "7d6588165503434088be28d5b777bb7a",
            "bc26fb93b3d34c26ab10b94035f81a86",
            "349f278f06fd479187716730e564018a",
            "eb8a7ccf3adf46f8991ca05538bf52a6",
            "86038ea1a53d47dba5adafe981c8a3b4",
            "a50f78f5c8254bdda191d4ad2172008a",
            "9edc499924a1459cbdbf5a6d2ec6f27c",
            "5a20956c30e04a95aadc3adfac751939",
            "39fcad7c783b4aab93462904d07e7e85",
            "ad09641fab04493d8c0ede403fb51a72",
            "5b118b411dd942c28a76b683ed9f5be4",
            "76ef8e5413144e339cb6771ad8a456d8",
            "e5313446e0734de199d5832bdb72b97a",
            "88b8da8515894553adc75bed9d89099a",
            "60c40ef411164b01801427e5b5b2d331",
            "a8bc7ed49ec545e6a41450e7ef256048",
            "2d6ef9ca68784e5ca691c61888f507e2",
            "6a9577849a2a4b33bc9d02e0c59647c4",
            "5dbcb775249a4e5f9e256b667ae6bd27",
            "3081ecfea6d54a24934c40b612e35869",
            "9e22673d0463498999a6cd587a6dd84e",
            "549857ca63b1406d9007f463edc36480",
            "8e3664cc2b9f400487ebf1aaf85ddaf4",
            "6741f6e4389a44b4a4b367d2259f5174",
            "bc732c8af9ee41b98cb6e0b4ff315f94",
            "03166b1c7dfa4ec5b09f4014c8cfd8f5",
            "dc16367ad4e54c909125227273225cb9",
            "3f7ff04a06c642f3b62fef2c3e7a009c",
            "237561ec12b4470cbf487c5429b2419f",
            "a1a65f2a42d54328ba329c2f11516516",
            "03d33320ff314c6b9bda5c3151623b47",
            "4f7ffd5dcff84e488f6f0e4454c4e2b2",
            "810d447752de45e7b37c738d4ede3b1f",
            "7b81e6acbaf44b09b107dc0d26736d6d",
            "1f231fa4fa1340eb86d8b583b255d9e5",
            "47369549ed1a47c594d488775804729c",
            "a60f9653c2784753aec0fc5009b8afc9",
            "c52603e271064c54b05bf57c89008922",
            "26ae9a3453064c15a11d48b37af22a45",
            "c480746b3bf34dd8ba1c95953f589fa5",
            "9406360cf2db4989984238904ab559b7",
            "64e5965695674bfbbe4674a4ac005ec4",
            "a37e20c274064bb5a10c552df35f5be3",
            "6994f7d8daa343ddbaf5851a9ced3df1",
            "98bd9736d6394da6b7e46c4f55909bd8",
            "e36086046d5a4e219bdd425bce54ff29",
            "51c3a907593c4aa49662f182c1b4bf52",
            "e7ca07e0313f4406997ea5bc18f3973e",
            "5f0c73c3d5af4019b6b472bb2982d84b",
            "e676b208c45e411d8a16e2bc02e8ffad",
            "6b64cdee5a5f47608b05cbc02122cb52",
            "60996775c77946349b1166409c919d0a",
            "9aa782d2025d48fcb021c9ace1a03c63",
            "b89ca59ad55143c3969a67831bd185df",
            "7bef511296fa410aa0e6a07245f25ed6",
            "b6a94e11ba2144f38debd9770cfadde8",
            "1f7f6676fb3b4b54bfbe9cc8538bee70",
            "ac905feea82145de86f0e597be501d66",
            "d5a3a1bc852345c08e3474e99562e757",
            "19652f29a18c454981642410a2c2b5d4",
            "af008c0d4d10460cbfe0d4ed7bcccfee",
            "72ea9c25c2464a629126d5af5dd0a8b5",
            "c32d85b0e0ab4c55a8a5a3460f4bf1a9",
            "fd6ff43cf8d0470ea24c4a857c08ec95",
            "c1caa53280f54bdea896b2f4f18506fb",
            "93aff741c413426388746f1a257f7a29",
            "1709a3eed7ac4aef8f16e418e1777556",
            "30900a51eca546c7b7a8ca1df0b0a58d",
            "d84d59141bbd4df19c3f4f3f839a61f9",
            "735c0805ef3745fe945cfff508bcccdc",
            "458dd41e9a3143bf900775cda4249d3b",
            "e18ac44e51a04616afb8d7a9184e7672",
            "1d3e01f99b824ab8829218168b38582d",
            "d0493d32818f441c9370372cd907c7e8",
            "6a76b2af5bdf492594c626bcb4720ee6",
            "d05c5cd7177f4853a9cf2b297a325b6d"
          ]
        },
        "id": "6fRwzfPbB7JO",
        "outputId": "ae23a99b-d4b6-4777-f04e-e7e961eaf6cd"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1mDownloading and preparing dataset ted_hrlr_translate/pt_to_en/1.0.0 (download: 124.94 MiB, generated: Unknown size, total: 124.94 MiB) to /root/tensorflow_datasets/ted_hrlr_translate/pt_to_en/1.0.0...\u001b[0m\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Dl Completed...: 0 url [00:00, ? url/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "a47a6b8bd0974681ac9d39203fa26915"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Dl Size...: 0 MiB [00:00, ? MiB/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "62b35e2df7984aa1bba29248380a2621"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Extraction completed...: 0 file [00:00, ? file/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "9d3628e72de04db2afd55ed758767d67"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\n",
            "\n",
            "\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "0 examples [00:00, ? examples/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "5b118b411dd942c28a76b683ed9f5be4"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Shuffling and writing examples to /root/tensorflow_datasets/ted_hrlr_translate/pt_to_en/1.0.0.incompleteIS3KG6/ted_hrlr_translate-train.tfrecord\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/51785 [00:00<?, ? examples/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "549857ca63b1406d9007f463edc36480"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "0 examples [00:00, ? examples/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "810d447752de45e7b37c738d4ede3b1f"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Shuffling and writing examples to /root/tensorflow_datasets/ted_hrlr_translate/pt_to_en/1.0.0.incompleteIS3KG6/ted_hrlr_translate-validation.tfrecord\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/1193 [00:00<?, ? examples/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "6994f7d8daa343ddbaf5851a9ced3df1"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "0 examples [00:00, ? examples/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "7bef511296fa410aa0e6a07245f25ed6"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Shuffling and writing examples to /root/tensorflow_datasets/ted_hrlr_translate/pt_to_en/1.0.0.incompleteIS3KG6/ted_hrlr_translate-test.tfrecord\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/1803 [00:00<?, ? examples/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "93aff741c413426388746f1a257f7a29"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1mDataset ted_hrlr_translate downloaded and prepared to /root/tensorflow_datasets/ted_hrlr_translate/pt_to_en/1.0.0. Subsequent calls will reuse this data.\u001b[0m\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "for pt_examples, en_examples in train_examples.batch(4).take(1):\n",
        "    for pt in pt_examples.numpy():\n",
        "        print(pt.decode('utf-8'))\n",
        "    print()\n",
        "    for en in en_examples.numpy():\n",
        "        print(en.decode('utf-8'))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QiRQetMcC98D",
        "outputId": "0d34252a-9f6e-4c2e-8285-8dc6615e6069"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "e quando melhoramos a procura , tiramos a única vantagem da impressão , que é a serendipidade .\n",
            "mas e se estes fatores fossem ativos ?\n",
            "mas eles não tinham a curiosidade de me testar .\n",
            "e esta rebeldia consciente é a razão pela qual eu , como agnóstica , posso ainda ter fé .\n",
            "\n",
            "and when you improve searchability , you actually take away the one advantage of print , which is serendipity .\n",
            "but what if it were active ?\n",
            "but they did n't test for curiosity .\n",
            "and this conscious defiance is why i , as an agnostic , can still have faith .\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model_name = 'ted_hrlr_translate_pt_en_converter'\n",
        "tf.keras.utils.get_file(\n",
        "    f'{model_name}.zip',\n",
        "    f'https://storage.googleapis.com/download.tensorflow.org/models/{model_name}.zip',\n",
        "    cache_dir='.', cache_subdir='', extract=True\n",
        ")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 107
        },
        "id": "xa3Z_t-bDWBn",
        "outputId": "a3df844e-02fc-4df8-b8ee-35d393326ef6"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/download.tensorflow.org/models/ted_hrlr_translate_pt_en_converter.zip\n",
            "188416/184801 [==============================] - 0s 0us/step\n",
            "196608/184801 [===============================] - 0s 0us/step\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'./ted_hrlr_translate_pt_en_converter.zip'"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            }
          },
          "metadata": {},
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "tokenizers = tf.saved_model.load(model_name)"
      ],
      "metadata": {
        "id": "MtzLjt7ODkbt"
      },
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "[item for item in dir(tokenizers.en) if not item.startswith('_')]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rkCuutNHDnhK",
        "outputId": "573b7194-81e0-479d-8e54-a3d13fd31242"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['detokenize',\n",
              " 'get_reserved_tokens',\n",
              " 'get_vocab_path',\n",
              " 'get_vocab_size',\n",
              " 'lookup',\n",
              " 'tokenize',\n",
              " 'tokenizer',\n",
              " 'vocab']"
            ]
          },
          "metadata": {},
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "for en in en_examples.numpy():\n",
        "    print(en.decode('utf-8'))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fwE_FzmGDsXl",
        "outputId": "5e539f97-51a8-4be9-fcbd-b96d40d45d9c"
      },
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "and when you improve searchability , you actually take away the one advantage of print , which is serendipity .\n",
            "but what if it were active ?\n",
            "but they did n't test for curiosity .\n",
            "and this conscious defiance is why i , as an agnostic , can still have faith .\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "encoded = tokenizers.en.tokenize(en_examples)\n",
        "\n",
        "for row in encoded.to_list():\n",
        "    print(row)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iatu3wcrDyd5",
        "outputId": "062b7b62-33d0-41c0-aef8-0fd8bb5d4df1"
      },
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[2, 72, 117, 79, 1259, 1491, 2362, 13, 79, 150, 184, 311, 71, 103, 2308, 74, 2679, 13, 148, 80, 55, 4840, 1434, 2423, 540, 15, 3]\n",
            "[2, 87, 90, 107, 76, 129, 1852, 30, 3]\n",
            "[2, 87, 83, 149, 50, 9, 56, 664, 85, 2512, 15, 3]\n",
            "[2, 72, 81, 2508, 2159, 3072, 1282, 80, 192, 45, 13, 100, 111, 6040, 3176, 3186, 13, 94, 235, 89, 1938, 15, 3]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "round_trip = tokenizers.en.detokenize(encoded)\n",
        "for line in round_trip.numpy():\n",
        "    print(line.decode('utf-8'))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AVFLEuHfD661",
        "outputId": "14cd24c4-d6f5-4ea2-89f3-d5d806aecbd4"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "and when you improve searchability , you actually take away the one advantage of print , which is serendipity .\n",
            "but what if it were active ?\n",
            "but they did n ' t test for curiosity .\n",
            "and this conscious defiance is why i , as an agnostic , can still have faith .\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "tokens = tokenizers.en.lookup(encoded)\n",
        "tokens"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "S4H7IzKlEA1i",
        "outputId": "d5dc50a8-484a-4ffa-ef35-820f392f60c1"
      },
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tf.RaggedTensor [[b'[START]', b'and', b'when', b'you', b'improve', b'search', b'##ability',\n",
              "  b',', b'you', b'actually', b'take', b'away', b'the', b'one', b'advantage',\n",
              "  b'of', b'print', b',', b'which', b'is', b's', b'##ere', b'##nd', b'##ip',\n",
              "  b'##ity', b'.', b'[END]']                                                 ,\n",
              " [b'[START]', b'but', b'what', b'if', b'it', b'were', b'active', b'?',\n",
              "  b'[END]']                                                           ,\n",
              " [b'[START]', b'but', b'they', b'did', b'n', b\"'\", b't', b'test', b'for',\n",
              "  b'curiosity', b'.', b'[END]']                                          ,\n",
              " [b'[START]', b'and', b'this', b'conscious', b'de', b'##fi', b'##ance',\n",
              "  b'is', b'why', b'i', b',', b'as', b'an', b'ag', b'##no', b'##stic', b',',\n",
              "  b'can', b'still', b'have', b'faith', b'.', b'[END]']                     ]>"
            ]
          },
          "metadata": {},
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "lengths = []\n",
        "for pt_examples, en_examples in train_examples.batch(1024):\n",
        "    pt_tokens = tokenizers.en.tokenize(pt_examples)\n",
        "    lengths.append(pt_tokens.row_lengths())\n",
        "\n",
        "    en_tokens = tokenizers.en.tokenize(en_examples)\n",
        "    lengths.append(en_tokens.row_lengths())\n",
        "    print('.', end = '', flush = True)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9GlUQHr-EJsX",
        "outputId": "ca1c0ed8-2cd8-4fe1-9859-efa8d68ae10e"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "..................................................."
