Chapter_6_1.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Chapter_6_1.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "machine_shape": "hm",
      "authorship_tag": "ABX9TyPU5WOzerQjkBk3mCd6ZiDV"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU",
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "e8674860168743d182954b3032e52116": {
          "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_6b4117b05e0a4393a5fe76d5256547e8",
              "IPY_MODEL_9c01b39de08c482ab249d025d8104d35",
              "IPY_MODEL_c492d3811d754f82bab252c2786cb5aa"
            ],
            "layout": "IPY_MODEL_1aaffb465fa64ad4a849db53fe842cab"
          }
        },
        "6b4117b05e0a4393a5fe76d5256547e8": {
          "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_529be0258c374d5cb71c202f6aa61380",
            "placeholder": "​",
            "style": "IPY_MODEL_bead54fb7f264a6b9e12f701d61f6912",
            "value": ""
          }
        },
        "9c01b39de08c482ab249d025d8104d35": {
          "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_88f1f25e788f46f59e0e9622dcd2624b",
            "max": 9912422,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_a3321acbf3a34ac0a9ab922b10b142cf",
            "value": 9912422
          }
        },
        "c492d3811d754f82bab252c2786cb5aa": {
          "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_7ea0df26ad3f4a5cb59c053f927ccd1d",
            "placeholder": "​",
            "style": "IPY_MODEL_ceb280580d7c4b4f9302ce6f8a0ed743",
            "value": " 9913344/? [00:00&lt;00:00, 48857347.72it/s]"
          }
        },
        "1aaffb465fa64ad4a849db53fe842cab": {
          "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
          }
        },
        "529be0258c374d5cb71c202f6aa61380": {
          "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
          }
        },
        "bead54fb7f264a6b9e12f701d61f6912": {
          "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": ""
          }
        },
        "88f1f25e788f46f59e0e9622dcd2624b": {
          "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
          }
        },
        "a3321acbf3a34ac0a9ab922b10b142cf": {
          "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": ""
          }
        },
        "7ea0df26ad3f4a5cb59c053f927ccd1d": {
          "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
          }
        },
        "ceb280580d7c4b4f9302ce6f8a0ed743": {
          "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": ""
          }
        },
        "a0df41e8802843c3ac11762a6883d112": {
          "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_e8f7ac2588154ee09134b950221c7d0e",
              "IPY_MODEL_8de2c6d0f9a642ce99fee5f3b3eb263d",
              "IPY_MODEL_c1eef64685f5451497e970f0452beda5"
            ],
            "layout": "IPY_MODEL_75864d0a7603438cb3fa2f3a58dd197f"
          }
        },
        "e8f7ac2588154ee09134b950221c7d0e": {
          "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_00b9711e3d6b4596ae903926cae6daec",
            "placeholder": "​",
            "style": "IPY_MODEL_4eca482ec68f4abe805118db7f238afd",
            "value": ""
          }
        },
        "8de2c6d0f9a642ce99fee5f3b3eb263d": {
          "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_f2c61d2964b641b6956d7372d9585393",
            "max": 28881,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_8c9d384ebe8748e19f9a69df129bd77a",
            "value": 28881
          }
        },
        "c1eef64685f5451497e970f0452beda5": {
          "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_957ab52a3c0e4378a6dfcd07ea07ebc0",
            "placeholder": "​",
            "style": "IPY_MODEL_24c062d2da024c9f850abdf01cc944b6",
            "value": " 29696/? [00:00&lt;00:00, 586937.71it/s]"
          }
        },
        "75864d0a7603438cb3fa2f3a58dd197f": {
          "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
          }
        },
        "00b9711e3d6b4596ae903926cae6daec": {
          "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
          }
        },
        "4eca482ec68f4abe805118db7f238afd": {
          "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": ""
          }
        },
        "f2c61d2964b641b6956d7372d9585393": {
          "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
          }
        },
        "8c9d384ebe8748e19f9a69df129bd77a": {
          "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": ""
          }
        },
        "957ab52a3c0e4378a6dfcd07ea07ebc0": {
          "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
          }
        },
        "24c062d2da024c9f850abdf01cc944b6": {
          "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": ""
          }
        },
        "53cf2c8a881844bfa8e78d07db8bab53": {
          "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_bc1f52743d214f7c8d924558f0f3117d",
              "IPY_MODEL_3d279f16d54e41d1a9efc25ff4c9dea8",
              "IPY_MODEL_ddb36cc64a8b4f238005a83f767f2e5c"
            ],
            "layout": "IPY_MODEL_6a7ed8262aed4c50b8540881c2842ea3"
          }
        },
        "bc1f52743d214f7c8d924558f0f3117d": {
          "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_a526012eba2843b3a235bf3dc06891c0",
            "placeholder": "​",
            "style": "IPY_MODEL_de9f5cbb595f49ea87f527272620d88c",
            "value": ""
          }
        },
        "3d279f16d54e41d1a9efc25ff4c9dea8": {
          "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_eeb0fa2a7cb144ac97c3a8560581157b",
            "max": 1648877,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_aef30f18ebd14557bd4dd966f29b72e4",
            "value": 1648877
          }
        },
        "ddb36cc64a8b4f238005a83f767f2e5c": {
          "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_bbd27ed05c424f00a21d4956ed8054a5",
            "placeholder": "​",
            "style": "IPY_MODEL_5c76b681247846cdafad502f1be227f3",
            "value": " 1649664/? [00:00&lt;00:00, 3037784.44it/s]"
          }
        },
        "6a7ed8262aed4c50b8540881c2842ea3": {
          "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
          }
        },
        "a526012eba2843b3a235bf3dc06891c0": {
          "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
          }
        },
        "de9f5cbb595f49ea87f527272620d88c": {
          "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": ""
          }
        },
        "eeb0fa2a7cb144ac97c3a8560581157b": {
          "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
          }
        },
        "aef30f18ebd14557bd4dd966f29b72e4": {
          "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": ""
          }
        },
        "bbd27ed05c424f00a21d4956ed8054a5": {
          "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
          }
        },
        "5c76b681247846cdafad502f1be227f3": {
          "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": ""
          }
        },
        "a437c33153d34129aaf147453c002583": {
          "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_ae78bedc199e480983109f5d5a5ec260",
              "IPY_MODEL_61da11ad5d4344fb98aaf10f50e0af14",
              "IPY_MODEL_a826a1ff4d434a799e123843c9c2b94b"
            ],
            "layout": "IPY_MODEL_282dcdb9e5104a75be70ef70dae7e9a4"
          }
        },
        "ae78bedc199e480983109f5d5a5ec260": {
          "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_dfac4163a8d44c20ae9c8fe475764136",
            "placeholder": "​",
            "style": "IPY_MODEL_077c9cd06ccb41bf9660cd35dee06443",
            "value": ""
          }
        },
        "61da11ad5d4344fb98aaf10f50e0af14": {
          "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_f86db3df1d3e46a2bdd95bcb77f2f002",
            "max": 4542,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_3af7730c06f448c7a0f2ecfaa53f178a",
            "value": 4542
          }
        },
        "a826a1ff4d434a799e123843c9c2b94b": {
          "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_087137fabbbc4889bc53820dfcb24cd1",
            "placeholder": "​",
            "style": "IPY_MODEL_a7c1f018cab94cd7801fcd3e205ab388",
            "value": " 5120/? [00:00&lt;00:00, 169484.22it/s]"
          }
        },
        "282dcdb9e5104a75be70ef70dae7e9a4": {
          "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
          }
        },
        "dfac4163a8d44c20ae9c8fe475764136": {
          "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
          }
        },
        "077c9cd06ccb41bf9660cd35dee06443": {
          "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": ""
          }
        },
        "f86db3df1d3e46a2bdd95bcb77f2f002": {
          "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
          }
        },
        "3af7730c06f448c7a0f2ecfaa53f178a": {
          "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": ""
          }
        },
        "087137fabbbc4889bc53820dfcb24cd1": {
          "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
          }
        },
        "a7c1f018cab94cd7801fcd3e205ab388": {
          "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": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "from torch.autograd import Variable\n",
        "import torch.optim as optim\n",
        "import torchvision\n",
        "from torchvision import datasets\n",
        "from torchvision import transforms\n",
        "import numpy as np\n",
        "\n",
        "torch.manual_seed(8)"
      ],
      "metadata": {
        "id": "csqGOVCeL6WR",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4f239efd-570b-429e-b3f9-32c90290d9eb"
      },
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<torch._C.Generator at 0x7ffadab231d0>"
            ]
          },
          "metadata": {},
          "execution_count": 1
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## MLP"
      ],
      "metadata": {
        "id": "aGd0qU1uJ9Kn"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "id": "Yz2LcA2Y4XaX",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7e4fc299-19f5-41cf-bb6f-a4f8d46f811f"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "4"
            ]
          },
          "metadata": {},
          "execution_count": 2
        }
      ],
      "source": [
        "class Dropout(nn.Module):\n",
        "    def __init__(self, dropout_rate=0.5):\n",
        "        super(Dropout, self).__init__()\n",
        "        self.dropout_rate = dropout_rate\n",
        "        # define the scaling factor to remaining nodes to compensate for dropped nodes\n",
        "        if self.dropout_rate < 1:\n",
        "            self.scaling_fct = 1.0 / (1.0-dropout_rate)\n",
        "        else:\n",
        "            self.scaling_fct = 0.0\n",
        "\n",
        "    def forward(self, input):\n",
        "        # only apply dropout during training mode\n",
        "        if not self.training:\n",
        "            return input\n",
        "        # indicator tensor of same size to label the selected nodes\n",
        "        selected_ind = torch.Tensor(input.shape).uniform_(0,1) > self.dropout_rate\n",
        "        # use GPU if available\n",
        "        if input.is_cuda:\n",
        "            selected_ind = Variable(selected_ind.type(torch.cuda.FloatTensor), requires_grad=False)\n",
        "        else:\n",
        "            selected_ind = Variable(selected_ind.type(torch.FloatTensor), requires_grad=False)\n",
        "        # drop out some nodes and scale up the rest\n",
        "        return torch.mul(selected_ind, input) * self.scaling_fct"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class MLP(nn.Module):\n",
        "    def __init__(self, hidden_layers=[800, 800], dropout_rates=[0, 0]):\n",
        "        super(MLP, self).__init__()\n",
        "        self.model = nn.Sequential()\n",
        "        self.model.add_module(\"input\", nn.Linear(28*28, hidden_layers[0]))\n",
        "        self.model.add_module(\"dropout_input\", Dropout(dropout_rate=dropout_rates[0]))\n",
        "        self.model.add_module(\"tanh\", nn.Tanh())\n",
        "        # add hidden layers\n",
        "        for i,d in enumerate(hidden_layers[:-1]):\n",
        "            self.model.add_module(\"dropout_hidden_\"+str(i+1), Dropout(dropout_rate=dropout_rates[1]))\n",
        "            self.model.add_module(\"hidden_\"+str(i+1), nn.Linear(hidden_layers[i], hidden_layers[i+1]))\n",
        "            self.model.add_module(\"tanh_hidden_\"+str(i+1), nn.Tanh())\n",
        "        self.model.add_module(\"final\",nn.Linear(hidden_layers[-1], 10))        \n",
        "        \n",
        "    def forward(self, x):\n",
        "        # flatten to one-dimensional array\n",
        "        x = x.view(x.shape[0], 28*28)\n",
        "        x = self.model(x)\n",
        "        return x"
      ],
      "metadata": {
        "id": "p76Ao5MSe90C"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model = MLP()\n",
        "print(model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GbyLwJaWJR7-",
        "outputId": "f087852a-f8af-4b97-f5aa-e006c851a745"
      },
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "MLP(\n",
            "  (model): Sequential(\n",
            "    (input): Linear(in_features=784, out_features=800, bias=True)\n",
            "    (dropout_input): Dropout()\n",
            "    (tanh): Tanh()\n",
            "    (dropout_hidden_1): Dropout()\n",
            "    (hidden_1): Linear(in_features=800, out_features=800, bias=True)\n",
            "    (tanh_hidden_1): Tanh()\n",
            "    (final): Linear(in_features=800, out_features=10, bias=True)\n",
            "  )\n",
            ")\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class MLPClassifier:\n",
        "    def __init__(self, hidden_layers=[800, 800], dropout_rates=[0, 0], batch_size=128, \\\n",
        "                 max_epoch=10, lr=0.1):\n",
        "        # save and initiate utility variables\n",
        "        self.hidden_layers = hidden_layers\n",
        "        self.dropout_rates = dropout_rates\n",
        "        self.batch_size = batch_size\n",
        "        self.max_epoch = max_epoch\n",
        "        self.model = MLP(hidden_layers=hidden_layers, dropout_rates=dropout_rates)\n",
        "        self.model.cuda()\n",
        "        self.criterion = nn.CrossEntropyLoss().cuda()\n",
        "        self.optimizer = optim.SGD(self.model.parameters(), lr=lr)\n",
        "        self.loss = []\n",
        "        self.test_accuracy = []\n",
        "        \n",
        "    def fit(self, trainset, testset, verbose=True, test_score_freq=5):\n",
        "        # prepare training and test sets\n",
        "        trainloader = torch.utils.data.DataLoader(trainset, batch_size=self.batch_size, shuffle=True)\n",
        "        testloader = torch.utils.data.DataLoader(testset, batch_size=len(testset), shuffle=False)\n",
        "        X_test, y_test = iter(testloader).next()\n",
        "        X_test = X_test.cuda()\n",
        "        # start training procedure, best run on GPU\n",
        "        for epoch in range(self.max_epoch):\n",
        "            running_loss = 0\n",
        "            for i, training_batch in enumerate(trainloader, 0):\n",
        "                X_train, y_train = training_batch\n",
        "                # send to GPU if available\n",
        "                X_train, y_train = Variable(X_train).cuda(), Variable(y_train).cuda()\n",
        "                # clear gradients\n",
        "                self.optimizer.zero_grad()\n",
        "                # make prediction\n",
        "                y_pred = self.model(X_train)\n",
        "                # calculate loss of current batch\n",
        "                loss = self.criterion(y_pred, y_train)\n",
        "                # perform back propagation to calculate gradients automatically\n",
        "                loss.backward()\n",
        "                # update weights\n",
        "                self.optimizer.step()\n",
        "                # add loss of different batches within the same epoch\n",
        "                running_loss += loss.item()\n",
        "            # store average loss per batch for all epochs\n",
        "            self.loss.append(running_loss / len(trainloader))\n",
        "            if verbose:\n",
        "                print('Epoch {} loss: {}'.format(epoch+1, self.loss[-1]))\n",
        "            if (epoch+1) % test_score_freq == 0:\n",
        "                # get predictions for the test set\n",
        "                y_test_pred = self.predict(X_test).cpu()\n",
        "                self.y_test_pred = y_test_pred\n",
        "                self.y_test = y_test\n",
        "                # store test set accuracy\n",
        "                self.test_accuracy.append(np.mean((y_test == y_test_pred).numpy()))\n",
        "                if verbose:\n",
        "                    print('Test accuracy: {}'.format(self.test_accuracy[-1]))\n",
        "        return self\n",
        "    \n",
        "    # make class predictions for test set\n",
        "    def predict(self, x):\n",
        "        model = self.model.eval()\n",
        "        outputs = model(Variable(x))\n",
        "        _, pred = torch.max(outputs.data, 1)\n",
        "        model = self.model.train()\n",
        "        return pred"
      ],
      "metadata": {
        "id": "w8EDr21gkMiv"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# load and transform MNIST\n",
        "transform = transforms.Compose([transforms.ToTensor(), \\\n",
        "                                transforms.Normalize([0.5], [0.5])])\n",
        "trainset = datasets.MNIST(root='data/', train=True, download=True, transform=transform)\n",
        "testset = datasets.MNIST(root='data/', train=False, transform=transform)\n",
        "# define three FCN models\n",
        "hidden_layers = [300, 300]\n",