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "all_lengths = np.concatenate(lengths)\n",
        "\n",
        "plt.hist(all_lengths, np.linspace(0, 500, 101))\n",
        "plt.ylim(plt.ylim())\n",
        "max_length = max(all_lengths)\n",
        "plt.plot([max_length, max_length], plt.ylim())\n",
        "plt.title(f'Max tokens per example: {max_length}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 298
        },
        "id": "6gegpyW8uJ6A",
        "outputId": "e3fba60d-68c9-4575-8ce9-55fc562d9358"
      },
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0.5, 1.0, 'Max tokens per example: 394')"
            ]
          },
          "metadata": {},
          "execution_count": 13
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "MAX_TOKENS = 128"
      ],
      "metadata": {
        "id": "z73HYb1cuo3M"
      },
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def filter_max_tokens(pt, en):\n",
        "    num_tokens = tf.maximum(tf.shape(pt)[1], tf.shape(en)[1])\n",
        "    return num_tokens < MAX_TOKENS"
      ],
      "metadata": {
        "id": "rJxW7v_jusSI"
      },
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def tokenize_pairs(pt, en):\n",
        "    pt = tokenizers.pt.tokenize(pt)\n",
        "    pt = pt.to_tensor()\n",
        "\n",
        "    en = tokenizers.en.tokenize(en)\n",
        "    en = en.to_tensor()\n",
        "    return pt, en"
      ],
      "metadata": {
        "id": "fFxUzw8Su5TL"
      },
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "for pt_examples, en_examples in train_examples.batch(1024):\n",
        "    print(tokenize_pairs(pt_examples, en_examples))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LH0l1XmLvE-w",
        "outputId": "3c564035-560e-4031-b56b-0cffbd382c33"
      },
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(<tf.Tensor: shape=(1024, 146), dtype=int64, numpy=\n",
            "array([[   2,   44,  115, ...,    0,    0,    0],\n",
            "       [   2,   99,   44, ...,    0,    0,    0],\n",
            "       [   2,   99,  131, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  131,  109, ...,    0,    0,    0],\n",
            "       [   2,   40, 1037, ...,    0,    0,    0],\n",
            "       [   2,  168,   15, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 171), dtype=int64, numpy=\n",
            "array([[  2,  72, 117, ...,   0,   0,   0],\n",
            "       [  2,  87,  90, ...,   0,   0,   0],\n",
            "       [  2,  87,  83, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  83,  86, ...,   0,   0,   0],\n",
            "       [  2,  71, 261, ...,   0,   0,   0],\n",
            "       [  2,  79, 125, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 244), dtype=int64, numpy=\n",
            "array([[   2,   89,  341, ...,    0,    0,    0],\n",
            "       [   2,  316,   14, ...,    0,    0,    0],\n",
            "       [   2,   10, 6052, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   44,  610, ...,    0,    0,    0],\n",
            "       [   2,  115,   89, ...,    0,    0,    0],\n",
            "       [   2,   99,  125, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 226), dtype=int64, numpy=\n",
            "array([[   2,   77,  237, ...,    0,    0,    0],\n",
            "       [   2,  157,   13, ...,    0,    0,    0],\n",
            "       [   2,   10, 4522, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   72,   79, ...,    0,    0,    0],\n",
            "       [   2,  117,   79, ...,    0,    0,    0],\n",
            "       [   2,   87,   81, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 130), dtype=int64, numpy=\n",
            "array([[   2,   54,  300, ...,    0,    0,    0],\n",
            "       [   2,   99,  115, ...,    0,    0,    0],\n",
            "       [   2,   54, 1320, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  771,   84, ...,    0,    0,    0],\n",
            "       [   2,  120,   86, ...,    0,    0,    0],\n",
            "       [   2,   90,   28, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 127), dtype=int64, numpy=\n",
            "array([[   2,   71, 2871, ...,    0,    0,    0],\n",
            "       [   2,   87,  117, ...,    0,    0,    0],\n",
            "       [   2,   71, 1129, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   45,  321, ...,    0,    0,    0],\n",
            "       [   2,   96,    9, ...,    0,    0,    0],\n",
            "       [   2, 1815,   27, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 110), dtype=int64, numpy=\n",
            "array([[   2,   90,  244, ...,    0,    0,    0],\n",
            "       [   2,   44,   44, ...,    0,    0,    0],\n",
            "       [   2,   40, 1599, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   99,  133, ...,    0,    0,    0],\n",
            "       [   2,  121, 7538, ...,    0,    0,    0],\n",
            "       [   2,  404, 3508, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 112), dtype=int64, numpy=\n",
            "array([[  2,  76, 126, ...,   0,   0,   0],\n",
            "       [  2,  72,  76, ...,   0,   0,   0],\n",
            "       [  2,  76, 563, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  87, 110, ...,   0,   0,   0],\n",
            "       [  2,  76,  84, ...,   0,   0,   0],\n",
            "       [  2,  45,  94, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 193), dtype=int64, numpy=\n",
            "array([[   2,   99,  107, ...,    0,    0,    0],\n",
            "       [   2,  108,   88, ...,    0,    0,    0],\n",
            "       [   2,  141,   91, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   57, 1879, ...,    0,    0,    0],\n",
            "       [   2,  125,  121, ...,    0,    0,    0],\n",
            "       [   2,   39,   39, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 208), dtype=int64, numpy=\n",
            "array([[   2,   87,   75, ...,    0,    0,    0],\n",
            "       [   2,   81,   80, ...,    0,    0,    0],\n",
            "       [   2,  172, 5243, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   54, 5162, ...,    0,    0,    0],\n",
            "       [   2,   81,   84, ...,    0,    0,    0],\n",
            "       [   2,   36,   36, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 146), dtype=int64, numpy=\n",
            "array([[   2, 1719,   84, ...,    0,    0,    0],\n",
            "       [   2,   10,  165, ...,    0,    0,    0],\n",
            "       [   2,   40, 3147, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  247,  424, ...,    0,    0,    0],\n",
            "       [   2,  259,  231, ...,    0,    0,    0],\n",
            "       [   2,  104, 7051, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 109), dtype=int64, numpy=\n",
            "array([[   2,   87,  110, ...,    0,    0,    0],\n",
            "       [   2,   10,  177, ...,    0,    0,    0],\n",
            "       [   2,  115, 4225, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  180,   74, ...,    0,    0,    0],\n",
            "       [   2,  110,  245, ...,    0,    0,    0],\n",
            "       [   2,   45,    9, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 160), dtype=int64, numpy=\n",
            "array([[   2,   40,  236, ...,    0,    0,    0],\n",
            "       [   2,   21,   16, ...,    0,    0,    0],\n",
            "       [   2,  160,   14, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   40,  138, ...,    0,    0,    0],\n",
            "       [   2,   99, 1906, ...,    0,    0,    0],\n",
            "       [   2,   99,   92, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 156), dtype=int64, numpy=\n",
            "array([[   2,   71,  236, ...,    0,    0,    0],\n",
            "       [   2,  796, 2339, ...,    0,    0,    0],\n",
            "       [   2,   82,   85, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   99,  430, ...,    0,    0,    0],\n",
            "       [   2,   87,  420, ...,    0,    0,    0],\n",
            "       [   2,   87,  354, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 264), dtype=int64, numpy=\n",
            "array([[   2,   40, 1319, ...,    0,    0,    0],\n",
            "       [   2,  245, 4074, ...,    0,    0,    0],\n",
            "       [   2,  360,   15, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  103,  200, ...,    0,    0,    0],\n",
            "       [   2,   44,  115, ...,    0,    0,    0],\n",
            "       [   2,   89,   88, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 294), dtype=int64, numpy=\n",
            "array([[   2,   71, 1261, ...,    0,    0,    0],\n",
            "       [   2,   72,   81, ...,    0,    0,    0],\n",
            "       [   2,   83,    9, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   77,  237, ...,    0,    0,    0],\n",
            "       [   2,   72,  117, ...,    0,    0,    0],\n",
            "       [   2,  107,   83, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 142), dtype=int64, numpy=\n",
            "array([[  2, 108,  88, ...,   0,   0,   0],\n",
            "       [  2, 425, 774, ...,   0,   0,   0],\n",
            "       [  2, 140,  14, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  39,  39, ...,   0,   0,   0],\n",
            "       [  2,  39,  39, ...,   0,   0,   0],\n",
            "       [  2, 333,  84, ...,   0,   0,   0]])>, <tf.Tensor: shape=(1024, 145), dtype=int64, numpy=\n",
            "array([[  2,  81,  80, ...,   0,   0,   0],\n",
            "       [  2,  78, 158, ...,   0,   0,   0],\n",
            "       [  2, 157,  13, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2, 120,  71, ...,   0,   0,   0],\n",
            "       [  2, 169,  30, ...,   0,   0,   0],\n",
            "       [  2,  45, 990, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 147), dtype=int64, numpy=\n",
            "array([[   2,   44,   86, ...,    0,    0,    0],\n",
            "       [   2, 5286, 1361, ...,    0,    0,    0],\n",
            "       [   2,  129,  425, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   91, 2301, ...,    0,    0,    0],\n",
            "       [   2, 1546,  135, ...,    0,    0,    0],\n",
            "       [   2,  310,   14, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 146), dtype=int64, numpy=\n",
            "array([[  2,  81,  80, ...,   0,   0,   0],\n",
            "       [  2, 139,  72, ...,   0,   0,   0],\n",
            "       [  2,  78, 101, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  72,  71, ...,   0,   0,   0],\n",
            "       [  2, 352, 124, ...,   0,   0,   0],\n",
            "       [  2, 575,  13, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 126), dtype=int64, numpy=\n",
            "array([[   2,  183,  195, ...,    0,    0,    0],\n",
            "       [   2,  107,  449, ...,    0,    0,    0],\n",
            "       [   2,  546, 5446, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  669,   14, ...,    0,    0,    0],\n",
            "       [   2,  130, 5902, ...,    0,    0,    0],\n",
            "       [   2,   44,   85, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 111), dtype=int64, numpy=\n",
            "array([[   2,   36,   36, ...,    0,    0,    0],\n",
            "       [   2,   72,   75, ...,    0,    0,    0],\n",
            "       [   2, 2720,  565, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  238,   13, ...,    0,    0,    0],\n",
            "       [   2,   78,   89, ...,    0,    0,    0],\n",
            "       [   2,   81,   80, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 168), dtype=int64, numpy=\n",
            "array([[   2,   44, 6968, ...,    0,    0,    0],\n",
            "       [   2,  342,  102, ...,    0,    0,    0],\n",
            "       [   2, 2733,   14, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  145,  459, ...,    0,    0,    0],\n",
            "       [   2, 2474,  220, ...,    0,    0,    0],\n",
            "       [   2,   89, 1066, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 180), dtype=int64, numpy=\n",
            "array([[   2,   72,  146, ...,    0,    0,    0],\n",
            "       [   2,   71,  121, ...,    0,    0,    0],\n",
            "       [   2, 3310,   13, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   78,   94, ...,    0,    0,    0],\n",
            "       [   2,   95,  165, ...,    0,    0,    0],\n",
            "       [   2,  107,   79, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 152), dtype=int64, numpy=\n",
            "array([[   2,  158,  141, ...,    0,    0,    0],\n",
            "       [   2, 2194,   15, ...,    0,    0,    0],\n",
            "       [   2,  162,   54, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   54,   65, ...,    0,    0,    0],\n",
            "       [   2,  162,  163, ...,    0,    0,    0],\n",
            "       [   2,  101,  235, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 138), dtype=int64, numpy=\n",
            "array([[   2,  172,   85, ...,    0,    0,    0],\n",
            "       [   2, 1072,   73, ...,    0,    0,    0],\n",
            "       [   2,  101,   45, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   71,  969, ...,    0,    0,    0],\n",
            "       [   2,   76,   84, ...,    0,    0,    0],\n",
            "       [   2,   85,  251, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 179), dtype=int64, numpy=\n",
            "array([[   2,   54, 4489, ...,    0,    0,    0],\n",
            "       [   2,   99,  556, ...,    0,    0,    0],\n",
            "       [   2,   44,  132, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   54,   44, ...,    0,    0,    0],\n",
            "       [   2,  411, 2802, ...,    0,    0,    0],\n",
            "       [   2,   99,   89, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 150), dtype=int64, numpy=\n",
            "array([[   2, 3862,    9, ...,    0,    0,    0],\n",
            "       [   2,   87,   79, ...,    0,    0,    0],\n",
            "       [   2,   72,   45, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2, 4797,  150, ...,    0,    0,    0],\n",
            "       [   2,  154,  103, ...,    0,    0,    0],\n",
            "       [   2,   87,  107, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 156), dtype=int64, numpy=\n",
            "array([[   2,   10,  165, ...,    0,    0,    0],\n",
            "       [   2,  103,  286, ...,    0,    0,    0],\n",
            "       [   2,   99,  151, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  106,   89, ...,    0,    0,    0],\n",
            "       [   2, 1138,   83, ...,    0,    0,    0],\n",
            "       [   2,  249,   16, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 148), dtype=int64, numpy=\n",
            "array([[   2,   10,  177, ...,    0,    0,    0],\n",
            "       [   2, 6577,   88, ...,    0,    0,    0],\n",
            "       [   2,   87,  146, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   37,  199, ...,    0,    0,    0],\n",
            "       [   2,   37, 1745, ...,    0,    0,    0],\n",
            "       [   2,  202,   79, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 168), dtype=int64, numpy=\n",
            "array([[  2,  10, 389, ...,   0,   0,   0],\n",
            "       [  2, 101, 235, ...,   0,   0,   0],\n",
            "       [  2,  44,  86, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  91, 448, ...,   0,   0,   0],\n",
            "       [  2,  92, 209, ...,   0,   0,   0],\n",
            "       [  2,  92, 507, ...,   0,   0,   0]])>, <tf.Tensor: shape=(1024, 180), dtype=int64, numpy=\n",
            "array([[  2,  10, 371, ...,   0,   0,   0],\n",
            "       [  2,  85, 251, ...,   0,   0,   0],\n",
            "       [  2,  76,  80, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2, 696,  89, ...,   0,   0,   0],\n",
            "       [  2,  82, 488, ...,   0,   0,   0],\n",
            "       [  2,  71, 336, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 160), dtype=int64, numpy=\n",
            "array([[   2,   54,  136, ...,    0,    0,    0],\n",
            "       [   2,  425, 1399, ...,    0,    0,    0],\n",