        "model_list = [MLPClassifier(hidden_layers, dropout_rates=[0, 0], max_epoch=500), \n",
        "        MLPClassifier(hidden_layers, dropout_rates=[0, 0.5], max_epoch=500),\n",
        "        MLPClassifier(hidden_layers, dropout_rates=[0.2, 0.5], max_epoch=500)]\n",
        "# start training\n",
        "[tmp_model.fit(trainset, testset,verbose=True) for tmp_model in model_list]\n",
        "# save torch models\n",
        "for ind, tmp_model in enumerate(model_list):\n",
        "    torch.save(tmp_model.model, 'mnist_mlp_'+str(ind)+'.pth')\n",
        "    tmp_model.test_accuracy = list(map(str, tmp_model.test_accuracy))\n",
        "# save test accuracy\n",
        "from google.colab import files\n",
        "open(\"mlp_test_accuracy.txt\",\"w\").write('\\n'.join([','.join(tmp_mlp.test_accuracy) for tmp_mlp in model_list]))\n",
        "files.download('mlp_test_accuracy.txt')"
      ],
      "metadata": {
        "id": "Vo3KECkrA1F3"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Load saved test errors to plot figures.\n",
        "mlp_test_accuracy = [accuracy_array.split(',') for accuracy_array in open(\"mlp_test_accuracy.txt\",\"r\").read().split('\\n')]\n",
        "mlp_test_accuracy = np.array(mlp_test_accuracy, dtype='f')\n",
        "\n",
        "labels = ['MLP no dropout', \n",
        "          'MLP 50% dropout in hidden layers', \n",
        "          'MLP 50% dropout in hidden layers + 20% in input layer']\n",
        "\n",
        "plt.figure(figsize=(8, 7))\n",
        "for i, r in enumerate(mlp_test_accuracy):\n",
        "    plt.plot(np.array(range(1, len(r)+1))*5, r, '.-', label=labels[i], alpha=0.6);\n",
        "plt.legend(loc=4);\n",
        "plt.xlabel('Epochs');\n",
        "plt.ylabel('Accuracy in test set');\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 442
        },
        "id": "b-iGFpk4Lu1B",
        "outputId": "03bbbab3-4969-4ccc-a81d-ca513d348f89"
      },
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 576x504 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## LeNet"
      ],
      "metadata": {
        "id": "uh82kCfBIjkK"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "class Flatten(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(Flatten, self).__init__()\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = x.view(x.size(0), -1)\n",
        "        return x\n",
        "\n",
        "class LeNet(nn.Module):\n",
        "    def __init__(self, dropout_rate=0.5):\n",
        "        super(LeNet, self).__init__()\n",
        "        self.model = nn.Sequential()\n",
        "        self.model.add_module('conv1', nn.Conv2d(1, 20, kernel_size=5, padding=2))\n",
        "        self.model.add_module('dropout1', nn.Dropout2d(p=dropout_rate))\n",
        "        self.model.add_module('maxpool1', nn.MaxPool2d(2, stride=2))\n",
        "        self.model.add_module('conv2', nn.Conv2d(20, 50, kernel_size=5, padding=2))\n",
        "        self.model.add_module('dropout2', nn.Dropout2d(p=dropout_rate))\n",
        "        self.model.add_module('maxpool2', nn.MaxPool2d(2, stride=2))\n",
        "        self.model.add_module('flatten', Flatten())\n",
        "        self.model.add_module('dense3', nn.Linear(50*7*7, 500))\n",
        "        self.model.add_module('relu3', nn.ReLU())\n",
        "        self.model.add_module('dropout3', nn.Dropout(p=dropout_rate))\n",
        "        self.model.add_module('final', nn.Linear(500, 10))\n",
        "        \n",
        "    def forward(self, x):\n",
        "        return self.model(x)"
      ],
      "metadata": {
        "id": "aCKeMECaE6PU"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model = LeNet()\n",
        "print(model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EXKqWpnfJbtV",
        "outputId": "243947e7-c6ad-4d5a-c695-2bda1b4ec867"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "LeNet(\n",
            "  (model): Sequential(\n",
            "    (conv1): Conv2d(1, 20, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
            "    (dropout1): Dropout2d(p=0.5, inplace=False)\n",
            "    (maxpool1): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (conv2): Conv2d(20, 50, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
            "    (dropout2): Dropout2d(p=0.5, inplace=False)\n",
            "    (maxpool2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (flatten): Flatten()\n",
            "    (dense3): Linear(in_features=2450, out_features=500, bias=True)\n",
            "    (relu3): ReLU()\n",
            "    (dropout3): Dropout(p=0.5, inplace=False)\n",
            "    (final): Linear(in_features=500, out_features=10, bias=True)\n",
            "  )\n",
            ")\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class LeNetClassifier:\n",
        "    def __init__(self, dropout_rate=0.5, batch_size=128, max_epoch=300, lr=0.01):\n",
        "        # save and initiate utility variables\n",
        "        self.batch_size = batch_size\n",
        "        self.max_epoch = max_epoch\n",
        "        self.lr = lr\n",
        "        self.model = LeNet(dropout_rate)\n",
        "        self.model.cuda()\n",
        "        self.criterion = nn.CrossEntropyLoss().cuda()\n",
        "        self.optimizer = optim.SGD(self.model.parameters(), lr=lr)\n",
        "        self.loss = []\n",
        "        self.test_accuracy = []\n",
        "        \n",
        "    def fit(self, trainset, testset, verbose=True, test_score_freq=5):\n",
        "        # prepare training and test sets\n",
        "        trainloader = torch.utils.data.DataLoader(trainset, batch_size=self.batch_size, shuffle=True)\n",
        "        testloader = torch.utils.data.DataLoader(testset, batch_size=len(testset), shuffle=False)\n",
        "        X_test, y_test = iter(testloader).next()\n",
        "        X_test = X_test.cuda()\n",
        "        # start training procedure, best run on GPU\n",
        "        for epoch in range(self.max_epoch):\n",
        "            running_loss = 0\n",
        "            for i, training_batch in enumerate(trainloader, 0):\n",
        "                X_train, y_train = training_batch\n",
        "                # send to GPU if available\n",
        "                X_train, y_train = Variable(X_train).cuda(), Variable(y_train).cuda()\n",
        "                # clear gradients\n",
        "                self.optimizer.zero_grad()\n",
        "                # make prediction\n",
        "                y_pred = self.model(X_train)\n",
        "                # calculate loss of current batch\n",
        "                loss = self.criterion(y_pred, y_train)\n",
        "                # perform back propagation to calculate gradients automatically\n",
        "                loss.backward()\n",
        "                # update weights\n",
        "                self.optimizer.step()\n",
        "                # add loss of different batches within the same epoch\n",
        "                running_loss += loss.item()\n",
        "            # store average loss per batch for all epochs\n",
        "            self.loss.append(running_loss / len(trainloader))\n",
        "            if verbose:\n",
        "                print('Epoch {} loss: {}'.format(epoch+1, self.loss[-1]))\n",
        "            if (epoch+1) % test_score_freq == 0:\n",
        "                # get predictions for the test set\n",
        "                y_test_pred = self.predict(X_test).cpu()\n",
        "                self.y_test_pred = y_test_pred\n",
        "                self.y_test = y_test\n",
        "                # store test set accuracy\n",
        "                self.test_accuracy.append(np.mean((y_test == y_test_pred).numpy()))\n",
        "                if verbose:\n",
        "                    print('Test accuracy: {}'.format(self.test_accuracy[-1]))\n",
        "        return self\n",
        "    \n",
        "    # make class predictions for test set\n",
        "    def predict(self, x):\n",
        "        model = self.model.eval()\n",
        "        outputs = model(Variable(x))\n",
        "        _, pred = torch.max(outputs.data, 1)\n",
        "        model = self.model.train()\n",
        "        return pred"
      ],
      "metadata": {
        "id": "MCTETVcAGYP0"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# load MNIST\n",
        "transform = transforms.Compose([transforms.ToTensor(), \\\n",
        "                                transforms.Normalize([0.5], [0.5])])\n",
        "\n",
        "trainset = datasets.MNIST(root='data/', train=True, download=True, transform=transform)\n",
        "testset = datasets.MNIST(root='data/', train=False, transform=transform)\n",
        "\n",
        "# Define networks\n",
        "lenet_list = [LeNetClassifier(dropout_rate=0, max_epoch=500),\n",
        "          LeNetClassifier(dropout_rate=0.5, max_epoch=500)]\n",
        "        \n",
        "# start training\n",
        "[lenet.fit(trainset, testset,verbose=True) for lenet in lenet_list]\n",
        "\n",
        "# Save torch models\n",
        "for ind, lenet in enumerate(lenet_list):\n",
        "    torch.save(lenet.model, 'mnist_lenet_'+str(ind)+'.pth')\n",
        "    # Prepare to save errors\n",
        "    lenet.test_accuracy = list(map(str, lenet.test_accuracy))\n",
        "\n",
        "# Save test accuracy to plot figures\n",
        "from google.colab import files\n",
        "open(\"lenet_test_accuracy.txt\",\"w\").write('\\n'.join([','.join(lenet.test_accuracy) for lenet in lenet_list])) \n",
        "files.download('lenet_test_accuracy.txt')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000,
          "referenced_widgets": [
            "e8674860168743d182954b3032e52116",
            "6b4117b05e0a4393a5fe76d5256547e8",
            "9c01b39de08c482ab249d025d8104d35",
            "c492d3811d754f82bab252c2786cb5aa",
            "1aaffb465fa64ad4a849db53fe842cab",
            "529be0258c374d5cb71c202f6aa61380",
            "bead54fb7f264a6b9e12f701d61f6912",
            "88f1f25e788f46f59e0e9622dcd2624b",
            "a3321acbf3a34ac0a9ab922b10b142cf",
            "7ea0df26ad3f4a5cb59c053f927ccd1d",
            "ceb280580d7c4b4f9302ce6f8a0ed743",
            "a0df41e8802843c3ac11762a6883d112",
            "e8f7ac2588154ee09134b950221c7d0e",
            "8de2c6d0f9a642ce99fee5f3b3eb263d",
            "c1eef64685f5451497e970f0452beda5",
            "75864d0a7603438cb3fa2f3a58dd197f",
            "00b9711e3d6b4596ae903926cae6daec",
            "4eca482ec68f4abe805118db7f238afd",
            "f2c61d2964b641b6956d7372d9585393",
            "8c9d384ebe8748e19f9a69df129bd77a",
            "957ab52a3c0e4378a6dfcd07ea07ebc0",
            "24c062d2da024c9f850abdf01cc944b6",
            "53cf2c8a881844bfa8e78d07db8bab53",
            "bc1f52743d214f7c8d924558f0f3117d",
            "3d279f16d54e41d1a9efc25ff4c9dea8",
            "ddb36cc64a8b4f238005a83f767f2e5c",
            "6a7ed8262aed4c50b8540881c2842ea3",
            "a526012eba2843b3a235bf3dc06891c0",
            "de9f5cbb595f49ea87f527272620d88c",
            "eeb0fa2a7cb144ac97c3a8560581157b",
            "aef30f18ebd14557bd4dd966f29b72e4",
            "bbd27ed05c424f00a21d4956ed8054a5",
            "5c76b681247846cdafad502f1be227f3",
            "a437c33153d34129aaf147453c002583",
            "ae78bedc199e480983109f5d5a5ec260",
            "61da11ad5d4344fb98aaf10f50e0af14",
            "a826a1ff4d434a799e123843c9c2b94b",
            "282dcdb9e5104a75be70ef70dae7e9a4",
            "dfac4163a8d44c20ae9c8fe475764136",
            "077c9cd06ccb41bf9660cd35dee06443",
            "f86db3df1d3e46a2bdd95bcb77f2f002",
            "3af7730c06f448c7a0f2ecfaa53f178a",
            "087137fabbbc4889bc53820dfcb24cd1",
            "a7c1f018cab94cd7801fcd3e205ab388"
          ]
        },
        "id": "LodmXF_4G3tW",
        "outputId": "3001e808-719a-424b-a781-e76228321714"
      },
      "execution_count": 5,
      "outputs": [
        {
          "metadata": {
            "tags": null
          },
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz\n",
            "Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to data/MNIST/raw/train-images-idx3-ubyte.gz\n"
          ]
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "e8674860168743d182954b3032e52116",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "  0%|          | 0/9912422 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "metadata": {
            "tags": null
          },
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Extracting data/MNIST/raw/train-images-idx3-ubyte.gz to data/MNIST/raw\n",
            "\n",
            "Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz\n",
            "Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to data/MNIST/raw/train-labels-idx1-ubyte.gz\n"
          ]
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "a0df41e8802843c3ac11762a6883d112",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "  0%|          | 0/28881 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "metadata": {
            "tags": null
          },
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Extracting data/MNIST/raw/train-labels-idx1-ubyte.gz to data/MNIST/raw\n",
            "\n",
            "Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz\n",
            "Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to data/MNIST/raw/t10k-images-idx3-ubyte.gz\n"
          ]
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "53cf2c8a881844bfa8e78d07db8bab53",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "  0%|          | 0/1648877 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "metadata": {
            "tags": null
          },
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Extracting data/MNIST/raw/t10k-images-idx3-ubyte.gz to data/MNIST/raw\n",
            "\n",
            "Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz\n",
            "Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to data/MNIST/raw/t10k-labels-idx1-ubyte.gz\n"
          ]
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "a437c33153d34129aaf147453c002583",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "  0%|          | 0/4542 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Extracting data/MNIST/raw/t10k-labels-idx1-ubyte.gz to data/MNIST/raw\n",
            "\n",
            "Epoch 1 loss: 0.8734755944659207\n",
            "Epoch 2 loss: 0.25499455889722683\n",
            "Epoch 3 loss: 0.17691441645214298\n",
            "Epoch 4 loss: 0.13335478362212302\n",
            "Epoch 5 loss: 0.1074233212267984\n",
            "Test accuracy: 0.9719\n",
            "Epoch 6 loss: 0.09015832378713688\n",
            "Epoch 7 loss: 0.07769234400830352\n",
            "Epoch 8 loss: 0.06894374141560942\n",
            "Epoch 9 loss: 0.06194850289896289\n",
            "Epoch 10 loss: 0.056854922630226436\n",
            "Test accuracy: 0.9833\n",
            "Epoch 11 loss: 0.052070346436520885\n",
            "Epoch 12 loss: 0.048230401526239\n",
            "Epoch 13 loss: 0.044735696005907016\n",
            "Epoch 14 loss: 0.04155575814368978\n",
            "Epoch 15 loss: 0.03947590064031006\n",
            "Test accuracy: 0.9853\n",
            "Epoch 16 loss: 0.03734043417925408\n",
            "Epoch 17 loss: 0.03513472151062858\n",
            "Epoch 18 loss: 0.032754330455498304\n",
            "Epoch 19 loss: 0.031057928516062847\n",
            "Epoch 20 loss: 0.02994254684554282\n",
            "Test accuracy: 0.9876\n",
            "Epoch 21 loss: 0.028243451235768224\n",
            "Epoch 22 loss: 0.026917095030985598\n",
            "Epoch 23 loss: 0.025541069304773102\n",
            "Epoch 24 loss: 0.024463233457548597\n",
            "Epoch 25 loss: 0.023183644527315236\n",
            "Test accuracy: 0.9872\n",
            "Epoch 26 loss: 0.022344336189430897\n",
            "Epoch 27 loss: 0.021032398700146185\n",
            "Epoch 28 loss: 0.020314977267262602\n",
            "Epoch 29 loss: 0.019402817469278872\n",
            "Epoch 30 loss: 0.018663828747956628\n",
            "Test accuracy: 0.9873\n",
            "Epoch 31 loss: 0.017791628293238326\n",
            "Epoch 32 loss: 0.017012981742100197\n",
            "Epoch 33 loss: 0.016563261533702123\n",
            "Epoch 34 loss: 0.01565039604812908\n",
            "Epoch 35 loss: 0.014748369362939998\n",
            "Test accuracy: 0.9891\n",
            "Epoch 36 loss: 0.014156498630761481\n",
            "Epoch 37 loss: 0.013731383913403143\n",
            "Epoch 38 loss: 0.012870678464809595\n",
            "Epoch 39 loss: 0.012385633124260982\n",
            "Epoch 40 loss: 0.011618607168791812\n",
            "Test accuracy: 0.9873\n",
            "Epoch 41 loss: 0.0116727897759104\n",
            "Epoch 42 loss: 0.011235563803763528\n",
            "Epoch 43 loss: 0.010403810025154074\n",
            "Epoch 44 loss: 0.009920095517632685\n",
            "Epoch 45 loss: 0.009715466507534777\n",
            "Test accuracy: 0.9898\n",
            "Epoch 46 loss: 0.009270596006802762\n",
            "Epoch 47 loss: 0.008816246607297186\n",
            "Epoch 48 loss: 0.008636509935467888\n",
            "Epoch 49 loss: 0.008025428297920034\n",
            "Epoch 50 loss: 0.0076814645499577805\n",
            "Test accuracy: 0.9895\n",
            "Epoch 51 loss: 0.007548760082922689\n",
            "Epoch 52 loss: 0.007370951194648764\n",
            "Epoch 53 loss: 0.006814685986073973\n",
            "Epoch 54 loss: 0.006633034228771202\n",
            "Epoch 55 loss: 0.006296814987784686\n",
            "Test accuracy: 0.9898\n",
            "Epoch 56 loss: 0.006150177571266495\n",
            "Epoch 57 loss: 0.005881705941487926\n",
            "Epoch 58 loss: 0.005632944476287087\n",
            "Epoch 59 loss: 0.005548419151542041\n",
            "Epoch 60 loss: 0.005229494547216234\n",
            "Test accuracy: 0.9893\n",
            "Epoch 61 loss: 0.005176115924322944\n",
            "Epoch 62 loss: 0.004974988532123039\n",
            "Epoch 63 loss: 0.004740974224204305\n",