            "       [   2,   10,  165, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   99,   54, ...,    0,    0,    0],\n",
            "       [   2,   44,   14, ...,    0,    0,    0],\n",
            "       [   2,   10,  165, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 139), dtype=int64, numpy=\n",
            "array([[  2,  71, 143, ...,   0,   0,   0],\n",
            "       [  2,  78, 158, ...,   0,   0,   0],\n",
            "       [  2,  10, 177, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  87,  71, ...,   0,   0,   0],\n",
            "       [  2,  72,  74, ...,   0,   0,   0],\n",
            "       [  2,  10, 177, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 121), dtype=int64, numpy=\n",
            "array([[  2, 122, 203, ...,   0,   0,   0],\n",
            "       [  2,  39,  39, ...,   0,   0,   0],\n",
            "       [  2, 121,  86, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2, 233, 313, ...,   0,   0,   0],\n",
            "       [  2,  39,  39, ...,   0,   0,   0],\n",
            "       [  2,  88, 282, ...,   0,   0,   0]])>, <tf.Tensor: shape=(1024, 109), dtype=int64, numpy=\n",
            "array([[   2,   82,   45, ...,    0,    0,    0],\n",
            "       [   2,   36,   36, ...,    0,    0,    0],\n",
            "       [   2,   75,    9, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  155,  287, ...,    0,    0,    0],\n",
            "       [   2, 2997,   88, ...,    0,    0,    0],\n",
            "       [   2,   45,    9, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 156), dtype=int64, numpy=\n",
            "array([[  2, 183,  61, ...,   0,   0,   0],\n",
            "       [  2, 108, 306, ...,   0,   0,   0],\n",
            "       [  2,  99,  88, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2, 199, 109, ...,   0,   0,   0],\n",
            "       [  2, 122,  14, ...,   0,   0,   0],\n",
            "       [  2,  39,  39, ...,   0,   0,   0]])>, <tf.Tensor: shape=(1024, 154), dtype=int64, numpy=\n",
            "array([[  2, 175,  58, ...,   0,   0,   0],\n",
            "       [  2,  82,  75, ...,   0,   0,   0],\n",
            "       [  2,  87,  13, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2, 116,  86, ...,   0,   0,   0],\n",
            "       [  2,  82,  45, ...,   0,   0,   0],\n",
            "       [  2,  36,  36, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 167), dtype=int64, numpy=\n",
            "array([[   2,   54, 1411, ...,    0,    0,    0],\n",
            "       [   2,  282,   85, ...,    0,    0,    0],\n",
            "       [   2,   10,   42, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2, 3421,   83, ...,    0,    0,    0],\n",
            "       [   2,  163, 1152, ...,    0,    0,    0],\n",
            "       [   2,  105,   44, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 170), dtype=int64, numpy=\n",
            "array([[   2, 1029,   84, ...,    0,    0,    0],\n",
            "       [   2,  125,   13, ...,    0,    0,    0],\n",
            "       [   2,   10,  189, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   78, 1057, ...,    0,    0,    0],\n",
            "       [   2,  175,   84, ...,    0,    0,    0],\n",
            "       [   2,   81,   80, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 116), dtype=int64, numpy=\n",
            "array([[   2,  100,  443, ...,    0,    0,    0],\n",
            "       [   2,   87,  208, ...,    0,    0,    0],\n",
            "       [   2,  703,   92, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   99,  115, ...,    0,    0,    0],\n",
            "       [   2,  363,  161, ...,    0,    0,    0],\n",
            "       [   2,   92, 6236, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 98), dtype=int64, numpy=\n",
            "array([[   2,  546,  122, ...,    0,    0,    0],\n",
            "       [   2,  107,   78, ...,    0,    0,    0],\n",
            "       [   2,   76, 5646, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   87,  117, ...,    0,    0,    0],\n",
            "       [   2,   81,  139, ...,    0,    0,    0],\n",
            "       [   2,   71, 2856, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 183), dtype=int64, numpy=\n",
            "array([[   2,   44,  225, ...,    0,    0,    0],\n",
            "       [   2,   44,  160, ...,    0,    0,    0],\n",
            "       [   2,   39,   39, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  342,  401, ...,    0,    0,    0],\n",
            "       [   2,  275,   83, ...,    0,    0,    0],\n",
            "       [   2,   99, 1719, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 215), dtype=int64, numpy=\n",
            "array([[  2,  72,  45, ...,   0,   0,   0],\n",
            "       [  2,  72,  82, ...,   0,   0,   0],\n",
            "       [  2,  72,  82, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2, 119, 170, ...,   0,   0,   0],\n",
            "       [  2,  78, 181, ...,   0,   0,   0],\n",
            "       [  2,  87, 732, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 186), dtype=int64, numpy=\n",
            "array([[   2,   10,  165, ...,    0,    0,    0],\n",
            "       [   2,  103, 2833, ...,    0,    0,    0],\n",
            "       [   2,   89, 5261, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2, 5537,   16, ...,    0,    0,    0],\n",
            "       [   2,   44,  151, ...,    0,    0,    0],\n",
            "       [   2,   40, 1614, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 172), dtype=int64, numpy=\n",
            "array([[   2,   10,  177, ...,    0,    0,    0],\n",
            "       [   2,   77,   71, ...,    0,    0,    0],\n",
            "       [   2,  107,   78, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  372,   13, ...,    0,    0,    0],\n",
            "       [   2,   72,  146, ...,    0,    0,    0],\n",
            "       [   2, 4715, 1628, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 120), dtype=int64, numpy=\n",
            "array([[   2,  125,   44, ...,    0,    0,    0],\n",
            "       [   2,  125,   44, ...,    0,    0,    0],\n",
            "       [   2,   39,   39, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   97,  437, ...,    0,    0,    0],\n",
            "       [   2,   90,  244, ...,    0,    0,    0],\n",
            "       [   2,   92, 1102, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 124), dtype=int64, numpy=\n",
            "array([[  2,  81,  80, ...,   0,   0,   0],\n",
            "       [  2,  81,  80, ...,   0,   0,   0],\n",
            "       [  2,  36,  36, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2, 119,  91, ...,   0,   0,   0],\n",
            "       [  2,  87, 168, ...,   0,   0,   0],\n",
            "       [  2,  72,  83, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 199), dtype=int64, numpy=\n",
            "array([[   2,   10, 2046, ...,    0,    0,    0],\n",
            "       [   2,  214,   84, ...,    0,    0,    0],\n",
            "       [   2,   86,  281, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  120,   84, ...,    0,    0,    0],\n",
            "       [   2,  105,   44, ...,    0,    0,    0],\n",
            "       [   2,   40, 1601, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 164), dtype=int64, numpy=\n",
            "array([[  2,  10, 932, ...,   0,   0,   0],\n",
            "       [  2,  72, 780, ...,   0,   0,   0],\n",
            "       [  2,  37, 174, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  82,  76, ...,   0,   0,   0],\n",
            "       [  2,  81,  80, ...,   0,   0,   0],\n",
            "       [  2, 118,  71, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 173), dtype=int64, numpy=\n",
            "array([[   2,   89,  296, ...,    0,    0,    0],\n",
            "       [   2,  140,   14, ...,    0,    0,    0],\n",
            "       [   2,  114,   88, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   97, 1279, ...,    0,    0,    0],\n",
            "       [   2,  411, 1627, ...,    0,    0,    0],\n",
            "       [   2,   44,   97, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 179), dtype=int64, numpy=\n",
            "array([[  2, 107, 461, ...,   0,   0,   0],\n",
            "       [  2, 157,  13, ...,   0,   0,   0],\n",
            "       [  2, 192,  98, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  78,  13, ...,   0,   0,   0],\n",
            "       [  2,  72, 917, ...,   0,   0,   0],\n",
            "       [  2, 120,  96, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 102), dtype=int64, numpy=\n",
            "array([[   2,  160,   14, ...,    0,    0,    0],\n",
            "       [   2, 5829,   85, ...,    0,    0,    0],\n",
            "       [   2, 4867,  746, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  122,  104, ...,    0,    0,    0],\n",
            "       [   2,   40,  169, ...,    0,    0,    0],\n",
            "       [   2,   39,   39, ...,    3,    0,    0]])>, <tf.Tensor: shape=(1024, 103), dtype=int64, numpy=\n",
            "array([[  2,  82, 107, ...,   0,   0,   0],\n",
            "       [  2, 260,  37, ...,   0,   0,   0],\n",
            "       [  2,  76, 563, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  82,  45, ...,   0,   0,   0],\n",
            "       [  2, 108, 467, ...,   0,   0,   0],\n",
            "       [  2,  36,  36, ...,   9,   9,   3]])>)\n",
            "(<tf.Tensor: shape=(1024, 151), dtype=int64, numpy=\n",
            "array([[   2,   99,   92, ...,    0,    0,    0],\n",
            "       [   2, 6386,   15, ...,    0,    0,    0],\n",
            "       [   2,   54,   84, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   44,   89, ...,    0,    0,    0],\n",
            "       [   2,  145, 2031, ...,    0,    0,    0],\n",
            "       [   2, 6425,   54, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 144), dtype=int64, numpy=\n",
            "array([[   2,   87, 5849, ...,    0,    0,    0],\n",
            "       [   2, 4997,   14, ...,    0,    0,    0],\n",
            "       [   2,   82,   90, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   72,  107, ...,    0,    0,    0],\n",
            "       [   2,   79,   94, ...,    0,    0,    0],\n",
            "       [   2,   45,  586, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 131), dtype=int64, numpy=\n",
            "array([[   2,  170,  136, ...,    0,    0,    0],\n",
            "       [   2,  101,  107, ...,    0,    0,    0],\n",
            "       [   2,  239,   89, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  110, 7205, ...,    0,    0,    0],\n",
            "       [   2,   10,  165, ...,    0,    0,    0],\n",
            "       [   2,   94,  325, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 119), dtype=int64, numpy=\n",
            "array([[   2,   77,   37, ...,    0,    0,    0],\n",
            "       [   2,   82,   78, ...,    0,    0,    0],\n",
            "       [   2,  623,  107, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  110,   13, ...,    0,    0,    0],\n",
            "       [   2,   10,  177, ...,    0,    0,    0],\n",
            "       [   2, 3531,   13, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 154), dtype=int64, numpy=\n",
            "array([[   2,   44,   44, ...,    0,    0,    0],\n",
            "       [   2,   54,  179, ...,    0,    0,    0],\n",
            "       [   2,  147,  967, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   44, 2695, ...,    0,    0,    0],\n",
            "       [   2,   85,  119, ...,    0,    0,    0],\n",
            "       [   2,  220, 4010, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 145), dtype=int64, numpy=\n",
            "array([[   2,   72,   96, ...,    0,    0,    0],\n",
            "       [   2,  185,  571, ...,    0,    0,    0],\n",
            "       [   2,   76,   94, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   72, 1615, ...,    0,    0,    0],\n",
            "       [   2,  110,  103, ...,    0,    0,    0],\n",
            "       [   2, 1244, 1650, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 184), dtype=int64, numpy=\n",
            "array([[   2,  490,   14, ...,    0,    0,    0],\n",
            "       [   2,   44,  453, ...,    0,    0,    0],\n",
            "       [   2,   88,  495, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  253, 2582, ...,    0,    0,    0],\n",
            "       [   2,  114,   44, ...,    0,    0,    0],\n",
            "       [   2,   44,   83, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 163), dtype=int64, numpy=\n",
            "array([[  2, 101, 169, ...,   0,   0,   0],\n",
            "       [  2,  72, 173, ...,   0,   0,   0],\n",
            "       [  2,  78, 145, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  82, 349, ...,   0,   0,   0],\n",
            "       [  2, 120,  81, ...,   0,   0,   0],\n",
            "       [  2,  72, 273, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 168), dtype=int64, numpy=\n",
            "array([[   2,  122,  502, ...,    0,    0,    0],\n",
            "       [   2,   39,   39, ...,    0,    0,    0],\n",
            "       [   2,   44,   14, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  122,   14, ...,    0,    0,    0],\n",
            "       [   2,  776, 2223, ...,    0,    0,    0],\n",
            "       [   2,   84,   98, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 173), dtype=int64, numpy=\n",
            "array([[  2,  82,  78, ...,   0,   0,   0],\n",
            "       [  2,  36,  36, ...,   0,   0,   0],\n",
            "       [  2,  72,  77, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  72,  82, ...,   0,   0,   0],\n",
            "       [  2, 655, 701, ...,   0,   0,   0],\n",
            "       [  2,  75, 161, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 166), dtype=int64, numpy=\n",
            "array([[   2,   99,  105, ...,    0,    0,    0],\n",
            "       [   2,   54,  185, ...,    0,    0,    0],\n",
            "       [   2,   39,   39, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   39,   39, ...,    0,    0,    0],\n",
            "       [   2,  131,  990, ...,    0,    0,    0],\n",
            "       [   2,   44, 2120, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 133), dtype=int64, numpy=\n",
            "array([[  2,  87, 115, ...,   0,   0,   0],\n",
            "       [  2, 135, 719, ...,   0,   0,   0],\n",
            "       [  2,  36,  36, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  77, 237, ...,   0,   0,   0],\n",
            "       [  2,  83, 184, ...,   0,   0,   0],\n",
            "       [  2,  76,   9, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 160), dtype=int64, numpy=\n",
            "array([[   2,   40,  650, ...,    0,    0,    0],\n",
            "       [   2,   44,  728, ...,    0,    0,    0],\n",
            "       [   2,   54, 1663, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  131,  137, ...,    0,    0,    0],\n",
            "       [   2,  100,  185, ...,    0,    0,    0],\n",
            "       [   2,  234,   40, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 158), dtype=int64, numpy=\n",
            "array([[  2,  72,  71, ...,   0,   0,   0],\n",
            "       [  2,  41, 787, ...,   0,   0,   0],\n",
            "       [  2,  71,  82, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  83,   9, ...,   0,   0,   0],\n",
            "       [  2,  77, 108, ...,   0,   0,   0],\n",
            "       [  2, 506,   9, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 140), dtype=int64, numpy=\n",
            "array([[   2,   44,  199, ...,    0,    0,    0],\n",
            "       [   2,   10,  178, ...,    0,    0,    0],\n",
            "       [   2,  228,  151, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  131,  290, ...,    0,    0,    0],\n",
            "       [   2,  245, 1863, ...,    0,    0,    0],\n",
            "       [   2,  777,   86, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 136), dtype=int64, numpy=\n",
            "array([[   2,   72,  146, ...,    0,    0,    0],\n",
            "       [   2,   10,  189, ...,    0,    0,    0],\n",