            "Epoch 64 loss: 0.004449652437251488\n",
            "Epoch 65 loss: 0.0043480494754639136\n",
            "Test accuracy: 0.99\n",
            "Epoch 66 loss: 0.00421733496949446\n",
            "Epoch 67 loss: 0.004147282982521108\n",
            "Epoch 68 loss: 0.0038872183709839003\n",
            "Epoch 69 loss: 0.003803929634221486\n",
            "Epoch 70 loss: 0.0036678645301047763\n",
            "Test accuracy: 0.9895\n",
            "Epoch 71 loss: 0.0033677027419191856\n",
            "Epoch 72 loss: 0.0034279730771658327\n",
            "Epoch 73 loss: 0.0032877660889080716\n",
            "Epoch 74 loss: 0.0031238482666955487\n",
            "Epoch 75 loss: 0.0030927048400331138\n",
            "Test accuracy: 0.9897\n",
            "Epoch 76 loss: 0.0028629825984122607\n",
            "Epoch 77 loss: 0.002729592125703107\n",
            "Epoch 78 loss: 0.0028176816049846573\n",
            "Epoch 79 loss: 0.0027042474273268552\n",
            "Epoch 80 loss: 0.0026197045182987703\n",
            "Test accuracy: 0.9902\n",
            "Epoch 81 loss: 0.0024543248844814043\n",
            "Epoch 82 loss: 0.002403876190741501\n",
            "Epoch 83 loss: 0.0023040794950036424\n",
            "Epoch 84 loss: 0.0022872818945678694\n",
            "Epoch 85 loss: 0.0021933861984573463\n",
            "Test accuracy: 0.9894\n",
            "Epoch 86 loss: 0.0021062783340071844\n",
            "Epoch 87 loss: 0.002026651689909579\n",
            "Epoch 88 loss: 0.0019230202792573775\n",
            "Epoch 89 loss: 0.001838905247111893\n",
            "Epoch 90 loss: 0.0017443431266900603\n",
            "Test accuracy: 0.9891\n",
            "Epoch 91 loss: 0.0017772447094750671\n",
            "Epoch 92 loss: 0.0016434392888411216\n",
            "Epoch 93 loss: 0.0015930592703751784\n",
            "Epoch 94 loss: 0.0015803825434186616\n",
            "Epoch 95 loss: 0.0016004900571095\n",
            "Test accuracy: 0.9896\n",
            "Epoch 96 loss: 0.0015381282031825155\n",
            "Epoch 97 loss: 0.0013886507551087988\n",
            "Epoch 98 loss: 0.0014875900104400147\n",
            "Epoch 99 loss: 0.0013822491651296423\n",
            "Epoch 100 loss: 0.0013946593594628134\n",
            "Test accuracy: 0.989\n",
            "Epoch 101 loss: 0.0013251168281147093\n",
            "Epoch 102 loss: 0.0012666496233368638\n",
            "Epoch 103 loss: 0.0011995334019868022\n",
            "Epoch 104 loss: 0.0012676538223063008\n",
            "Epoch 105 loss: 0.001220522083947795\n",
            "Test accuracy: 0.9903\n",
            "Epoch 106 loss: 0.0011821169093216353\n",
            "Epoch 107 loss: 0.0011250372211359986\n",
            "Epoch 108 loss: 0.0010641637408800523\n",
            "Epoch 109 loss: 0.001068597865413933\n",
            "Epoch 110 loss: 0.0010496152002726301\n",
            "Test accuracy: 0.9899\n",
            "Epoch 111 loss: 0.0010451461062453544\n",
            "Epoch 112 loss: 0.0010460075069565018\n",
            "Epoch 113 loss: 0.0009951985673684162\n",
            "Epoch 114 loss: 0.0009402800699432707\n",
            "Epoch 115 loss: 0.0009644912551865535\n",
            "Test accuracy: 0.9898\n",
            "Epoch 116 loss: 0.0008924795382588889\n",
            "Epoch 117 loss: 0.0009112088759030118\n",
            "Epoch 118 loss: 0.000861992768306997\n",
            "Epoch 119 loss: 0.0008756484472107684\n",
            "Epoch 120 loss: 0.0008129431507632713\n",
            "Test accuracy: 0.9897\n",
            "Epoch 121 loss: 0.0007978421026234652\n",
            "Epoch 122 loss: 0.0007856784649077555\n",
            "Epoch 123 loss: 0.0008127821247806206\n",
            "Epoch 124 loss: 0.0007677520752079532\n",
            "Epoch 125 loss: 0.0007343102765737721\n",
            "Test accuracy: 0.9897\n",
            "Epoch 126 loss: 0.000743891711143947\n",
            "Epoch 127 loss: 0.000766702914570914\n",
            "Epoch 128 loss: 0.0007116246272653176\n",
            "Epoch 129 loss: 0.0007033823865365073\n",
            "Epoch 130 loss: 0.0007116294140658609\n",
            "Test accuracy: 0.9895\n",
            "Epoch 131 loss: 0.0006857647728457896\n",
            "Epoch 132 loss: 0.0006601599477528622\n",
            "Epoch 133 loss: 0.000658069284513377\n",
            "Epoch 134 loss: 0.0006486577503756731\n",
            "Epoch 135 loss: 0.0006200059939410475\n",
            "Test accuracy: 0.9897\n",
            "Epoch 136 loss: 0.000605843147171291\n",
            "Epoch 137 loss: 0.0006030655387768657\n",
            "Epoch 138 loss: 0.0006017368372891309\n",
            "Epoch 139 loss: 0.0005907516377750719\n",
            "Epoch 140 loss: 0.0005633915580493504\n",
            "Test accuracy: 0.9899\n",
            "Epoch 141 loss: 0.0005625185870103412\n",
            "Epoch 142 loss: 0.0005582482192062877\n",
            "Epoch 143 loss: 0.0005456628401542573\n",
            "Epoch 144 loss: 0.0005475252686517663\n",
            "Epoch 145 loss: 0.0005465879178745164\n",
            "Test accuracy: 0.99\n",
            "Epoch 146 loss: 0.0005160421596631581\n",
            "Epoch 147 loss: 0.00052488633986701\n",
            "Epoch 148 loss: 0.0005110946867813809\n",
            "Epoch 149 loss: 0.0005006784432789999\n",
            "Epoch 150 loss: 0.0004942564779152495\n",
            "Test accuracy: 0.9903\n",
            "Epoch 151 loss: 0.0005042049822287212\n",
            "Epoch 152 loss: 0.00048614828818648216\n",
            "Epoch 153 loss: 0.00048655093492975935\n",
            "Epoch 154 loss: 0.00046700758449470765\n",
            "Epoch 155 loss: 0.00044972100448130984\n",
            "Test accuracy: 0.9897\n",
            "Epoch 156 loss: 0.00045300952980783913\n",
            "Epoch 157 loss: 0.0004504888255586242\n",
            "Epoch 158 loss: 0.00046013967586719227\n",
            "Epoch 159 loss: 0.000436718644248604\n",
            "Epoch 160 loss: 0.0004272147441735064\n",
            "Test accuracy: 0.9895\n",
            "Epoch 161 loss: 0.00043491415709198574\n",
            "Epoch 162 loss: 0.00042099983455778754\n",
            "Epoch 163 loss: 0.0004133253698032769\n",
            "Epoch 164 loss: 0.00041383580909067666\n",
            "Epoch 165 loss: 0.0004089354135589783\n",
            "Test accuracy: 0.9894\n",
            "Epoch 166 loss: 0.0004074375262359348\n",
            "Epoch 167 loss: 0.0003938697121104661\n",
            "Epoch 168 loss: 0.00039679538479180267\n",
            "Epoch 169 loss: 0.0003838010168019653\n",
            "Epoch 170 loss: 0.00037580780831677194\n",
            "Test accuracy: 0.9899\n",
            "Epoch 171 loss: 0.00036324140816773186\n",
            "Epoch 172 loss: 0.00038223261377843805\n",
            "Epoch 173 loss: 0.0003671931177752979\n",
            "Epoch 174 loss: 0.0003715348260179693\n",
            "Epoch 175 loss: 0.00037599450917389835\n",
            "Test accuracy: 0.9896\n",
            "Epoch 176 loss: 0.0003552726835833983\n",
            "Epoch 177 loss: 0.00034881151799542415\n",
            "Epoch 178 loss: 0.0003504139598481376\n",
            "Epoch 179 loss: 0.000346984335594053\n",
            "Epoch 180 loss: 0.00034227897832043693\n",
            "Test accuracy: 0.9898\n",
            "Epoch 181 loss: 0.0003399632933151748\n",
            "Epoch 182 loss: 0.00033036186135338515\n",
            "Epoch 183 loss: 0.0003323611821085071\n",
            "Epoch 184 loss: 0.0003251354505383307\n",
            "Epoch 185 loss: 0.00032039980800936685\n",
            "Test accuracy: 0.9899\n",
            "Epoch 186 loss: 0.0003221541809190447\n",
            "Epoch 187 loss: 0.0003162279599215439\n",
            "Epoch 188 loss: 0.0003137067856642776\n",
            "Epoch 189 loss: 0.0003091926857823075\n",
            "Epoch 190 loss: 0.0003078801653804852\n",
            "Test accuracy: 0.9901\n",
            "Epoch 191 loss: 0.0003032891258207296\n",
            "Epoch 192 loss: 0.00029907327216627705\n",
            "Epoch 193 loss: 0.00030186948880193883\n",
            "Epoch 194 loss: 0.0002964080231303049\n",
            "Epoch 195 loss: 0.00029353922318012585\n",
            "Test accuracy: 0.9901\n",
            "Epoch 196 loss: 0.0002900353878118691\n",
            "Epoch 197 loss: 0.00028891432019761206\n",
            "Epoch 198 loss: 0.00028869155302459876\n",
            "Epoch 199 loss: 0.00028248359864744085\n",
            "Epoch 200 loss: 0.00028254082202371186\n",
            "Test accuracy: 0.9902\n",
            "Epoch 201 loss: 0.0002772420422401282\n",
            "Epoch 202 loss: 0.0002745975291587972\n",
            "Epoch 203 loss: 0.0002771715538545435\n",
            "Epoch 204 loss: 0.0002706285308268212\n",
            "Epoch 205 loss: 0.0002633307795534733\n",
            "Test accuracy: 0.9896\n",
            "Epoch 206 loss: 0.0002658996579306567\n",
            "Epoch 207 loss: 0.00026289963406520706\n",
            "Epoch 208 loss: 0.00026412074877181464\n",
            "Epoch 209 loss: 0.0002604106179690991\n",
            "Epoch 210 loss: 0.0002585133446674459\n",
            "Test accuracy: 0.9899\n",
            "Epoch 211 loss: 0.00025519301494774494\n",
            "Epoch 212 loss: 0.00025085340514399625\n",
            "Epoch 213 loss: 0.00024725325529715593\n",
            "Epoch 214 loss: 0.0002457588952912667\n",
            "Epoch 215 loss: 0.000248919400854299\n",
            "Test accuracy: 0.9898\n",
            "Epoch 216 loss: 0.0002459869800073407\n",
            "Epoch 217 loss: 0.00024222757386061725\n",
            "Epoch 218 loss: 0.00024089907110426208\n",
            "Epoch 219 loss: 0.0002395353031344425\n",
            "Epoch 220 loss: 0.00023847814812769411\n",
            "Test accuracy: 0.9897\n",
            "Epoch 221 loss: 0.00023200802983166198\n",
            "Epoch 222 loss: 0.0002311512432314786\n",
            "Epoch 223 loss: 0.00022999930671449545\n",
            "Epoch 224 loss: 0.000228949570173388\n",
            "Epoch 225 loss: 0.00022434524824053285\n",
            "Test accuracy: 0.9901\n",
            "Epoch 226 loss: 0.00022508231887341239\n",
            "Epoch 227 loss: 0.00022374639488923826\n",
            "Epoch 228 loss: 0.00022317862074289693\n",
            "Epoch 229 loss: 0.00021944194278926272\n",
            "Epoch 230 loss: 0.00021613141201250478\n",
            "Test accuracy: 0.9897\n",
            "Epoch 231 loss: 0.0002142868674872046\n",
            "Epoch 232 loss: 0.00021471255782964328\n",
            "Epoch 233 loss: 0.00021205593563743737\n",
            "Epoch 234 loss: 0.00021172581279385139\n",
            "Epoch 235 loss: 0.00020781882893503408\n",
            "Test accuracy: 0.9899\n",
            "Epoch 236 loss: 0.0002066519539830881\n",
            "Epoch 237 loss: 0.00020800267903628558\n",
            "Epoch 238 loss: 0.00020584093705025393\n",
            "Epoch 239 loss: 0.00020273355550907697\n",
            "Epoch 240 loss: 0.00020074669722921625\n",
            "Test accuracy: 0.99\n",
            "Epoch 241 loss: 0.00020333577084316986\n",
            "Epoch 242 loss: 0.00019845381696294873\n",
            "Epoch 243 loss: 0.00019897413263269298\n",
            "Epoch 244 loss: 0.00019735056576231541\n",
            "Epoch 245 loss: 0.00019612344880373022\n",
            "Test accuracy: 0.99\n",
            "Epoch 246 loss: 0.00019470531649835187\n",
            "Epoch 247 loss: 0.00019197622500377078\n",
            "Epoch 248 loss: 0.00019127577309184982\n",
            "Epoch 249 loss: 0.00019149663321882252\n",
            "Epoch 250 loss: 0.00018985764201448325\n",
            "Test accuracy: 0.9901\n",
            "Epoch 251 loss: 0.0001880373309299401\n",
            "Epoch 252 loss: 0.0001869795120895053\n",
            "Epoch 253 loss: 0.0001829841884149639\n",
            "Epoch 254 loss: 0.0001849121788706855\n",
            "Epoch 255 loss: 0.00018171450576530355\n",
            "Test accuracy: 0.9898\n",
            "Epoch 256 loss: 0.0001813306985434992\n",
            "Epoch 257 loss: 0.00018025952950231284\n",
            "Epoch 258 loss: 0.00017767613003624472\n",
            "Epoch 259 loss: 0.0001762203951068071\n",
            "Epoch 260 loss: 0.00017701759107502688\n",
            "Test accuracy: 0.9898\n",
            "Epoch 261 loss: 0.0001765170718344438\n",
            "Epoch 262 loss: 0.00017420804384074532\n",
            "Epoch 263 loss: 0.00017382390735131619\n",
            "Epoch 264 loss: 0.00017141299477230058\n",
            "Epoch 265 loss: 0.00016918058155491913\n",
            "Test accuracy: 0.9901\n",
            "Epoch 266 loss: 0.0001707480957599625\n",
            "Epoch 267 loss: 0.00016751998237183693\n",
            "Epoch 268 loss: 0.00016694804081629654\n",
            "Epoch 269 loss: 0.00016675432388530375\n",
            "Epoch 270 loss: 0.00016585435909251123\n",
            "Test accuracy: 0.9896\n",
            "Epoch 271 loss: 0.00016348502243944705\n",
            "Epoch 272 loss: 0.0001648623473380185\n",
            "Epoch 273 loss: 0.0001613051654799756\n",
            "Epoch 274 loss: 0.00016064056580265726\n",
            "Epoch 275 loss: 0.00015924631767219344\n",
            "Test accuracy: 0.9896\n",
            "Epoch 276 loss: 0.00016059858159478958\n",
            "Epoch 277 loss: 0.00015793779916147303\n",
            "Epoch 278 loss: 0.00015703183739041474\n",
            "Epoch 279 loss: 0.00015636943538923552\n",
            "Epoch 280 loss: 0.00015670904464264835\n",
            "Test accuracy: 0.9898\n",
            "Epoch 281 loss: 0.00015557556298032932\n",
            "Epoch 282 loss: 0.00015427907643875274\n",
            "Epoch 283 loss: 0.0001528119090911019\n",
            "Epoch 284 loss: 0.0001529646584567129\n",
            "Epoch 285 loss: 0.00015084623509247205\n",
            "Test accuracy: 0.9898\n",
            "Epoch 286 loss: 0.00015151951379605756\n",
            "Epoch 287 loss: 0.00014708533120771455\n",
            "Epoch 288 loss: 0.00015128302492231052\n",
            "Epoch 289 loss: 0.0001474813864676487\n",
            "Epoch 290 loss: 0.00014827774398647135\n",
            "Test accuracy: 0.9899\n",
            "Epoch 291 loss: 0.00014509627836077092\n",
            "Epoch 292 loss: 0.00014607655122278817\n",
            "Epoch 293 loss: 0.00014498933935997297\n",
            "Epoch 294 loss: 0.00014503879830531498\n",
            "Epoch 295 loss: 0.00014378180856177333\n",
            "Test accuracy: 0.9898\n",
            "Epoch 296 loss: 0.00014292478349896338\n",
            "Epoch 297 loss: 0.00014254539774153442\n",
            "Epoch 298 loss: 0.00014150091231837725\n",
            "Epoch 299 loss: 0.00013925262378613543\n",
            "Epoch 300 loss: 0.00013943422729003198\n",
            "Test accuracy: 0.9896\n",
            "Epoch 301 loss: 0.0001383411693889535\n",
            "Epoch 302 loss: 0.00013850700062916917\n",
            "Epoch 303 loss: 0.00013745807310759677\n",
            "Epoch 304 loss: 0.00013762184401482698\n",
            "Epoch 305 loss: 0.0001359222042587472\n",
            "Test accuracy: 0.9897\n",
            "Epoch 306 loss: 0.00013528893435605087\n",
            "Epoch 307 loss: 0.0001342591089771521\n",
            "Epoch 308 loss: 0.0001332523312014216\n",
            "Epoch 309 loss: 0.00013372164921301733\n",
            "Epoch 310 loss: 0.00013194315712827763\n",
            "Test accuracy: 0.9897\n",
            "Epoch 311 loss: 0.0001319010319521248\n",
            "Epoch 312 loss: 0.00013092831960733096\n",
            "Epoch 313 loss: 0.00012942743826530157\n",
            "Epoch 314 loss: 0.00012844168976522242\n",
            "Epoch 315 loss: 0.00012927547351655196\n",
            "Test accuracy: 0.9895\n",
            "Epoch 316 loss: 0.00012889899177328173\n",
            "Epoch 317 loss: 0.00012730648697968695\n",
            "Epoch 318 loss: 0.00012652584525733515\n",
            "Epoch 319 loss: 0.00012584851744843137\n",
            "Epoch 320 loss: 0.00012535928309787619\n",
            "Test accuracy: 0.9898\n",
            "Epoch 321 loss: 0.00012555442421612468\n",
            "Epoch 322 loss: 0.00012498910079512788\n",
            "Epoch 323 loss: 0.00012460129424975044\n",
            "Epoch 324 loss: 0.0001229900999072564\n",
            "Epoch 325 loss: 0.00012329987759822077\n",
            "Test accuracy: 0.9898\n",
            "Epoch 326 loss: 0.00012292772232682213\n",
            "Epoch 327 loss: 0.0001218748336884534\n",
            "Epoch 328 loss: 0.00012146053092046694\n",
            "Epoch 329 loss: 0.00011965184831545159\n",
            "Epoch 330 loss: 0.00012044444017011943\n",
            "Test accuracy: 0.9897\n",
            "Epoch 331 loss: 0.0001197379070248962\n",
            "Epoch 332 loss: 0.00011922535149434227\n",
            "Epoch 333 loss: 0.00011895211399121429\n",
            "Epoch 334 loss: 0.00011741355365011414\n",
            "Epoch 335 loss: 0.00011748686547085712\n",
            "Test accuracy: 0.9897\n",
            "Epoch 336 loss: 0.00011627184971451831\n",
            "Epoch 337 loss: 0.00011604210696245387\n",
            "Epoch 338 loss: 0.00011576356498903952\n",
            "Epoch 339 loss: 0.00011484638585211726\n",
            "Epoch 340 loss: 0.00011408907741074636\n",
            "Test accuracy: 0.9898\n",
            "Epoch 341 loss: 0.00011335645015641039\n",
            "Epoch 342 loss: 0.00011351038264130697\n",
            "Epoch 343 loss: 0.00011261673697067626\n",
            "Epoch 344 loss: 0.00011251988710901913\n",
            "Epoch 345 loss: 0.00011198378780254701\n",
            "Test accuracy: 0.9899\n",
            "Epoch 346 loss: 0.00011151526052193448\n",
            "Epoch 347 loss: 0.0001107424499760784\n",
            "Epoch 348 loss: 0.00011080044080686162\n",
            "Epoch 349 loss: 0.00010863318594632906\n",
            "Epoch 350 loss: 0.00011014545219977528\n",
            "Test accuracy: 0.9897\n",
            "Epoch 351 loss: 0.00010910178202759784\n",
            "Epoch 352 loss: 0.00010839256495163195\n",