            "       [   2, 5427,  221, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   83,  129, ...,    0,    0,    0],\n",
            "       [   2,   72,   75, ...,    0,    0,    0],\n",
            "       [   2,   37,  174, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 104), dtype=int64, numpy=\n",
            "array([[   2,  170, 1658, ...,    0,    0,    0],\n",
            "       [   2,  105,  121, ...,    0,    0,    0],\n",
            "       [   2,   39,   39, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  122, 5286, ...,    0,    0,    0],\n",
            "       [   2,   44,   97, ...,    0,    0,    0],\n",
            "       [   2,  191,  651, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 115), dtype=int64, numpy=\n",
            "array([[  2,  72, 102, ...,   0,   0,   0],\n",
            "       [  2,  81,  84, ...,   0,   0,   0],\n",
            "       [  2,  87, 192, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  82,  45, ...,   0,   0,   0],\n",
            "       [  2,  72, 100, ...,   0,   0,   0],\n",
            "       [  2,  82, 186, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 148), dtype=int64, numpy=\n",
            "array([[   2,  249,   16, ...,    0,    0,    0],\n",
            "       [   2,  191,  139, ...,    0,    0,    0],\n",
            "       [   2,  132, 1499, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   54,  142, ...,    0,    0,    0],\n",
            "       [   2,   56, 2063, ...,    0,    0,    0],\n",
            "       [   2,  104,  225, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 132), dtype=int64, numpy=\n",
            "array([[   2,  202,   79, ...,    0,    0,    0],\n",
            "       [   2,  186,    9, ...,    0,    0,    0],\n",
            "       [   2,   78,  166, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   99,  182, ...,    0,    0,    0],\n",
            "       [   2, 6798,   15, ...,    0,    0,    0],\n",
            "       [   2,  110,   13, ...,    3,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 189), dtype=int64, numpy=\n",
            "array([[  2,  39,  39, ...,   0,   0,   0],\n",
            "       [  2, 568,  14, ...,   0,   0,   0],\n",
            "       [  2, 122,  14, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  99, 151, ...,   0,   0,   0],\n",
            "       [  2,  10, 178, ...,   0,   0,   0],\n",
            "       [  2,  44, 106, ...,   0,   0,   0]])>, <tf.Tensor: shape=(1024, 177), dtype=int64, numpy=\n",
            "array([[  2,  79, 112, ...,   0,   0,   0],\n",
            "       [  2, 112, 327, ...,   0,   0,   0],\n",
            "       [  2,  82, 165, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  87,  45, ...,   0,   0,   0],\n",
            "       [  2,  10, 189, ...,   0,   0,   0],\n",
            "       [  2,  72,  81, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 160), dtype=int64, numpy=\n",
            "array([[   2,   44, 2263, ...,    0,    0,    0],\n",
            "       [   2,   39,   39, ...,    0,    0,    0],\n",
            "       [   2,   44, 6720, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   10, 4294, ...,    0,    0,    0],\n",
            "       [   2,  108,   44, ...,    0,    0,    0],\n",
            "       [   2,  105,   98, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 157), dtype=int64, numpy=\n",
            "array([[   2,   76,    9, ...,    0,    0,    0],\n",
            "       [   2,   45, 6323, ...,    0,    0,    0],\n",
            "       [   2,   72,   78, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   10,  177, ...,    0,    0,    0],\n",
            "       [   2,   82,   75, ...,    0,    0,    0],\n",
            "       [   2,   83,    9, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 195), dtype=int64, numpy=\n",
            "array([[   2,   41, 5389, ...,    0,    0,    0],\n",
            "       [   2,   88,   44, ...,    0,    0,    0],\n",
            "       [   2,   39,   39, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  120,   85, ...,    0,    0,    0],\n",
            "       [   2,  631,   95, ...,    0,    0,    0],\n",
            "       [   2,   90,  318, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 191), dtype=int64, numpy=\n",
            "array([[   2, 6623,  400, ...,    0,    0,    0],\n",
            "       [   2,   81,   80, ...,    0,    0,    0],\n",
            "       [   2,   36,   36, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   96,   80, ...,    0,    0,    0],\n",
            "       [   2,   76,  921, ...,    0,    0,    0],\n",
            "       [   2,  201,  124, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 133), dtype=int64, numpy=\n",
            "array([[   2,  346,   84, ...,    0,    0,    0],\n",
            "       [   2, 3064, 1373, ...,    0,    0,    0],\n",
            "       [   2,  609,  103, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   44,  151, ...,    0,    0,    0],\n",
            "       [   2,  115,   89, ...,    0,    0,    0],\n",
            "       [   2,   54, 3271, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 142), dtype=int64, numpy=\n",
            "array([[   2,  110,   13, ...,    0,    0,    0],\n",
            "       [   2,  936, 5558, ...,    0,    0,    0],\n",
            "       [   2,  133,   95, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   45,  146, ...,    0,    0,    0],\n",
            "       [   2,  117,   79, ...,    0,    0,    0],\n",
            "       [   2, 6637,   80, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 138), dtype=int64, numpy=\n",
            "array([[  2,  99,  98, ...,   0,   0,   0],\n",
            "       [  2,  39,  39, ...,   0,   0,   0],\n",
            "       [  2,  44, 132, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2, 100, 142, ...,   0,   0,   0],\n",
            "       [  2, 280, 916, ...,   0,   0,   0],\n",
            "       [  2, 101, 107, ...,   0,   0,   0]])>, <tf.Tensor: shape=(1024, 133), dtype=int64, numpy=\n",
            "array([[  2,  87,  78, ...,   0,   0,   0],\n",
            "       [  2,  36,  36, ...,   0,   0,   0],\n",
            "       [  2,  72,  76, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  77,  99, ...,   0,   0,   0],\n",
            "       [  2,  45,   9, ...,   0,   0,   0],\n",
            "       [  2,  82, 107, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 223), dtype=int64, numpy=\n",
            "array([[   2,   44,  416, ...,    0,    0,    0],\n",
            "       [   2,  183,  163, ...,    0,    0,    0],\n",
            "       [   2,  698,   83, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   39,   39, ...,    0,    0,    0],\n",
            "       [   2,   53,  818, ...,    0,    0,    0],\n",
            "       [   2,   91, 1279, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 172), dtype=int64, numpy=\n",
            "array([[   2,   72,  146, ...,    0,    0,    0],\n",
            "       [   2,  175,   84, ...,    0,    0,    0],\n",
            "       [   2,   45,    9, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   71,  152, ...,    0,    0,    0],\n",
            "       [   2,   50,  601, ...,    0,    0,    0],\n",
            "       [   2,  377, 1377, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 141), dtype=int64, numpy=\n",
            "array([[   2,  118, 2333, ...,    0,    0,    0],\n",
            "       [   2,   39,   39, ...,    0,    0,    0],\n",
            "       [   2,  460, 1735, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   54,  929, ...,    0,    0,    0],\n",
            "       [   2,  137,   40, ...,    0,    0,    0],\n",
            "       [   2,  503, 1393, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 139), dtype=int64, numpy=\n",
            "array([[   2,  790,  500, ...,    0,    0,    0],\n",
            "       [   2,   71, 2760, ...,    0,    0,    0],\n",
            "       [   2,  335, 2993, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2, 4367, 4656, ...,    0,    0,    0],\n",
            "       [   2,   86,   79, ...,    0,    0,    0],\n",
            "       [   2,  174, 1524, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 291), dtype=int64, numpy=\n",
            "array([[  2,  24,  44, ...,   0,   0,   0],\n",
            "       [  2, 107, 902, ...,   0,   0,   0],\n",
            "       [  2,  39,  39, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2, 101, 235, ...,   0,   0,   0],\n",
            "       [  2,  47, 266, ...,   0,   0,   0],\n",
            "       [  2,  44, 151, ...,   0,   0,   0]])>, <tf.Tensor: shape=(1024, 320), dtype=int64, numpy=\n",
            "array([[  2, 512,  80, ...,   0,   0,   0],\n",
            "       [  2,  72,  75, ...,   0,   0,   0],\n",
            "       [  2,  45,  91, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  85, 251, ...,   0,   0,   0],\n",
            "       [  2,  44, 240, ...,   0,   0,   0],\n",
            "       [  2,  72, 146, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 194), dtype=int64, numpy=\n",
            "array([[   2, 4372,  146, ...,    0,    0,    0],\n",
            "       [   2,   99,   88, ...,    0,    0,    0],\n",
            "       [   2,   92, 4272, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  623,  106, ...,    0,    0,    0],\n",
            "       [   2,  123,   54, ...,    0,    0,    0],\n",
            "       [   2,  101,  443, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 208), dtype=int64, numpy=\n",
            "array([[   2,   78,  302, ...,    0,    0,    0],\n",
            "       [   2,   87,   83, ...,    0,    0,    0],\n",
            "       [   2, 2038,   88, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   78,  129, ...,    0,    0,    0],\n",
            "       [   2,   76,    9, ...,    0,    0,    0],\n",
            "       [   2,   72, 6780, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 155), dtype=int64, numpy=\n",
            "array([[   2,  104, 1936, ...,    0,    0,    0],\n",
            "       [   2,   44,  102, ...,    0,    0,    0],\n",
            "       [   2,   44,  115, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  105,   44, ...,    0,    0,    0],\n",
            "       [   2,   39,   39, ...,    0,    0,    0],\n",
            "       [   2,   54,  171, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 140), dtype=int64, numpy=\n",
            "array([[   2,   45,    9, ...,    0,    0,    0],\n",
            "       [   2,   76,    9, ...,    0,    0,    0],\n",
            "       [   2,   72,  117, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   81,   80, ...,    0,    0,    0],\n",
            "       [   2,   45,  505, ...,    0,    0,    0],\n",
            "       [   2,   71, 6657, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 156), dtype=int64, numpy=\n",
            "array([[   2,   39,   39, ...,    0,    0,    0],\n",
            "       [   2, 1740,  446, ...,    0,    0,    0],\n",
            "       [   2,   54,   84, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,  151,   83, ...,    0,    0,    0],\n",
            "       [   2, 1630,  436, ...,    0,    0,    0],\n",
            "       [   2,   44,  229, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 154), dtype=int64, numpy=\n",
            "array([[   2,   82,   71, ...,    0,    0,    0],\n",
            "       [   2,  680, 5558, ...,    0,    0,    0],\n",
            "       [   2,   90,  234, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   72,  411, ...,    0,    0,    0],\n",
            "       [   2,   52, 2713, ...,    0,    0,    0],\n",
            "       [   2,   72,   45, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(1024, 124), dtype=int64, numpy=\n",
            "array([[   2, 1719,   14, ...,    0,    0,    0],\n",
            "       [   2,  202,   92, ...,    0,    0,    0],\n",
            "       [   2,  127, 2982, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   44,  108, ...,    0,    0,    0],\n",
            "       [   2,  121,   86, ...,    0,    0,    0],\n",
            "       [   2,  931,   40, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 118), dtype=int64, numpy=\n",
            "array([[  2,  79, 132, ...,   0,   0,   0],\n",
            "       [  2, 101,  71, ...,   0,   0,   0],\n",
            "       [  2,  72, 115, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  72,  81, ...,   0,   0,   0],\n",
            "       [  2,  76,  84, ...,   0,   0,   0],\n",
            "       [  2,  78, 212, ...,   0,   0,   0]])>)\n",
            "(<tf.Tensor: shape=(1024, 158), dtype=int64, numpy=\n",
            "array([[   2,   44,   88, ...,    0,    0,    0],\n",
            "       [   2,   44,   54, ...,    0,    0,    0],\n",
            "       [   2, 5932,   28, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   10,  165, ...,    0,    0,    0],\n",
            "       [   2,  619,   15, ...,    0,    0,    0],\n",
            "       [   2,  260,  309, ...,    0,    0,    0]])>, <tf.Tensor: shape=(1024, 173), dtype=int64, numpy=\n",
            "array([[   2,   72,   98, ...,    0,    0,    0],\n",
            "       [   2,   76,    9, ...,    0,    0,    0],\n",
            "       [   2, 4998,   27, ...,    0,    0,    0],\n",
            "       ...,\n",
            "       [   2,   36,   36, ...,    0,    0,    0],\n",
            "       [   2,  186,  114, ...,    0,    0,    0],\n",
            "       [   2,  169,  110, ...,    0,    0,    0]])>)\n",
            "(<tf.Tensor: shape=(585, 147), dtype=int64, numpy=\n",
            "array([[  2,  88, 120, ...,   0,   0,   0],\n",
            "       [  2, 175, 153, ...,   0,   0,   0],\n",
            "       [  2, 101, 105, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2, 103, 770, ...,   0,   0,   0],\n",
            "       [  2, 133,  14, ...,   0,   0,   0],\n",
            "       [  2,  91, 301, ...,   0,   0,   0]])>, <tf.Tensor: shape=(585, 134), dtype=int64, numpy=\n",
            "array([[  2, 110,  13, ...,   0,   0,   0],\n",
            "       [  2,  45, 628, ...,   0,   0,   0],\n",
            "       [  2,  87, 140, ...,   0,   0,   0],\n",
            "       ...,\n",
            "       [  2,  77,  71, ...,   0,   0,   0],\n",
            "       [  2, 110,  13, ...,   0,   0,   0],\n",
            "       [  2,  99, 474, ...,   0,   0,   0]])>)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "BUFFER_SIZE = 20000\n",
        "BATCH_SIZE = 64"
      ],
      "metadata": {
        "id": "97qPE2aGvOs8"
      },
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def make_batches(ds):\n",
        "    return (\n",
        "        ds\n",
        "        .cache()\n",
        "        .shuffle(BUFFER_SIZE)\n",
        "        .batch(BATCH_SIZE)\n",
        "        .map(tokenize_pairs, num_parallel_calls = tf.data.AUTOTUNE)\n",
        "        .filter(filter_max_tokens)\n",
        "        .prefetch(tf.data.AUTOTUNE)\n",
        "    )\n",
        "\n",
        "train_batches = make_batches(train_examples)\n",
        "val_batches = make_batches(val_examples)"
      ],
      "metadata": {
        "id": "qVlgOxxAvUeZ"
      },
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def get_angles(pos, i, d_model):\n",
        "    angle_rates = 1/np.power(10000, (2 * (i//2)) / np.float32(d_model))\n",
        "    return pos * angle_rates\n",
        "\n",
        "def positional_encoding(position, d_model):\n",
        "    angle_rads = get_angles(np.arange(position)[:, np.newaxis],\n",
        "                            np.arange(d_model)[np.newaxis, :],\n",
        "                            d_model)\n",
        "    angle_rads[:, 0::2] = np.sin(angle_rads[:, 0::2])\n",
        "    angle_rads[:, 1::2] = np.cos(angle_rads[:, 1::2])\n",
        "    pos_encoding = angle_rads[np.newaxis, ...]\n",
        "    return tf.cast(pos_encoding, dtype = tf.float32)"
      ],
      "metadata": {
        "id": "RqhtOQ21vXAR"
      },
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(positional_encoding(5, 768))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Eus1HE5iANJr",