            "Epoch 353 loss: 0.00010755659279218575\n",
            "Epoch 354 loss: 0.00010708641074801841\n",
            "Epoch 355 loss: 0.00010731129632447822\n",
            "Test accuracy: 0.9899\n",
            "Epoch 356 loss: 0.00010642300261590135\n",
            "Epoch 357 loss: 0.0001055887479964758\n",
            "Epoch 358 loss: 0.00010479604781667421\n",
            "Epoch 359 loss: 0.00010550155305035525\n",
            "Epoch 360 loss: 0.00010496305788613355\n",
            "Test accuracy: 0.9895\n",
            "Epoch 361 loss: 0.00010437494915859211\n",
            "Epoch 362 loss: 0.00010414541094829854\n",
            "Epoch 363 loss: 0.00010319019196616174\n",
            "Epoch 364 loss: 0.00010377249368784003\n",
            "Epoch 365 loss: 0.00010223741761124627\n",
            "Test accuracy: 0.9897\n",
            "Epoch 366 loss: 0.0001018638531392563\n",
            "Epoch 367 loss: 0.0001020551626366827\n",
            "Epoch 368 loss: 0.0001008373677617209\n",
            "Epoch 369 loss: 0.00010070817733261631\n",
            "Epoch 370 loss: 0.00010043677642168672\n",
            "Test accuracy: 0.9897\n",
            "Epoch 371 loss: 0.00010007547233133253\n",
            "Epoch 372 loss: 9.982201953089568e-05\n",
            "Epoch 373 loss: 9.89095591345217e-05\n",
            "Epoch 374 loss: 9.917254927564026e-05\n",
            "Epoch 375 loss: 9.855275019130949e-05\n",
            "Test accuracy: 0.9897\n",
            "Epoch 376 loss: 9.862442776448743e-05\n",
            "Epoch 377 loss: 9.773344852676488e-05\n",
            "Epoch 378 loss: 9.699886654384994e-05\n",
            "Epoch 379 loss: 9.665236649316342e-05\n",
            "Epoch 380 loss: 9.676670477100935e-05\n",
            "Test accuracy: 0.9897\n",
            "Epoch 381 loss: 9.601030158520288e-05\n",
            "Epoch 382 loss: 9.599405992965811e-05\n",
            "Epoch 383 loss: 9.536973582654454e-05\n",
            "Epoch 384 loss: 9.487439267690136e-05\n",
            "Epoch 385 loss: 9.435142090508581e-05\n",
            "Test accuracy: 0.9895\n",
            "Epoch 386 loss: 9.39263312687475e-05\n",
            "Epoch 387 loss: 9.392146924048442e-05\n",
            "Epoch 388 loss: 9.345961005893225e-05\n",
            "Epoch 389 loss: 9.26214253983651e-05\n",
            "Epoch 390 loss: 9.274691836269138e-05\n",
            "Test accuracy: 0.9897\n",
            "Epoch 391 loss: 9.187640790439077e-05\n",
            "Epoch 392 loss: 9.246921911683829e-05\n",
            "Epoch 393 loss: 9.154926050859722e-05\n",
            "Epoch 394 loss: 9.14552038481868e-05\n",
            "Epoch 395 loss: 9.054366834889946e-05\n",
            "Test accuracy: 0.9896\n",
            "Epoch 396 loss: 9.039580461429646e-05\n",
            "Epoch 397 loss: 8.988709021317175e-05\n",
            "Epoch 398 loss: 9.034840674538883e-05\n",
            "Epoch 399 loss: 8.967948359496192e-05\n",
            "Epoch 400 loss: 8.885664977757276e-05\n",
            "Test accuracy: 0.9896\n",
            "Epoch 401 loss: 8.902088246364842e-05\n",
            "Epoch 402 loss: 8.849526574646425e-05\n",
            "Epoch 403 loss: 8.727357674340263e-05\n",
            "Epoch 404 loss: 8.78701074854499e-05\n",
            "Epoch 405 loss: 8.737814806818181e-05\n",
            "Test accuracy: 0.9898\n",
            "Epoch 406 loss: 8.703012787418532e-05\n",
            "Epoch 407 loss: 8.684542708268827e-05\n",
            "Epoch 408 loss: 8.640109659106894e-05\n",
            "Epoch 409 loss: 8.648100576711648e-05\n",
            "Epoch 410 loss: 8.568312084720643e-05\n",
            "Test accuracy: 0.9895\n",
            "Epoch 411 loss: 8.563953962700069e-05\n",
            "Epoch 412 loss: 8.50836319603095e-05\n",
            "Epoch 413 loss: 8.478455624635356e-05\n",
            "Epoch 414 loss: 8.403700840819108e-05\n",
            "Epoch 415 loss: 8.401369508664347e-05\n",
            "Test accuracy: 0.9896\n",
            "Epoch 416 loss: 8.374153128709784e-05\n",
            "Epoch 417 loss: 8.384850227728724e-05\n",
            "Epoch 418 loss: 8.361674037474889e-05\n",
            "Epoch 419 loss: 8.318405250769975e-05\n",
            "Epoch 420 loss: 8.249051955902714e-05\n",
            "Test accuracy: 0.9897\n",
            "Epoch 421 loss: 8.28219530048402e-05\n",
            "Epoch 422 loss: 8.188101737822993e-05\n",
            "Epoch 423 loss: 8.192866457522002e-05\n",
            "Epoch 424 loss: 8.164574804379536e-05\n",
            "Epoch 425 loss: 8.144412748555166e-05\n",
            "Test accuracy: 0.9896\n",
            "Epoch 426 loss: 8.098621794084923e-05\n",
            "Epoch 427 loss: 8.026232672975955e-05\n",
            "Epoch 428 loss: 8.090468739487684e-05\n",
            "Epoch 429 loss: 8.015313425125013e-05\n",
            "Epoch 430 loss: 7.970062255081545e-05\n",
            "Test accuracy: 0.9896\n",
            "Epoch 431 loss: 8.022181364508576e-05\n",
            "Epoch 432 loss: 7.899261142797892e-05\n",
            "Epoch 433 loss: 7.900035118280412e-05\n",
            "Epoch 434 loss: 7.872003028199157e-05\n",
            "Epoch 435 loss: 7.849800700231751e-05\n",
            "Test accuracy: 0.9898\n",
            "Epoch 436 loss: 7.845039211178156e-05\n",
            "Epoch 437 loss: 7.791944428753341e-05\n",
            "Epoch 438 loss: 7.78936736557736e-05\n",
            "Epoch 439 loss: 7.775401493053146e-05\n",
            "Epoch 440 loss: 7.741467570105364e-05\n",
            "Test accuracy: 0.9895\n",
            "Epoch 441 loss: 7.693841708122951e-05\n",
            "Epoch 442 loss: 7.706502119792183e-05\n",
            "Epoch 443 loss: 7.634065777824956e-05\n",
            "Epoch 444 loss: 7.602947585752356e-05\n",
            "Epoch 445 loss: 7.629511814773239e-05\n",
            "Test accuracy: 0.9896\n",
            "Epoch 446 loss: 7.545716507155632e-05\n",
            "Epoch 447 loss: 7.561675696983339e-05\n",
            "Epoch 448 loss: 7.532017766490815e-05\n",
            "Epoch 449 loss: 7.485969020889321e-05\n",
            "Epoch 450 loss: 7.512422971771723e-05\n",
            "Test accuracy: 0.9895\n",
            "Epoch 451 loss: 7.518193683705804e-05\n",
            "Epoch 452 loss: 7.449395635426697e-05\n",
            "Epoch 453 loss: 7.412156616150191e-05\n",
            "Epoch 454 loss: 7.390970652196536e-05\n",
            "Epoch 455 loss: 7.31514474178041e-05\n",
            "Test accuracy: 0.9895\n",
            "Epoch 456 loss: 7.346222723569955e-05\n",
            "Epoch 457 loss: 7.327434540649017e-05\n",
            "Epoch 458 loss: 7.308239069129402e-05\n",
            "Epoch 459 loss: 7.251896387993462e-05\n",
            "Epoch 460 loss: 7.256174293066145e-05\n",
            "Test accuracy: 0.9897\n",
            "Epoch 461 loss: 7.20150147268497e-05\n",
            "Epoch 462 loss: 7.21664994012935e-05\n",
            "Epoch 463 loss: 7.16215487156722e-05\n",
            "Epoch 464 loss: 7.137268616529553e-05\n",
            "Epoch 465 loss: 7.116299883264394e-05\n",
            "Test accuracy: 0.9897\n",
            "Epoch 466 loss: 7.095811201276807e-05\n",
            "Epoch 467 loss: 7.085432730806571e-05\n",
            "Epoch 468 loss: 7.070420118527263e-05\n",
            "Epoch 469 loss: 6.988395422558048e-05\n",
            "Epoch 470 loss: 7.014359218581508e-05\n",
            "Test accuracy: 0.9895\n",
            "Epoch 471 loss: 6.98653978960812e-05\n",
            "Epoch 472 loss: 6.953097932382312e-05\n",
            "Epoch 473 loss: 6.923894396666036e-05\n",
            "Epoch 474 loss: 6.891011497782432e-05\n",
            "Epoch 475 loss: 6.879082884263732e-05\n",
            "Test accuracy: 0.9894\n",
            "Epoch 476 loss: 6.883605342195968e-05\n",
            "Epoch 477 loss: 6.844135630753062e-05\n",
            "Epoch 478 loss: 6.845008457704549e-05\n",
            "Epoch 479 loss: 6.796922867300021e-05\n",
            "Epoch 480 loss: 6.820900044835098e-05\n",
            "Test accuracy: 0.9895\n",
            "Epoch 481 loss: 6.76765082342265e-05\n",
            "Epoch 482 loss: 6.752987400750208e-05\n",
            "Epoch 483 loss: 6.750093043318727e-05\n",
            "Epoch 484 loss: 6.710543697057867e-05\n",
            "Epoch 485 loss: 6.666384480606623e-05\n",
            "Test accuracy: 0.9897\n",
            "Epoch 486 loss: 6.668735220122673e-05\n",
            "Epoch 487 loss: 6.651442486906387e-05\n",
            "Epoch 488 loss: 6.589426959247758e-05\n",
            "Epoch 489 loss: 6.624708875272097e-05\n",
            "Epoch 490 loss: 6.57237095427928e-05\n",
            "Test accuracy: 0.9894\n",
            "Epoch 491 loss: 6.57941057718634e-05\n",
            "Epoch 492 loss: 6.571586010040982e-05\n",
            "Epoch 493 loss: 6.540577297665901e-05\n",
            "Epoch 494 loss: 6.527590365533124e-05\n",
            "Epoch 495 loss: 6.53534688138576e-05\n",
            "Test accuracy: 0.9896\n",
            "Epoch 496 loss: 6.46950506428444e-05\n",
            "Epoch 497 loss: 6.466965588304075e-05\n",
            "Epoch 498 loss: 6.443915369296182e-05\n",