        "outputId": "287e564b-9fb7-47c3-8330-bf7949466989"
      },
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tf.Tensor(\n",
            "[[[ 0.0000000e+00  1.0000000e+00  0.0000000e+00 ...  1.0000000e+00\n",
            "    0.0000000e+00  1.0000000e+00]\n",
            "  [ 8.4147096e-01  5.4030228e-01  8.2843077e-01 ...  1.0000000e+00\n",
            "    1.0242752e-04  1.0000000e+00]\n",
            "  [ 9.0929741e-01 -4.1614684e-01  9.2799401e-01 ...  1.0000000e+00\n",
            "    2.0485504e-04  1.0000000e+00]\n",
            "  [ 1.4112000e-01 -9.8999250e-01  2.1109235e-01 ...  9.9999994e-01\n",
            "    3.0728255e-04  9.9999994e-01]\n",
            "  [-7.5680250e-01 -6.5364361e-01 -6.9153196e-01 ...  9.9999994e-01\n",
            "    4.0971008e-04  9.9999994e-01]]], shape=(1, 5, 768), dtype=float32)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "n, d = 2048, 512\n",
        "pos_encoding = np.squeeze(positional_encoding(n, d))\n",
        "\n",
        "pos_encoding = tf.transpose(pos_encoding)\n",
        "# pos_encoding = tf.reshape(pos_encoding, (n, d//2, 2))\n",
        "# pos_encoding = tf.transpose(pos_encoding, (2, 1, 0))\n",
        "# pos_encoding = tf.reshape(pos_encoding, (d, n))\n",
        "print(pos_encoding)\n",
        "\n",
        "plt.pcolormesh(pos_encoding, cmap='RdBu')\n",
        "plt.ylabel('Depth')\n",
        "plt.xlabel('Position')\n",
        "plt.colorbar()\n",
        "plt.show()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 526
        },
        "id": "0pKrTLE5AcRh",
        "outputId": "9dcb6bcd-b259-4fd5-f7b9-30b3d34be5cb"
      },
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tf.Tensor(\n",
            "[[ 0.0000000e+00  8.4147096e-01  9.0929741e-01 ...  1.7589758e-01\n",
            "  -7.3331332e-01 -9.6831930e-01]\n",
            " [ 1.0000000e+00  5.4030228e-01 -4.1614684e-01 ... -9.8440850e-01\n",
            "  -6.7989087e-01  2.4971525e-01]\n",
            " [ 0.0000000e+00  8.2185620e-01  9.3641472e-01 ... -1.8608274e-01\n",
            "   7.0149130e-01  9.8535496e-01]\n",
            " ...\n",
            " [ 1.0000000e+00  1.0000000e+00  1.0000000e+00 ...  9.7595036e-01\n",
            "   9.7592694e-01  9.7590351e-01]\n",
            " [ 0.0000000e+00  1.0366329e-04  2.0732658e-04 ...  2.1040717e-01\n",
            "   2.1050851e-01  2.1060985e-01]\n",
            " [ 1.0000000e+00  1.0000000e+00  1.0000000e+00 ...  9.7761387e-01\n",
            "   9.7759205e-01  9.7757018e-01]], shape=(512, 2048), dtype=float32)\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def create_padding_mask(seq):\n",
        "    seq = tf.cast(tf.math.equal(seq, 0), tf.float32)\n",
        "\n",
        "    return seq[:, tf.newaxis, tf.newaxis, :] # (batch_size, 1, 1, seq_len)"
      ],
      "metadata": {
        "id": "DTifnPAlAxWb"
      },
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x = tf.constant([[7, 6, 0, 0, 1], [1, 2, 3, 0, 0], [0, 0, 0, 4, 5]])\n",
        "create_padding_mask(x)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Ct98FOd_Cyx3",
        "outputId": "3c340c9a-0b7e-4bcc-92ef-f2ef0af49fb4"
      },
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tf.Tensor: shape=(3, 1, 1, 5), dtype=float32, numpy=\n",
              "array([[[[0., 0., 1., 1., 0.]]],\n",
              "\n",
              "\n",
              "       [[[0., 0., 0., 1., 1.]]],\n",
              "\n",
              "\n",
              "       [[[1., 1., 1., 0., 0.]]]], dtype=float32)>"
            ]
          },
          "metadata": {},
          "execution_count": 24
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def create_look_ahead_mask(size):\n",
        "    mask = 1 - tf.linalg.band_part(tf.ones((size, size)), -1, 0)\n",
        "    return mask"
      ],
      "metadata": {
        "id": "s3v6Pi5jC2HG"
      },
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x = tf.random.uniform((1, 3))\n",
        "print(x)\n",
        "temp = create_look_ahead_mask(x.shape[1])\n",
        "print(temp)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_WwD4t7ZDrgO",
        "outputId": "26263ed3-fffc-4909-ab8b-639f46fb87f1"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tf.Tensor([[0.99497867 0.38387823 0.7204126 ]], shape=(1, 3), dtype=float32)\n",
            "tf.Tensor(\n",
            "[[0. 1. 1.]\n",
            " [0. 0. 1.]\n",
            " [0. 0. 0.]], shape=(3, 3), dtype=float32)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def scaled_dot_product_attention(q, k, v, mask):\n",
        "    \"\"\"Calculate the attention weights.\n",
        "    q, k, v must have matching leading dimensions.\n",
        "    k, v must have matching penultimate dimension, i.e.: seq_len_k = seq_len_v.\n",
        "    The mask has different shapes depending on its type(padding or look ahead)\n",
        "    but it must be broadcastable for addition.\n",
        "\n",
        "    Args:\n",
        "        q: query shape == (..., seq_len_q, depth)\n",
        "        k: key shape == (..., seq_len_k, depth)\n",
        "        v: value shape == (..., seq_len_v, depth_v)\n",
        "        mask: Float tensor with shape broadcastable \n",
        "            to (..., seq_len_q, seq_len_k). Defaults to None.\n",
        "\n",
        "    Returns:\n",
        "        output, attention_weights\n",
        "    \"\"\"\n",
        "\n",
        "    matmul_qk = tf.matmul(q, k, transpose_b = True) # (..., seq_len_q, seq_len_k)\n",
        "\n",
        "    # scale matmul_qk\n",
        "    dk = tf.cast(tf.shape(k)[-1], tf.float32)\n",
        "    scaled_attention_logits = matmul_qk / tf.math.sqrt(dk)\n",
        "\n",
        "    if mask is not None:\n",
        "        scaled_attention_logits += (mask*-1e9)\n",
        "    \n",
        "    # softmax is normalized on the last axis (seq_len_k) so that the scores\n",
        "    # add up to 1.\n",
        "    attention_weights = tf.nn.softmax(scaled_attention_logits, axis = -1) # (..., seq_len_q, seq_len_k)\n",
        "\n",
        "    output = tf.matmul(attention_weights, v) # (..., seq_len_q, depth_v)\n",
        "\n",
        "    return output, attention_weights\n",
        "\n",
        "def print_out(q, k, v):\n",
        "    temp_out, temp_attn = scaled_dot_product_attention(\n",
        "        q, k, v, None\n",
        "    )\n",
        "    print('Attention weights are:')\n",
        "    print(temp_attn)\n",
        "    print('Output is:')\n",
        "    print(temp_out)"
      ],
      "metadata": {
        "id": "hInMHga_Dtb3"
      },
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "temp_k = tf.constant([[10, 0, 0],\n",
        "                      [0, 10, 0],\n",
        "                      [0, 0, 10],\n",
        "                      [0, 0, 10]], dtype=tf.float32)  # (4, 3)\n",
        "\n",
        "temp_v = tf.constant([[1, 0],\n",
        "                      [10, 0],\n",
        "                      [100, 5],\n",
        "                      [1000, 6]], dtype=tf.float32)  # (4, 2)\n",
        "\n",
        "temp_q = tf.constant([[0, 10, 0]], dtype = tf.float32) # (1, 3)\n",
        "\n",
        "print_out(temp_q, temp_k, temp_v)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EW96mY-FH_9Q",
        "outputId": "29dcb75b-5b50-457b-9a1c-ce8bc7fd8722"
      },
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Attention weights are:\n",
            "tf.Tensor([[8.433274e-26 1.000000e+00 8.433274e-26 8.433274e-26]], shape=(1, 4), dtype=float32)\n",
            "Output is:\n",
            "tf.Tensor([[1.000000e+01 9.276601e-25]], shape=(1, 2), dtype=float32)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# This query aligns with a repeated key (third and fourth),\n",
        "# so all associated values get averaged.\n",
        "temp_q = tf.constant([[0, 0, 10]], dtype=tf.float32)  # (1, 3)\n",
        "print_out(temp_q, temp_k, temp_v)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2fRNzZHHICsM",
        "outputId": "4ae5f8cc-d3d1-46e8-a60d-5094985d9600"
      },
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Attention weights are:\n",
            "tf.Tensor([[4.216637e-26 4.216637e-26 5.000000e-01 5.000000e-01]], shape=(1, 4), dtype=float32)\n",
            "Output is:\n",
            "tf.Tensor([[550.    5.5]], shape=(1, 2), dtype=float32)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# This query aligns equally with the first and second key,\n",
        "# so their values get averaged.\n",
        "temp_q = tf.constant([[10, 10, 0]], dtype=tf.float32)  # (1, 3)\n",
        "print_out(temp_q, temp_k, temp_v)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "j9Riy_sKIdsE",
        "outputId": "d017ac71-fd61-45ec-ebc5-3161225f51ea"
      },
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Attention weights are:\n",
            "tf.Tensor([[5.000000e-01 5.000000e-01 4.216637e-26 4.216637e-26]], shape=(1, 4), dtype=float32)\n",
            "Output is:\n",
            "tf.Tensor([[5.5000000e+00 4.6383005e-25]], shape=(1, 2), dtype=float32)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "temp_q = tf.constant([[0, 0, 10],\n",
        "                      [0, 10, 0],\n",
        "                      [10, 10, 0]], dtype=tf.float32)  # (3, 3)\n",
        "print_out(temp_q, temp_k, temp_v)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7ZLCd3sjImy-",
        "outputId": "dd50fd38-40a6-45e0-fada-35990232ddbe"
      },
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Attention weights are:\n",
            "tf.Tensor(\n",
            "[[4.216637e-26 4.216637e-26 5.000000e-01 5.000000e-01]\n",
            " [8.433274e-26 1.000000e+00 8.433274e-26 8.433274e-26]\n",
            " [5.000000e-01 5.000000e-01 4.216637e-26 4.216637e-26]], shape=(3, 4), dtype=float32)\n",
            "Output is:\n",
            "tf.Tensor(\n",
            "[[5.5000000e+02 5.5000000e+00]\n",
            " [1.0000000e+01 9.2766009e-25]\n",
            " [5.5000000e+00 4.6383005e-25]], shape=(3, 2), dtype=float32)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class MultiHeadAttention(tf.keras.layers.Layer):\n",
        "    def __init__(self, *, d_model, num_heads):\n",
        "        super(MultiHeadAttention, self).__init__()\n",
        "        self.num_heads = num_heads\n",
        "        self.d_model = d_model\n",
        "\n",
        "        assert d_model % self.num_heads == 0\n",
        "\n",
        "        self.depth = d_model // self.num_heads\n",
        "\n",
        "        self.wq = tf.keras.layers.Dense(d_model)\n",
        "        self.wk = tf.keras.layers.Dense(d_model)\n",
        "        self.wv = tf.keras.layers.Dense(d_model)\n",
        "\n",
        "        self.dense = tf.keras.layers.Dense(d_model)\n",
        "\n",
        "    def split_heads(self, x, batch_size):\n",
        "        \"\"\"Split the last dimension into (num_heads, depth).\n",
        "        Transpose the result so that the shape is (batch_size, num_heads, seq_len, depth)\n",
        "        \"\"\"\n",
        "        x = tf.reshape(x, (batch_size, -1, self.num_heads, self.depth))\n",
        "        return tf.transpose(x, perm = [0, 2, 1, 3])\n",
        "\n",
        "    def call(self, v, k, q, mask):\n",
        "        batch_size = tf.shape(q)[0]\n",
        "\n",
        "        q = self.wq(q) # (batch_size, seq_len, d_model)\n",
        "        k = self.wk(k) # (batch_size, seq_len, d_model)\n",
        "        v = self.wv(v) # (batch_size, seq_len, d_model)\n",
        "\n",
        "        q = self.split_heads(q, batch_size) # (batch_size, num_heads, seq_len, depth)\n",
        "        k = self.split_heads(k, batch_size) # (batch_size, num_heads, seq_len, depth)\n",
        "        v = self.split_heads(v, batch_size) # (batch_size, num_heads, seq_len, depth)\n",
        "\n",
        "        # scaled_attentions.shape == (batch_size, num_heads, seq_len_q, depth)\n",
        "        # attention_weights.shape == (batch_size, num_heads, seq_len_q, seq_len_k)\n",
        "        scaled_attention, attention_weights = scaled_dot_product_attention(\n",
        "            q, k, v, mask\n",
        "        )\n",
        "\n",
        "        scaled_attention = tf.transpose(scaled_attention, perm = [0, 2, 1, 3]) # (batch_size, seq_len_q, num_heads, depth)\n",
        "\n",
        "        concat_attention = tf.reshape(scaled_attention,\n",
        "                                      (batch_size, -1, self.d_model)) # (batch_size, seq_len_q, d_model)\n",
        "        \n",
        "        output = self.dense(concat_attention)\n",
        "\n",
        "        return output, attention_weights"
      ],
      "metadata": {
        "id": "3DZTphwUIqQA"
      },
      "execution_count": 32,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "temp_mha = MultiHeadAttention(d_model = 512, num_heads = 8)\n",
        "y = tf.random.uniform((1, 60, 512)) # (batch_size, encoder_sequence, d_model)\n",
        "out, attn = temp_mha(y, k=y, q=y, mask = None)\n",
        "out.shape, attn.shape"
      ],
      "metadata": {
        "id": "uZdIwlj5ku9C",
        "outputId": "eac073da-df1a-4f04-bd45-eb614db13d64",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(TensorShape([1, 60, 512]), TensorShape([1, 8, 60, 60]))"
            ]
          },
          "metadata": {},
          "execution_count": 33
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def point_wise_feed_forward_network(d_model, dff):\n",
        "    return tf.keras.Sequential([\n",
        "        tf.keras.layers.Dense(dff, activation = 'relu'),\n",
        "        tf.keras.layers.Dense(d_model)\n",
        "    ])"
      ],
      "metadata": {
        "id": "leLWV28Dk-kt"
      },
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "class EncoderLayer(tf.keras.layers.Layer):\n",
        "    def __init__(self, *, d_model, num_heads, dff, rate = 0.1):\n",
        "        super(EncoderLayer, self).__init__()\n",
        "\n",
        "        self.mha = MultiHeadAttention(d_model = d_model, num_heads = num_heads)\n",
        "        self.ffn = point_wise_feed_forward_network(d_model, dff)\n",
        "\n",
        "        self.layernorm1 = tf.keras.layers.LayerNormalization(epsilon = 1e-6)\n",
        "        self.layernorm2 = tf.keras.layers.LayerNormalization(epsilon = 1e-6)\n",
        "\n",
        "        self.dropout1 = tf.keras.layers.Dropout(rate)\n",
        "        self.dropout2 = tf.keras.layers.Dropout(rate)\n",
        "\n",
        "    def call(self, x, training, mask):\n",
        "\n",
        "        attn_output, _ = self.mha(x, x, x, mask) # (batch_size, input_seq_len, d_model)\n",
        "        attn_output = self.dropout1(attn_output, training = training)\n",
        "        out1 = self.layernorm1(x + attn_output) # (batch_size, input_seq_len, d_model)\n",
        "\n",
        "        ffn_output = self.ffn(out1) # (batch_size, input_seq_len, d_model)\n",
        "        ffn_output = self.dropout2(ffn_output, training = training)\n",
        "        out2 = self.layernorm2(out1 + ffn_output) # (batch_size, input_seq_len, d_model)\n",
        "\n",
        "        return out2\n"
      ],
      "metadata": {
        "id": "-i_bbybUmL1-"
      },
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "sample_encoder_layer = EncoderLayer(d_model = 512, num_heads = 8, dff = 2048)\n",
        "\n",
        "sample_encoder_layer_output = sample_encoder_layer(\n",
        "    tf.random.uniform((64, 43, 512)), False, None)\n",
        "\n",
        "sample_encoder_layer_output.shape"
      ],
      "metadata": {
        "id": "zQTFxMvapc6k",
        "outputId": "10693b4e-7201-4beb-a6ee-607f719e01b0",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "TensorShape([64, 43, 512])"
            ]
          },
          "metadata": {},
          "execution_count": 36
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class DecoderLayer(tf.keras.layers.Layer):\n",
        "    def __init__(self, *, d_model, num_heads, dff, rate = 0.1):\n",
        "        super(DecoderLayer, self).__init__()\n",
        "\n",
        "        self.mha1 = MultiHeadAttention(d_model = d_model, num_heads = num_heads)\n",
        "        self.mha2 = MultiHeadAttention(d_model = d_model, num_heads = num_heads)\n",
        "\n",
        "        self.ffn = point_wise_feed_forward_network(d_model, dff)\n",
        "\n",
        "        self.layernorm1 = tf.keras.layers.LayerNormalization(epsilon = 1e-6)\n",
        "        self.layernorm2 = tf.keras.layers.LayerNormalization(epsilon = 1e-6)\n",
        "        self.layernorm3 = tf.keras.layers.LayerNormalization(epsilon = 1e-6)\n",
        "\n",
        "        self.dropout1 = tf.keras.layers.Dropout(rate)\n",
        "        self.dropout2 = tf.keras.layers.Dropout(rate)\n",
        "        self.dropout3 = tf.keras.layers.Dropout(rate)\n",
        "\n",
        "    def call(self, x, enc_output, training, look_ahead_mask, padding_mask):\n",
        "        # enc_output.shape == (batch_size, input_seq_len, d_model)\n",
        "\n",
        "        attn1, attn_weights_block1 = self.mha1(x, x, x, look_ahead_mask) # (batch_size, target_seq_len, d_model)\n",
        "        att1 = self.dropout1(attn1, training = training)\n",
        "        out1 = self.layernorm1(attn1 + x)\n",
        "\n",
        "        attn2, attn_weights_block2 = self.mha2(\n",
        "            enc_output, enc_output, out1, padding_mask) # (batch_size, target_seq_len, d_model)\n",
        "        att2 = self.dropout2(attn2, training = training)\n",
        "        out2 = self.layernorm2(attn2 + out1)  # (batch_size, target_seq_len, d_model)\n",
        "\n",
        "        ffn_output = self.ffn(out2)  # (batch_size, target_seq_len, d_model)\n",
        "        ffn_output = self.dropout3(ffn_output, training=training)\n",
        "        out3 = self.layernorm3(ffn_output + out2)  # (batch_size, target_seq_len, d_model)\n",
        "\n",
        "        return out3, attn_weights_block1, attn_weights_block2"
      ],
      "metadata": {
        "id": "PDbhg5r2p8px"
      },
      "execution_count": 37,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "sample_decoder_layer = DecoderLayer(d_model = 512, num_heads = 8, dff = 2048)\n",
        "\n",
        "sample_decoder_layer_output, _, _ = sample_decoder_layer(\n",
        "    tf.random.uniform((64, 50, 512)), sample_encoder_layer_output,\n",
        "    False, None, None)\n",
        "\n",
        "sample_decoder_layer_output.shape"
      ],
      "metadata": {
        "id": "vZvn6iuZvFIE",
        "outputId": "75784157-afc0-4195-9256-a539a2f70ba2",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "TensorShape([64, 50, 512])"
            ]
          },
          "metadata": {},
          "execution_count": 38
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class Encoder(tf.keras.layers.Layer):\n",
        "    def __init__(self, *, num_layers, d_model, num_heads, dff, input_vocab_size,\n",
        "                 rate = 0.1):\n",
        "        super(Encoder, self).__init__()\n",
        "        self.d_model = d_model\n",
        "        self.num_layers = num_layers\n",
        "        self.embedding = tf.keras.layers.Embedding(input_vocab_size, d_model)\n",
        "        self.pos_encoding = positional_encoding(MAX_TOKENS, self.d_model)\n",
        "\n",
        "        self.enc_layers = [\n",
        "            EncoderLayer(d_model = d_model, num_heads = num_heads, dff = dff,\n",
        "                         rate = rate)\n",
        "            for _ in range(num_layers)\n",
        "        ]\n",
        "\n",
        "        self.dropout = tf.keras.layers.Dropout(rate)\n",
        "\n",
        "    def call(self, x, training, mask):\n",
        "        seq_len = tf.shape(x)[1]\n",
        "\n",
        "        # adding embedding and position encoding\n",
        "        x = self.embedding(x)\n",
        "        x *= tf.math.sqrt(tf.cast(self.d_model, tf.float32))\n",
        "        x += self.pos_encoding[:, :seq_len, :]\n",
        "        x = self.dropout(x, training = training)\n",
        "\n",
        "        for i in range(self.num_layers):\n",
        "            x = self.enc_layers[i](x, training, mask)\n",
        "\n",
        "        return x"
      ],
      "metadata": {
        "id": "W_gz9Skyvbtp"
      },
      "execution_count": 39,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "sample_encoder = Encoder(num_layers = 2, d_model = 512, num_heads = 8, dff = 2048,\n",
        "                         input_vocab_size = 8500)\n",
        "\n",
        "temp_input = tf.random.uniform((64, 62), dtype = tf.int64, minval = 0, maxval = 200)\n",
        "\n",
        "sample_encoder_output = sample_encoder(temp_input, training = False, mask = None)\n",
        "\n",
        "print(sample_encoder_output.shape) # (batch_size, input_seq_len, d_model)"
      ],
      "metadata": {
        "id": "TFHxlFsZwxlO",
        "outputId": "a89303f2-f9bc-4d34-d57b-a4cfa8d732fc",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(64, 62, 512)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class Decoder(tf.keras.layers.Layer):\n",
        "    def __init__(self, *, num_layers, d_model, num_heads, dff, target_vocab_size,\n",
        "                 rate = 0.1):\n",
        "        super(Decoder, self).__init__()\n",
        "\n",
        "        self.d_model = d_model\n",
        "        self.num_layers = num_layers\n",
        "\n",
        "        self.embedding = tf.keras.layers.Embedding(target_vocab_size, d_model)\n",
        "        self.pos_encoding = positional_encoding(MAX_TOKENS, d_model)\n",
        "\n",
        "        self.dec_layers = [\n",
        "            DecoderLayer(d_model=d_model, num_heads=num_heads, dff=dff, rate=rate)\n",
        "            for _ in range(num_layers)]\n",
        "        self.dropout = tf.keras.layers.Dropout(rate)\n",
        "\n",
        "    def call(self, x, enc_output, training, look_ahead_mask, padding_mask):\n",
        "        seq_len = tf.shape(x)[1]\n",
        "        attention_weights = {}\n",
        "\n",
        "        x = self.embedding(x) # (batch_size, target_seq_len, d_model)\n",
        "        x *= tf.math.sqrt(tf.cast(self.d_model, tf.float32))\n",
        "        x += self.pos_encoding[:, :seq_len, :]\n",
        "\n",
        "        x = self.dropout(x, training = training)\n",
        "\n",
        "        for i in range(self.num_layers):\n",
        "            x, block1, block2 = self.dec_layers[i](x, enc_output, training,\n",
        "                                                   look_ahead_mask, padding_mask)\n",
        "            attention_weights[f\"decoder_layer{i+1}_block1\"] = block1\n",
        "            attention_weights[f\"decoder_layer{i+1}_block2\"] = block2\n",
        "\n",
        "        # x.shape == (batch_size, target_seq_len, d_model)\n",
        "        return x, attention_weights"
      ],
      "metadata": {
        "id": "ZaCkSn92xJJg"
      },
      "execution_count": 41,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "sample_decoder = Decoder(num_layers = 2, d_model = 512, num_heads = 8, dff = 2048,\n",
        "                         target_vocab_size = 8000)\n",
        "\n",
        "temp_input = tf.random.uniform((64, 26))\n",
        "\n",
        "output, attn = sample_decoder(temp_input, enc_output = sample_encoder_output,\n",
        "                              training = False,\n",
        "                              look_ahead_mask = None,\n",
        "                              padding_mask = None\n",
        "                              )\n",
        "\n",
        "output.shape, attn['decoder_layer2_block2'].shape"
      ],
      "metadata": {
        "id": "_glP2VAs2Tvh",
        "outputId": "5d11ee75-57af-414a-cb28-93481f7ea15d",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(TensorShape([64, 26, 512]), TensorShape([64, 8, 26, 62]))"
            ]
          },
          "metadata": {},
          "execution_count": 42
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class Transformer(tf.keras.Model):\n",
        "    def __init__(self, *, num_layers, d_model, num_heads, dff, input_vocab_size,\n",
        "                 target_vocab_size, rate = 0.1):\n",
        "        super().__init__()\n",
        "        self.encoder = Encoder(num_layers = num_layers, d_model = d_model,\n",
        "                               num_heads = num_heads, dff = dff,\n",
        "                               input_vocab_size = input_vocab_size, rate = rate)\n",
        "\n",
        "        self.decoder = Decoder(num_layers = num_layers, d_model = d_model,\n",
        "                               num_heads = num_heads, dff = dff,\n",
        "                               target_vocab_size = target_vocab_size, rate = rate)\n",
        "        \n",
        "        self.final_layer = tf.keras.layers.Dense(target_vocab_size)\n",
        "\n",
        "\n",
        "    def call(self, inputs, training):\n",
        "        inp, tar = inputs\n",
        "        \n",
        "        padding_mask, look_ahead_mask = self.create_masks(inp, tar)\n",
        "\n",
        "        enc_output = self.encoder(inp, training, padding_mask)\n",
        "\n",
        "        dec_output, attention_weights = self.decoder(\n",
        "            tar, enc_output, training, look_ahead_mask, padding_mask)\n",
        "        \n",
        "        final_output = self.final_layer(dec_output)\n",
        "\n",
        "        return final_output, attention_weights\n",
        "\n",
        "    def create_masks(self, inp, tar):\n",
        "        padding_mask = create_padding_mask(inp)\n",
        "\n",
        "        look_ahead_mask = create_look_ahead_mask(tf.shape(tar)[1])\n",
        "        dec_target_padding_mask = create_padding_mask(tar)\n",
        "        look_ahead_mask = tf.maximum(dec_target_padding_mask, look_ahead_mask)\n",
        "\n",
        "        return padding_mask, look_ahead_mask\n",
        "\n"
      ],
      "metadata": {
        "id": "heEYotL425MC"
      },
      "execution_count": 43,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "sample_transformer = Transformer(\n",
        "    num_layers = 2, d_model = 512, num_heads = 8, dff = 2048, \n",
        "    input_vocab_size = 8500, target_vocab_size = 8000\n",
        ")\n",
        "\n",
        "temp_input = tf.random.uniform((64, 38), dtype=tf.int64, minval=0, maxval=200)\n",
        "temp_target = tf.random.uniform((64, 36), dtype=tf.int64, minval=0, maxval=200)\n",
        "\n",
        "fn_out, _ = sample_transformer([temp_input, temp_target], training=False)\n",
        "\n",
        "fn_out.shape"
      ],
      "metadata": {
        "id": "IPccfWAN5Aaf",
        "outputId": "95ea2da0-516e-49ec-9a58-5ab5ac9d88ef",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 44,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "TensorShape([64, 36, 8000])"
            ]
          },
          "metadata": {},
          "execution_count": 44
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "sample_transformer.summary()"
      ],
      "metadata": {
        "id": "_XtSSAhL5QXw",
        "outputId": "c8f4826f-c20d-41f2-f3a8-ab14a34fc003",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 45,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"transformer\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " encoder_1 (Encoder)         multiple                  10656768  \n",
            "                                                                 \n",
            " decoder_1 (Decoder)         multiple                  12504064  \n",
            "                                                                 \n",
            " dense_84 (Dense)            multiple                  4104000   \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 27,264,832\n",
            "Trainable params: 27,264,832\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "num_layers = 4\n",
        "d_model = 128\n",
        "dff = 512\n",
        "num_heads = 8\n",
        "dropout_rate = 0.1"
      ],
      "metadata": {
        "id": "jWE8D4Dn-24X"
      },
      "execution_count": 46,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "class CustomSchedule(tf.keras.optimizers.schedules.LearningRateSchedule):\n",
        "  def __init__(self, d_model, warmup_steps=4000):\n",
        "    super(CustomSchedule, self).__init__()\n",
        "\n",
        "    self.d_model = d_model\n",
        "    self.d_model = tf.cast(self.d_model, tf.float32)\n",
        "\n",
        "    self.warmup_steps = warmup_steps\n",
        "\n",
        "  def __call__(self, step):\n",
        "    arg1 = tf.math.rsqrt(step)\n",
        "    arg2 = step * (self.warmup_steps ** -1.5)\n",
        "\n",
        "    return tf.math.rsqrt(self.d_model) * tf.math.minimum(arg1, arg2)"
      ],
      "metadata": {
        "id": "-8m4YneW5gHp"
      },
      "execution_count": 47,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "learning_rate = CustomSchedule(d_model)\n",
        "\n",
        "optimizer = tf.keras.optimizers.Adam(learning_rate, beta_1=0.9, beta_2=0.98,\n",
        "                                     epsilon=1e-9)"
      ],
      "metadata": {
        "id": "UjwNcH466eqo"
      },
      "execution_count": 48,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "temp_learning_rate_schedule = CustomSchedule(d_model)\n",
        "\n",
        "plt.plot(temp_learning_rate_schedule(tf.range(40000, dtype=tf.float32)))\n",
        "plt.ylabel('Learning Rate')\n",
        "plt.xlabel('Train Step')"
      ],
      "metadata": {
        "id": "ffJ1c5W5-y9A",
        "outputId": "7ad3f886-744d-4b39-eb85-73b4a0b842d6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 296
        }
      },
      "execution_count": 49,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0.5, 0, 'Train Step')"
            ]
          },
          "metadata": {},
          "execution_count": 49
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "loss_object = tf.keras.losses.SparseCategoricalCrossentropy(\n",
        "    from_logits=True, reduction='none')"
      ],
      "metadata": {
        "id": "yKqFR3MN--MI"
      },
      "execution_count": 50,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def loss_function(real, pred):\n",
        "  mask = tf.math.logical_not(tf.math.equal(real, 0))\n",
        "  loss_ = loss_object(real, pred)\n",
        "\n",
        "  mask = tf.cast(mask, dtype=loss_.dtype)\n",
        "  loss_ *= mask\n",
        "\n",
        "  return tf.reduce_sum(loss_)/tf.reduce_sum(mask)\n",
        "\n",
        "\n",
        "def accuracy_function(real, pred):\n",
        "  accuracies = tf.equal(real, tf.argmax(pred, axis=2))\n",
        "\n",
        "  mask = tf.math.logical_not(tf.math.equal(real, 0))\n",
        "  accuracies = tf.math.logical_and(mask, accuracies)\n",
        "\n",
        "  accuracies = tf.cast(accuracies, dtype=tf.float32)\n",
        "  mask = tf.cast(mask, dtype=tf.float32)\n",
        "  return tf.reduce_sum(accuracies)/tf.reduce_sum(mask)"
      ],
      "metadata": {
        "id": "jyC2ImI3_A9y"
      },
      "execution_count": 51,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_loss = tf.keras.metrics.Mean(name='train_loss')\n",
        "train_accuracy = tf.keras.metrics.Mean(name='train_accuracy')"
      ],
      "metadata": {
        "id": "7gUPutKo_FqB"
      },
      "execution_count": 52,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "transformer = Transformer(\n",
        "    num_layers=num_layers,\n",
        "    d_model=d_model,\n",
        "    num_heads=num_heads,\n",
        "    dff=dff,\n",
        "    input_vocab_size=tokenizers.pt.get_vocab_size().numpy(),\n",
        "    target_vocab_size=tokenizers.en.get_vocab_size().numpy(),\n",
        "    rate=dropout_rate)"
      ],
      "metadata": {
        "id": "5s75WQe__HAI"
      },
      "execution_count": 53,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "checkpoint_path = './checkpoints/train'\n",
        "\n",
        "ckpt = tf.train.Checkpoint(transformer=transformer,\n",
        "                           optimizer=optimizer)\n",
        "\n",
        "ckpt_manager = tf.train.CheckpointManager(ckpt, checkpoint_path, max_to_keep=5)\n",
        "\n",
        "# if a checkpoint exists, restore the latest checkpoint.\n",
        "if ckpt_manager.latest_checkpoint:\n",
        "  ckpt.restore(ckpt_manager.latest_checkpoint)\n",
        "  print('Latest checkpoint restored!!')"
      ],
      "metadata": {
        "id": "9EbMUzco_IPW"
      },
      "execution_count": 54,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "EPOCHS = 20"
      ],
      "metadata": {
        "id": "qCrq-8QQ_KMC"
      },
      "execution_count": 55,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_step_signature = [\n",
        "    tf.TensorSpec(shape=(None, None), dtype=tf.int64),\n",
        "    tf.TensorSpec(shape=(None, None), dtype=tf.int64),\n",
        "]\n",
        "\n",
        "\n",
        "@tf.function(input_signature=train_step_signature)\n",
        "def train_step(inp, tar):\n",
        "  tar_inp = tar[:, :-1]\n",
        "  tar_real = tar[:, 1:]\n",
        "\n",
        "  with tf.GradientTape() as tape:\n",
        "    predictions, _ = transformer([inp, tar_inp],\n",
        "                                 training = True)\n",
        "    loss = loss_function(tar_real, predictions)\n",
        "\n",
        "  gradients = tape.gradient(loss, transformer.trainable_variables)\n",
        "  optimizer.apply_gradients(zip(gradients, transformer.trainable_variables))\n",
        "\n",
        "  train_loss(loss)\n",
        "  train_accuracy(accuracy_function(tar_real, predictions))"
      ],
      "metadata": {
        "id": "DwV8CP5u_X6x"
      },
      "execution_count": 56,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "for epoch in range(EPOCHS):\n",
        "  start = time.time()\n",
        "\n",
        "  train_loss.reset_states()\n",
        "  train_accuracy.reset_states()\n",
        "\n",
        "  # inp -> portuguese, tar -> english\n",
        "  for (batch, (inp, tar)) in enumerate(train_batches):\n",
        "    train_step(inp, tar)\n",
        "\n",
        "    if batch % 50 == 0:\n",
        "      print(f'Epoch {epoch + 1} Batch {batch} Loss {train_loss.result():.4f} Accuracy {train_accuracy.result():.4f}')\n",
        "\n",
        "  if (epoch + 1) % 5 == 0:\n",
        "    ckpt_save_path = ckpt_manager.save()\n",
        "    print(f'Saving checkpoint for epoch {epoch+1} at {ckpt_save_path}')\n",
        "\n",
        "  print(f'Epoch {epoch + 1} Loss {train_loss.result():.4f} Accuracy {train_accuracy.result():.4f}')\n",
        "\n",
        "  print(f'Time taken for 1 epoch: {time.time() - start:.2f} secs\\n')"
      ],
      "metadata": {
        "id": "nZSnqOFx_YsB",
        "outputId": "eb746585-7192-4d28-ddb8-10da6b9924ea",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": 57,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1 Batch 0 Loss 8.8513 Accuracy 0.0000\n",
            "Epoch 1 Batch 50 Loss 8.7748 Accuracy 0.0111\n",
            "Epoch 1 Batch 100 Loss 8.6558 Accuracy 0.0350\n",
            "Epoch 1 Batch 150 Loss 8.5323 Accuracy 0.0425\n",
            "Epoch 1 Batch 200 Loss 8.3868 Accuracy 0.0468\n",
            "Epoch 1 Batch 250 Loss 8.2163 Accuracy 0.0544\n",
            "Epoch 1 Batch 300 Loss 8.0239 Accuracy 0.0619\n",
            "Epoch 1 Batch 350 Loss 7.8276 Accuracy 0.0693\n",
            "Epoch 1 Batch 400 Loss 7.6422 Accuracy 0.0759\n",
            "Epoch 1 Batch 450 Loss 7.4756 Accuracy 0.0820\n",
            "Epoch 1 Batch 500 Loss 7.3303 Accuracy 0.0878\n",
            "Epoch 1 Batch 550 Loss 7.1985 Accuracy 0.0946\n",
            "Epoch 1 Batch 600 Loss 7.0754 Accuracy 0.1018\n",
            "Epoch 1 Batch 650 Loss 6.9596 Accuracy 0.1087\n",
            "Epoch 1 Batch 700 Loss 6.8511 Accuracy 0.1149\n",
            "Epoch 1 Loss 6.8463 Accuracy 0.1152\n",
            "Time taken for 1 epoch: 83.27 secs\n",
            "\n",
            "Epoch 2 Batch 0 Loss 5.3399 Accuracy 0.1967\n",
            "Epoch 2 Batch 50 Loss 5.3443 Accuracy 0.2034\n",
            "Epoch 2 Batch 100 Loss 5.3110 Accuracy 0.2062\n",
            "Epoch 2 Batch 150 Loss 5.2606 Accuracy 0.2111\n",
            "Epoch 2 Batch 200 Loss 5.2250 Accuracy 0.2151\n",
            "Epoch 2 Batch 250 Loss 5.1919 Accuracy 0.2187\n",
            "Epoch 2 Batch 300 Loss 5.1570 Accuracy 0.2220\n",
            "Epoch 2 Batch 350 Loss 5.1296 Accuracy 0.2249\n",
            "Epoch 2 Batch 400 Loss 5.1055 Accuracy 0.2274\n",
            "Epoch 2 Batch 450 Loss 5.0822 Accuracy 0.2298\n",
            "Epoch 2 Batch 500 Loss 5.0588 Accuracy 0.2322\n",
            "Epoch 2 Batch 550 Loss 5.0352 Accuracy 0.2345\n",
            "Epoch 2 Batch 600 Loss 5.0132 Accuracy 0.2366\n",
            "Epoch 2 Batch 650 Loss 4.9932 Accuracy 0.2383\n",
            "Epoch 2 Loss 4.9735 Accuracy 0.2399\n",
            "Time taken for 1 epoch: 70.30 secs\n",
            "\n",
            "Epoch 3 Batch 0 Loss 4.5685 Accuracy 0.2828\n",
            "Epoch 3 Batch 50 Loss 4.6527 Accuracy 0.2690\n",
            "Epoch 3 Batch 100 Loss 4.6519 Accuracy 0.2701\n",
            "Epoch 3 Batch 150 Loss 4.6407 Accuracy 0.2710\n",
            "Epoch 3 Batch 200 Loss 4.6316 Accuracy 0.2711\n",
            "Epoch 3 Batch 250 Loss 4.6141 Accuracy 0.2731\n",
            "Epoch 3 Batch 300 Loss 4.6005 Accuracy 0.2742\n",
            "Epoch 3 Batch 350 Loss 4.5871 Accuracy 0.2756\n",
            "Epoch 3 Batch 400 Loss 4.5737 Accuracy 0.2769\n",
            "Epoch 3 Batch 450 Loss 4.5570 Accuracy 0.2785\n",
            "Epoch 3 Batch 500 Loss 4.5414 Accuracy 0.2804\n",
            "Epoch 3 Batch 550 Loss 4.5274 Accuracy 0.2821\n",
            "Epoch 3 Batch 600 Loss 4.5108 Accuracy 0.2840\n",
            "Epoch 3 Batch 650 Loss 4.4948 Accuracy 0.2859\n",
            "Epoch 3 Batch 700 Loss 4.4765 Accuracy 0.2877\n",
            "Epoch 3 Loss 4.4765 Accuracy 0.2877\n",
            "Time taken for 1 epoch: 68.48 secs\n",
            "\n",
            "Epoch 4 Batch 0 Loss 4.2778 Accuracy 0.3267\n",
            "Epoch 4 Batch 50 Loss 4.1788 Accuracy 0.3190\n",
            "Epoch 4 Batch 100 Loss 4.1460 Accuracy 0.3229\n",
            "Epoch 4 Batch 150 Loss 4.1424 Accuracy 0.3243\n",
            "Epoch 4 Batch 200 Loss 4.1266 Accuracy 0.3264\n",
            "Epoch 4 Batch 250 Loss 4.1063 Accuracy 0.3286\n",
            "Epoch 4 Batch 300 Loss 4.0873 Accuracy 0.3310\n",
            "Epoch 4 Batch 350 Loss 4.0687 Accuracy 0.3336\n",
            "Epoch 4 Batch 400 Loss 4.0483 Accuracy 0.3362\n",
            "Epoch 4 Batch 450 Loss 4.0314 Accuracy 0.3385\n",
            "Epoch 4 Batch 500 Loss 4.0156 Accuracy 0.3406\n",
            "Epoch 4 Batch 550 Loss 3.9990 Accuracy 0.3426\n",
            "Epoch 4 Batch 600 Loss 3.9792 Accuracy 0.3450\n",
            "Epoch 4 Batch 650 Loss 3.9660 Accuracy 0.3470\n",
            "Epoch 4 Batch 700 Loss 3.9497 Accuracy 0.3492\n",
            "Epoch 4 Loss 3.9485 Accuracy 0.3494\n",
            "Time taken for 1 epoch: 68.72 secs\n",
            "\n",
            "Epoch 5 Batch 0 Loss 3.6878 Accuracy 0.3819\n",
            "Epoch 5 Batch 50 Loss 3.6475 Accuracy 0.3835\n",
            "Epoch 5 Batch 100 Loss 3.6384 Accuracy 0.3861\n",
            "Epoch 5 Batch 150 Loss 3.6052 Accuracy 0.3900\n",
            "Epoch 5 Batch 200 Loss 3.5866 Accuracy 0.3915\n",
            "Epoch 5 Batch 250 Loss 3.5756 Accuracy 0.3936\n",
            "Epoch 5 Batch 300 Loss 3.5587 Accuracy 0.3957\n",
            "Epoch 5 Batch 350 Loss 3.5461 Accuracy 0.3973\n",
            "Epoch 5 Batch 400 Loss 3.5319 Accuracy 0.3994\n",
            "Epoch 5 Batch 450 Loss 3.5112 Accuracy 0.4020\n",
            "Epoch 5 Batch 500 Loss 3.5016 Accuracy 0.4031\n",
            "Epoch 5 Batch 550 Loss 3.4908 Accuracy 0.4045\n",
            "Epoch 5 Batch 600 Loss 3.4795 Accuracy 0.4059\n",
            "Epoch 5 Batch 650 Loss 3.4683 Accuracy 0.4075\n",
            "Epoch 5 Batch 700 Loss 3.4565 Accuracy 0.4090\n",
            "Saving checkpoint for epoch 5 at ./checkpoints/train/ckpt-1\n",
            "Epoch 5 Loss 3.4565 Accuracy 0.4090\n",
            "Time taken for 1 epoch: 68.29 secs\n",
            "\n",
            "Epoch 6 Batch 0 Loss 2.9731 Accuracy 0.4737\n",
            "Epoch 6 Batch 50 Loss 3.1677 Accuracy 0.4415\n",
            "Epoch 6 Batch 100 Loss 3.1600 Accuracy 0.4423\n",
            "Epoch 6 Batch 150 Loss 3.1442 Accuracy 0.4453\n",
            "Epoch 6 Batch 200 Loss 3.1381 Accuracy 0.4458\n",
            "Epoch 6 Batch 250 Loss 3.1324 Accuracy 0.4465\n",
            "Epoch 6 Batch 300 Loss 3.1256 Accuracy 0.4474\n",
            "Epoch 6 Batch 350 Loss 3.1168 Accuracy 0.4486\n",
            "Epoch 6 Batch 400 Loss 3.1092 Accuracy 0.4498\n",
            "Epoch 6 Batch 450 Loss 3.1013 Accuracy 0.4510\n",
            "Epoch 6 Batch 500 Loss 3.0919 Accuracy 0.4522\n",
            "Epoch 6 Batch 550 Loss 3.0835 Accuracy 0.4534\n",
            "Epoch 6 Batch 600 Loss 3.0749 Accuracy 0.4547\n",
            "Epoch 6 Batch 650 Loss 3.0679 Accuracy 0.4555\n",
            "Epoch 6 Batch 700 Loss 3.0603 Accuracy 0.4564\n",
            "Epoch 6 Loss 3.0598 Accuracy 0.4565\n",
            "Time taken for 1 epoch: 68.97 secs\n",
            "\n",
            "Epoch 7 Batch 0 Loss 2.8807 Accuracy 0.4930\n",
            "Epoch 7 Batch 50 Loss 2.8177 Accuracy 0.4823\n",
            "Epoch 7 Batch 100 Loss 2.7906 Accuracy 0.4872\n",
            "Epoch 7 Batch 150 Loss 2.7866 Accuracy 0.4882\n",
            "Epoch 7 Batch 200 Loss 2.7796 Accuracy 0.4896\n",
            "Epoch 7 Batch 250 Loss 2.7759 Accuracy 0.4902\n",
            "Epoch 7 Batch 300 Loss 2.7650 Accuracy 0.4920\n",
            "Epoch 7 Batch 350 Loss 2.7548 Accuracy 0.4937\n",
            "Epoch 7 Batch 400 Loss 2.7424 Accuracy 0.4954\n",
            "Epoch 7 Batch 450 Loss 2.7358 Accuracy 0.4959\n",
            "Epoch 7 Batch 500 Loss 2.7267 Accuracy 0.4970\n",
            "Epoch 7 Batch 550 Loss 2.7193 Accuracy 0.4982\n",
            "Epoch 7 Batch 600 Loss 2.7148 Accuracy 0.4990\n",
            "Epoch 7 Batch 650 Loss 2.7071 Accuracy 0.5003\n",
            "Epoch 7 Loss 2.7035 Accuracy 0.5008\n",
            "Time taken for 1 epoch: 67.78 secs\n",
            "\n",
            "Epoch 8 Batch 0 Loss 2.3174 Accuracy 0.5434\n",
            "Epoch 8 Batch 50 Loss 2.4558 Accuracy 0.5290\n",
            "Epoch 8 Batch 100 Loss 2.4579 Accuracy 0.5307\n",
            "Epoch 8 Batch 150 Loss 2.4525 Accuracy 0.5323\n",
            "Epoch 8 Batch 200 Loss 2.4603 Accuracy 0.5315\n",
            "Epoch 8 Batch 250 Loss 2.4562 Accuracy 0.5322\n",
            "Epoch 8 Batch 300 Loss 2.4538 Accuracy 0.5330\n",
            "Epoch 8 Batch 350 Loss 2.4487 Accuracy 0.5339\n",
            "Epoch 8 Batch 400 Loss 2.4461 Accuracy 0.5340\n",
            "Epoch 8 Batch 450 Loss 2.4421 Accuracy 0.5349\n",
            "Epoch 8 Batch 500 Loss 2.4374 Accuracy 0.5355\n",
            "Epoch 8 Batch 550 Loss 2.4358 Accuracy 0.5358\n",
            "Epoch 8 Batch 600 Loss 2.4323 Accuracy 0.5365\n",
            "Epoch 8 Batch 650 Loss 2.4302 Accuracy 0.5369\n",
            "Epoch 8 Loss 2.4275 Accuracy 0.5374\n",
            "Time taken for 1 epoch: 68.08 secs\n",
            "\n",
            "Epoch 9 Batch 0 Loss 2.3380 Accuracy 0.5372\n",
            "Epoch 9 Batch 50 Loss 2.2437 Accuracy 0.5591\n",
            "Epoch 9 Batch 100 Loss 2.2310 Accuracy 0.5614\n",
            "Epoch 9 Batch 150 Loss 2.2394 Accuracy 0.5607\n",
            "Epoch 9 Batch 200 Loss 2.2359 Accuracy 0.5613\n",
            "Epoch 9 Batch 250 Loss 2.2332 Accuracy 0.5620\n",
            "Epoch 9 Batch 300 Loss 2.2319 Accuracy 0.5628\n",
            "Epoch 9 Batch 350 Loss 2.2344 Accuracy 0.5625\n",
            "Epoch 9 Batch 400 Loss 2.2361 Accuracy 0.5621\n",
            "Epoch 9 Batch 450 Loss 2.2350 Accuracy 0.5623\n",
            "Epoch 9 Batch 500 Loss 2.2291 Accuracy 0.5632\n",
            "Epoch 9 Batch 550 Loss 2.2298 Accuracy 0.5631\n",
            "Epoch 9 Batch 600 Loss 2.2303 Accuracy 0.5632\n",
            "Epoch 9 Batch 650 Loss 2.2311 Accuracy 0.5633\n",
            "Epoch 9 Loss 2.2298 Accuracy 0.5637\n",
            "Time taken for 1 epoch: 68.31 secs\n",
            "\n",
            "Epoch 10 Batch 0 Loss 2.3006 Accuracy 0.5281\n",
            "Epoch 10 Batch 50 Loss 2.0563 Accuracy 0.5886\n",
            "Epoch 10 Batch 100 Loss 2.0530 Accuracy 0.5890\n",
            "Epoch 10 Batch 150 Loss 2.0577 Accuracy 0.5885\n",
            "Epoch 10 Batch 200 Loss 2.0626 Accuracy 0.5875\n",
            "Epoch 10 Batch 250 Loss 2.0682 Accuracy 0.5866\n",
            "Epoch 10 Batch 300 Loss 2.0689 Accuracy 0.5866\n",
            "Epoch 10 Batch 350 Loss 2.0706 Accuracy 0.5865\n",
            "Epoch 10 Batch 400 Loss 2.0721 Accuracy 0.5861\n",
            "Epoch 10 Batch 450 Loss 2.0728 Accuracy 0.5860\n",
            "Epoch 10 Batch 500 Loss 2.0748 Accuracy 0.5857\n",
            "Epoch 10 Batch 550 Loss 2.0715 Accuracy 0.5861\n",
            "Epoch 10 Batch 600 Loss 2.0737 Accuracy 0.5857\n",
            "Epoch 10 Batch 650 Loss 2.0741 Accuracy 0.5859\n",
            "Epoch 10 Batch 700 Loss 2.0747 Accuracy 0.5858\n",
            "Saving checkpoint for epoch 10 at ./checkpoints/train/ckpt-2\n",
            "Epoch 10 Loss 2.0743 Accuracy 0.5859\n",
            "Time taken for 1 epoch: 69.09 secs\n",
            "\n",
            "Epoch 11 Batch 0 Loss 1.8995 Accuracy 0.6065\n",
            "Epoch 11 Batch 50 Loss 1.9370 Accuracy 0.6033\n",
            "Epoch 11 Batch 100 Loss 1.9290 Accuracy 0.6049\n",
            "Epoch 11 Batch 150 Loss 1.9289 Accuracy 0.6061\n",
            "Epoch 11 Batch 200 Loss 1.9352 Accuracy 0.6051\n",
            "Epoch 11 Batch 250 Loss 1.9425 Accuracy 0.6038\n",
            "Epoch 11 Batch 300 Loss 1.9456 Accuracy 0.6037\n",
            "Epoch 11 Batch 350 Loss 1.9485 Accuracy 0.6035\n",
            "Epoch 11 Batch 400 Loss 1.9476 Accuracy 0.6035\n",
            "Epoch 11 Batch 450 Loss 1.9466 Accuracy 0.6038\n",
            "Epoch 11 Batch 500 Loss 1.9490 Accuracy 0.6034\n",
            "Epoch 11 Batch 550 Loss 1.9505 Accuracy 0.6032\n",
            "Epoch 11 Batch 600 Loss 1.9500 Accuracy 0.6031\n",
            "Epoch 11 Batch 650 Loss 1.9490 Accuracy 0.6034\n",
            "Epoch 11 Batch 700 Loss 1.9508 Accuracy 0.6033\n",
            "Epoch 11 Loss 1.9503 Accuracy 0.6033\n",
            "Time taken for 1 epoch: 68.82 secs\n",
            "\n",
            "Epoch 12 Batch 0 Loss 1.8364 Accuracy 0.6140\n",
            "Epoch 12 Batch 50 Loss 1.8291 Accuracy 0.6204\n",
            "Epoch 12 Batch 100 Loss 1.8157 Accuracy 0.6221\n",
            "Epoch 12 Batch 150 Loss 1.8181 Accuracy 0.6220\n",
            "Epoch 12 Batch 200 Loss 1.8226 Accuracy 0.6213\n",
            "Epoch 12 Batch 250 Loss 1.8303 Accuracy 0.6205\n",
            "Epoch 12 Batch 300 Loss 1.8293 Accuracy 0.6205\n",
            "Epoch 12 Batch 350 Loss 1.8324 Accuracy 0.6202\n",
            "Epoch 12 Batch 400 Loss 1.8331 Accuracy 0.6202\n",
            "Epoch 12 Batch 450 Loss 1.8373 Accuracy 0.6195\n",
            "Epoch 12 Batch 500 Loss 1.8370 Accuracy 0.6196\n",
            "Epoch 12 Batch 550 Loss 1.8404 Accuracy 0.6191\n",
            "Epoch 12 Batch 600 Loss 1.8455 Accuracy 0.6183\n",
            "Epoch 12 Batch 650 Loss 1.8452 Accuracy 0.6184\n",
            "Epoch 12 Batch 700 Loss 1.8485 Accuracy 0.6180\n",
            "Epoch 12 Loss 1.8487 Accuracy 0.6180\n",
            "Time taken for 1 epoch: 68.69 secs\n",
            "\n",
            "Epoch 13 Batch 0 Loss 1.8177 Accuracy 0.6089\n",
            "Epoch 13 Batch 50 Loss 1.6969 Accuracy 0.6406\n",
            "Epoch 13 Batch 100 Loss 1.7162 Accuracy 0.6371\n",
            "Epoch 13 Batch 150 Loss 1.7210 Accuracy 0.6364\n",
            "Epoch 13 Batch 200 Loss 1.7237 Accuracy 0.6353\n",
            "Epoch 13 Batch 250 Loss 1.7237 Accuracy 0.6355\n",
            "Epoch 13 Batch 300 Loss 1.7277 Accuracy 0.6350\n",
            "Epoch 13 Batch 350 Loss 1.7354 Accuracy 0.6336\n",
            "Epoch 13 Batch 400 Loss 1.7399 Accuracy 0.6328\n",
            "Epoch 13 Batch 450 Loss 1.7398 Accuracy 0.6334\n",
            "Epoch 13 Batch 500 Loss 1.7430 Accuracy 0.6331\n",
            "Epoch 13 Batch 550 Loss 1.7462 Accuracy 0.6325\n",
            "Epoch 13 Batch 600 Loss 1.7505 Accuracy 0.6320\n",
            "Epoch 13 Batch 650 Loss 1.7536 Accuracy 0.6316\n",
            "Epoch 13 Batch 700 Loss 1.7574 Accuracy 0.6312\n",
            "Epoch 13 Loss 1.7569 Accuracy 0.6312\n",
            "Time taken for 1 epoch: 68.58 secs\n",
            "\n",
            "Epoch 14 Batch 0 Loss 1.5848 Accuracy 0.6644\n",
            "Epoch 14 Batch 50 Loss 1.6331 Accuracy 0.6508\n",
            "Epoch 14 Batch 100 Loss 1.6386 Accuracy 0.6490\n",
            "Epoch 14 Batch 150 Loss 1.6419 Accuracy 0.6483\n",
            "Epoch 14 Batch 200 Loss 1.6446 Accuracy 0.6482\n",
            "Epoch 14 Batch 250 Loss 1.6516 Accuracy 0.6475\n",
            "Epoch 14 Batch 300 Loss 1.6545 Accuracy 0.6471\n",
            "Epoch 14 Batch 350 Loss 1.6545 Accuracy 0.6474\n",
            "Epoch 14 Batch 400 Loss 1.6593 Accuracy 0.6464\n",
            "Epoch 14 Batch 450 Loss 1.6638 Accuracy 0.6458\n",
            "Epoch 14 Batch 500 Loss 1.6655 Accuracy 0.6456\n",
            "Epoch 14 Batch 550 Loss 1.6695 Accuracy 0.6448\n",
            "Epoch 14 Batch 600 Loss 1.6744 Accuracy 0.6441\n",
            "Epoch 14 Batch 650 Loss 1.6773 Accuracy 0.6437\n",
            "Epoch 14 Loss 1.6794 Accuracy 0.6433\n",
            "Time taken for 1 epoch: 68.40 secs\n",
            "\n",
            "Epoch 15 Batch 0 Loss 1.5551 Accuracy 0.6496\n",
            "Epoch 15 Batch 50 Loss 1.5718 Accuracy 0.6589\n",
            "Epoch 15 Batch 100 Loss 1.5792 Accuracy 0.6574\n",
            "Epoch 15 Batch 150 Loss 1.5767 Accuracy 0.6590\n",
            "Epoch 15 Batch 200 Loss 1.5840 Accuracy 0.6579\n",
            "Epoch 15 Batch 250 Loss 1.5839 Accuracy 0.6580\n",
            "Epoch 15 Batch 300 Loss 1.5931 Accuracy 0.6563\n",
            "Epoch 15 Batch 350 Loss 1.5948 Accuracy 0.6559\n",
            "Epoch 15 Batch 400 Loss 1.5966 Accuracy 0.6556\n",
            "Epoch 15 Batch 450 Loss 1.5990 Accuracy 0.6552\n",
            "Epoch 15 Batch 500 Loss 1.6020 Accuracy 0.6547\n",
            "Epoch 15 Batch 550 Loss 1.6064 Accuracy 0.6540\n",
            "Epoch 15 Batch 600 Loss 1.6108 Accuracy 0.6532\n",
            "Epoch 15 Batch 650 Loss 1.6126 Accuracy 0.6530\n",
            "Epoch 15 Batch 700 Loss 1.6165 Accuracy 0.6526\n",
            "Saving checkpoint for epoch 15 at ./checkpoints/train/ckpt-3\n",
            "Epoch 15 Loss 1.6165 Accuracy 0.6526\n",
            "Time taken for 1 epoch: 68.75 secs\n",
            "\n",
            "Epoch 16 Batch 0 Loss 1.2931 Accuracy 0.7187\n",
            "Epoch 16 Batch 50 Loss 1.4978 Accuracy 0.6713\n",
            "Epoch 16 Batch 100 Loss 1.4979 Accuracy 0.6721\n",
            "Epoch 16 Batch 150 Loss 1.5075 Accuracy 0.6703\n",
            "Epoch 16 Batch 200 Loss 1.5154 Accuracy 0.6690\n",
            "Epoch 16 Batch 250 Loss 1.5189 Accuracy 0.6683\n",
            "Epoch 16 Batch 300 Loss 1.5242 Accuracy 0.6676\n",
            "Epoch 16 Batch 350 Loss 1.5298 Accuracy 0.6666\n",
            "Epoch 16 Batch 400 Loss 1.5328 Accuracy 0.6662\n",
            "Epoch 16 Batch 450 Loss 1.5363 Accuracy 0.6657\n",
            "Epoch 16 Batch 500 Loss 1.5412 Accuracy 0.6649\n",
            "Epoch 16 Batch 550 Loss 1.5460 Accuracy 0.6642\n",
            "Epoch 16 Batch 600 Loss 1.5477 Accuracy 0.6640\n",
            "Epoch 16 Batch 650 Loss 1.5513 Accuracy 0.6635\n",
            "Epoch 16 Loss 1.5555 Accuracy 0.6628\n",
            "Time taken for 1 epoch: 67.88 secs\n",
            "\n",
            "Epoch 17 Batch 0 Loss 1.5840 Accuracy 0.6578\n",
            "Epoch 17 Batch 50 Loss 1.4771 Accuracy 0.6750\n",
            "Epoch 17 Batch 100 Loss 1.4600 Accuracy 0.6778\n",
            "Epoch 17 Batch 150 Loss 1.4642 Accuracy 0.6772\n",
            "Epoch 17 Batch 200 Loss 1.4697 Accuracy 0.6761\n",
            "Epoch 17 Batch 250 Loss 1.4748 Accuracy 0.6756\n",
            "Epoch 17 Batch 300 Loss 1.4791 Accuracy 0.6747\n",
            "Epoch 17 Batch 350 Loss 1.4825 Accuracy 0.6741\n",
            "Epoch 17 Batch 400 Loss 1.4872 Accuracy 0.6732\n",
            "Epoch 17 Batch 450 Loss 1.4882 Accuracy 0.6730\n",
            "Epoch 17 Batch 500 Loss 1.4906 Accuracy 0.6727\n",
            "Epoch 17 Batch 550 Loss 1.4941 Accuracy 0.6723\n",
            "Epoch 17 Batch 600 Loss 1.4963 Accuracy 0.6719\n",
            "Epoch 17 Batch 650 Loss 1.5003 Accuracy 0.6714\n",
            "Epoch 17 Batch 700 Loss 1.5043 Accuracy 0.6707\n",
            "Epoch 17 Loss 1.5043 Accuracy 0.6707\n",
            "Time taken for 1 epoch: 68.43 secs\n",
            "\n",
            "Epoch 18 Batch 0 Loss 1.2439 Accuracy 0.7111\n",
            "Epoch 18 Batch 50 Loss 1.4093 Accuracy 0.6860\n",
            "Epoch 18 Batch 100 Loss 1.3981 Accuracy 0.6872\n",
            "Epoch 18 Batch 150 Loss 1.4110 Accuracy 0.6853\n",
            "Epoch 18 Batch 200 Loss 1.4195 Accuracy 0.6844\n",
            "Epoch 18 Batch 250 Loss 1.4249 Accuracy 0.6830\n",
            "Epoch 18 Batch 300 Loss 1.4249 Accuracy 0.6835\n",
            "Epoch 18 Batch 350 Loss 1.4302 Accuracy 0.6825\n",
            "Epoch 18 Batch 400 Loss 1.4341 Accuracy 0.6819\n",
            "Epoch 18 Batch 450 Loss 1.4358 Accuracy 0.6818\n",
            "Epoch 18 Batch 500 Loss 1.4388 Accuracy 0.6815\n",
            "Epoch 18 Batch 550 Loss 1.4430 Accuracy 0.6807\n",
            "Epoch 18 Batch 600 Loss 1.4466 Accuracy 0.6802\n",
            "Epoch 18 Batch 650 Loss 1.4503 Accuracy 0.6798\n",
            "Epoch 18 Batch 700 Loss 1.4538 Accuracy 0.6792\n",
            "Epoch 18 Loss 1.4543 Accuracy 0.6792\n",
            "Time taken for 1 epoch: 68.30 secs\n",
            "\n",
            "Epoch 19 Batch 0 Loss 1.3061 Accuracy 0.7069\n",
            "Epoch 19 Batch 50 Loss 1.3676 Accuracy 0.6930\n",
            "Epoch 19 Batch 100 Loss 1.3652 Accuracy 0.6935\n",
            "Epoch 19 Batch 150 Loss 1.3649 Accuracy 0.6940\n",
            "Epoch 19 Batch 200 Loss 1.3722 Accuracy 0.6926\n",
            "Epoch 19 Batch 250 Loss 1.3757 Accuracy 0.6922\n",
            "Epoch 19 Batch 300 Loss 1.3825 Accuracy 0.6911\n",
            "Epoch 19 Batch 350 Loss 1.3871 Accuracy 0.6901\n",
            "Epoch 19 Batch 400 Loss 1.3873 Accuracy 0.6900\n",
            "Epoch 19 Batch 450 Loss 1.3904 Accuracy 0.6893\n",
            "Epoch 19 Batch 500 Loss 1.3945 Accuracy 0.6884\n",
            "Epoch 19 Batch 550 Loss 1.3976 Accuracy 0.6879\n",
            "Epoch 19 Batch 600 Loss 1.4013 Accuracy 0.6874\n",
            "Epoch 19 Batch 650 Loss 1.4055 Accuracy 0.6870\n",
            "Epoch 19 Batch 700 Loss 1.4094 Accuracy 0.6863\n",
            "Epoch 19 Loss 1.4102 Accuracy 0.6862\n",
            "Time taken for 1 epoch: 68.43 secs\n",
            "\n",
            "Epoch 20 Batch 0 Loss 1.3297 Accuracy 0.6838\n",
            "Epoch 20 Batch 50 Loss 1.3146 Accuracy 0.7028\n",
            "Epoch 20 Batch 100 Loss 1.3178 Accuracy 0.7011\n",
            "Epoch 20 Batch 150 Loss 1.3219 Accuracy 0.7009\n",
            "Epoch 20 Batch 200 Loss 1.3265 Accuracy 0.7000\n",
            "Epoch 20 Batch 250 Loss 1.3328 Accuracy 0.6984\n",
            "Epoch 20 Batch 300 Loss 1.3405 Accuracy 0.6972\n",
            "Epoch 20 Batch 350 Loss 1.3433 Accuracy 0.6967\n",
            "Epoch 20 Batch 400 Loss 1.3470 Accuracy 0.6959\n",
            "Epoch 20 Batch 450 Loss 1.3514 Accuracy 0.6952\n",
            "Epoch 20 Batch 500 Loss 1.3578 Accuracy 0.6940\n",
            "Epoch 20 Batch 550 Loss 1.3625 Accuracy 0.6933\n",
            "Epoch 20 Batch 600 Loss 1.3654 Accuracy 0.6931\n",
            "Epoch 20 Batch 650 Loss 1.3683 Accuracy 0.6926\n",
            "Epoch 20 Batch 700 Loss 1.3703 Accuracy 0.6922\n",
            "Saving checkpoint for epoch 20 at ./checkpoints/train/ckpt-4\n",
            "Epoch 20 Loss 1.3704 Accuracy 0.6922\n",
            "Time taken for 1 epoch: 68.66 secs\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        ""
      ],
      "metadata": {
        "id": "7FZ3a-e7_bew"
      },
      "execution_count": 57,
      "outputs": []
    }
  ]
}