            "Epoch 499 loss: 6.395956272345604e-05\n",
            "Epoch 500 loss: 6.404499710381472e-05\n",
            "Test accuracy: 0.9894\n",
            "Epoch 1 loss: 1.1338889460955093\n",
            "Epoch 2 loss: 0.42787778469672333\n",
            "Epoch 3 loss: 0.32473039846303364\n",
            "Epoch 4 loss: 0.26945817219550167\n",
            "Epoch 5 loss: 0.2301835380415164\n",
            "Test accuracy: 0.962\n",
            "Epoch 6 loss: 0.2042802950657253\n",
            "Epoch 7 loss: 0.18479759023705525\n",
            "Epoch 8 loss: 0.17028088721512222\n",
            "Epoch 9 loss: 0.15938842889151847\n",
            "Epoch 10 loss: 0.14895583863165587\n",
            "Test accuracy: 0.977\n",
            "Epoch 11 loss: 0.14233228525341446\n",
            "Epoch 12 loss: 0.13513584721730207\n",
            "Epoch 13 loss: 0.1315915005754179\n",
            "Epoch 14 loss: 0.12478073602164987\n",
            "Epoch 15 loss: 0.11886938446080253\n",
            "Test accuracy: 0.9805\n",
            "Epoch 16 loss: 0.11265652375355331\n",
            "Epoch 17 loss: 0.10973826380792076\n",
            "Epoch 18 loss: 0.10555406761115421\n",
            "Epoch 19 loss: 0.10024141310565253\n",
            "Epoch 20 loss: 0.0983172802945603\n",
            "Test accuracy: 0.9832\n",
            "Epoch 21 loss: 0.09616811288349918\n",
            "Epoch 22 loss: 0.09424510809864953\n",
            "Epoch 23 loss: 0.09026612249264584\n",
            "Epoch 24 loss: 0.08760075623006709\n",
            "Epoch 25 loss: 0.08727240310247138\n",
            "Test accuracy: 0.987\n",
            "Epoch 26 loss: 0.08389668294917672\n",
            "Epoch 27 loss: 0.08224718102307589\n",
            "Epoch 28 loss: 0.08095363941766433\n",
            "Epoch 29 loss: 0.07953192938619585\n",
            "Epoch 30 loss: 0.07585840444686189\n",
            "Test accuracy: 0.9875\n",
            "Epoch 31 loss: 0.07467354587845203\n",
            "Epoch 32 loss: 0.07642049992731068\n",
            "Epoch 33 loss: 0.07182240587816055\n",
            "Epoch 34 loss: 0.07100905719072198\n",
            "Epoch 35 loss: 0.06863162411984478\n",
            "Test accuracy: 0.989\n",
            "Epoch 36 loss: 0.0666731806185994\n",
            "Epoch 37 loss: 0.06915818648670972\n",
            "Epoch 38 loss: 0.06578271564985834\n",
            "Epoch 39 loss: 0.06565705417240383\n",
            "Epoch 40 loss: 0.06422647347113788\n",
            "Test accuracy: 0.9898\n",
            "Epoch 41 loss: 0.061670315048810265\n",
            "Epoch 42 loss: 0.06253628398595588\n",
            "Epoch 43 loss: 0.05997599960008918\n",
            "Epoch 44 loss: 0.05958979503948615\n",
            "Epoch 45 loss: 0.05929726023258749\n",
            "Test accuracy: 0.9895\n",
            "Epoch 46 loss: 0.056766854689470424\n",
            "Epoch 47 loss: 0.05716317571969683\n",
            "Epoch 48 loss: 0.05543392951062112\n",
            "Epoch 49 loss: 0.05408175500816723\n",
            "Epoch 50 loss: 0.053653857487041365\n",
            "Test accuracy: 0.9899\n",
            "Epoch 51 loss: 0.05674328045574015\n",
            "Epoch 52 loss: 0.05414404050171026\n",
            "Epoch 53 loss: 0.05218117259470607\n",
            "Epoch 54 loss: 0.05192797518035433\n",
            "Epoch 55 loss: 0.05272416113865878\n",
            "Test accuracy: 0.9903\n",
            "Epoch 56 loss: 0.05166061436221289\n",
            "Epoch 57 loss: 0.04957368986895566\n",
            "Epoch 58 loss: 0.050018718084896295\n",
            "Epoch 59 loss: 0.04935013254516637\n",
            "Epoch 60 loss: 0.04766480225934657\n",
            "Test accuracy: 0.9908\n",
            "Epoch 61 loss: 0.049057648232234505\n",
            "Epoch 62 loss: 0.04586813719506298\n",
            "Epoch 63 loss: 0.0456472231916138\n",
            "Epoch 64 loss: 0.04572271855214416\n",
            "Epoch 65 loss: 0.045206833802568694\n",
            "Test accuracy: 0.991\n",
            "Epoch 66 loss: 0.044415357639329994\n",
            "Epoch 67 loss: 0.04433395916760873\n",
            "Epoch 68 loss: 0.045399340937958596\n",
            "Epoch 69 loss: 0.04278519620566861\n",
            "Epoch 70 loss: 0.04402073548296526\n",
            "Test accuracy: 0.991\n",
            "Epoch 71 loss: 0.043577188081832835\n",
            "Epoch 72 loss: 0.041515837056633795\n",
            "Epoch 73 loss: 0.04146904396568375\n",
            "Epoch 74 loss: 0.04285411434923249\n",
            "Epoch 75 loss: 0.03961126002466191\n",
            "Test accuracy: 0.9905\n",
            "Epoch 76 loss: 0.03982575385144223\n",
            "Epoch 77 loss: 0.04055697489569563\n",
            "Epoch 78 loss: 0.03949345019645989\n",
            "Epoch 79 loss: 0.03824067733057939\n",
            "Epoch 80 loss: 0.03779217454397292\n",
            "Test accuracy: 0.9914\n",
            "Epoch 81 loss: 0.04001471313185243\n",
            "Epoch 82 loss: 0.040099363119752485\n",
            "Epoch 83 loss: 0.038604663426318624\n",
            "Epoch 84 loss: 0.03659843270982634\n",
            "Epoch 85 loss: 0.036995203611550176\n",
            "Test accuracy: 0.9911\n",
            "Epoch 86 loss: 0.036882913999581164\n",
            "Epoch 87 loss: 0.03565731563065241\n",
            "Epoch 88 loss: 0.03576804460214971\n",
            "Epoch 89 loss: 0.03508952154238313\n",
            "Epoch 90 loss: 0.03554657848277834\n",
            "Test accuracy: 0.992\n",
            "Epoch 91 loss: 0.034473830605171986\n",
            "Epoch 92 loss: 0.03502965920559093\n",
            "Epoch 93 loss: 0.03466239051599461\n",
            "Epoch 94 loss: 0.0354841884517514\n",
            "Epoch 95 loss: 0.03305268121102869\n",
            "Test accuracy: 0.9913\n",
            "Epoch 96 loss: 0.032449340552780655\n",
            "Epoch 97 loss: 0.034010177150543434\n",
            "Epoch 98 loss: 0.031015933908796166\n",
            "Epoch 99 loss: 0.033320754463635464\n",
            "Epoch 100 loss: 0.032377252291673535\n",
            "Test accuracy: 0.9915\n",
            "Epoch 101 loss: 0.03249114214716706\n",
            "Epoch 102 loss: 0.03241914376141881\n",
            "Epoch 103 loss: 0.03230301869153452\n",
            "Epoch 104 loss: 0.03195598389534379\n",
            "Epoch 105 loss: 0.03181712495198827\n",
            "Test accuracy: 0.9914\n",
            "Epoch 106 loss: 0.029579017970607732\n",
            "Epoch 107 loss: 0.031230117438106872\n",
            "Epoch 108 loss: 0.029863887113521\n",
            "Epoch 109 loss: 0.029689208271531965\n",
            "Epoch 110 loss: 0.02955757157464049\n",
            "Test accuracy: 0.9916\n",
            "Epoch 111 loss: 0.029771624149869818\n",
            "Epoch 112 loss: 0.029777772231911147\n",
            "Epoch 113 loss: 0.02878972028852748\n",
            "Epoch 114 loss: 0.02839655909977202\n",
            "Epoch 115 loss: 0.030281638553086073\n",
            "Test accuracy: 0.9919\n",
            "Epoch 116 loss: 0.030140136333325428\n",
            "Epoch 117 loss: 0.02779193117624836\n",
            "Epoch 118 loss: 0.02812682587141072\n",
            "Epoch 119 loss: 0.02839841885563693\n",
            "Epoch 120 loss: 0.028479263499957014\n",
            "Test accuracy: 0.9919\n",
            "Epoch 121 loss: 0.028472506456813818\n",
            "Epoch 122 loss: 0.02799951964006352\n",
            "Epoch 123 loss: 0.027144876516499578\n",
            "Epoch 124 loss: 0.026457881658220852\n",
            "Epoch 125 loss: 0.026080896997494676\n",
            "Test accuracy: 0.9918\n",
            "Epoch 126 loss: 0.02675356994842741\n",
            "Epoch 127 loss: 0.025330825850043866\n",
            "Epoch 128 loss: 0.02736897615225378\n",
            "Epoch 129 loss: 0.026971890063865033\n",
            "Epoch 130 loss: 0.025839661344834196\n",
            "Test accuracy: 0.9918\n",
            "Epoch 131 loss: 0.025878187412443732\n",
            "Epoch 132 loss: 0.025052264088274302\n",
            "Epoch 133 loss: 0.026326901496830844\n",
            "Epoch 134 loss: 0.024137601324134686\n",
            "Epoch 135 loss: 0.024991859903440935\n",
            "Test accuracy: 0.9923\n",
            "Epoch 136 loss: 0.024777440839871042\n",
            "Epoch 137 loss: 0.025199957597287877\n",
            "Epoch 138 loss: 0.02463335260663972\n",
            "Epoch 139 loss: 0.024027177800130504\n",
            "Epoch 140 loss: 0.02404835441612454\n",
            "Test accuracy: 0.9927\n",
            "Epoch 141 loss: 0.024487943100462012\n",
            "Epoch 142 loss: 0.02325738094504307\n",
            "Epoch 143 loss: 0.022205177654967958\n",
            "Epoch 144 loss: 0.02317792589692816\n",
            "Epoch 145 loss: 0.023364551647850222\n",
            "Test accuracy: 0.9923\n",
            "Epoch 146 loss: 0.023049165137799213\n",
            "Epoch 147 loss: 0.024128842567439194\n",
            "Epoch 148 loss: 0.0208647567793719\n",
            "Epoch 149 loss: 0.022457257215951834\n",
            "Epoch 150 loss: 0.023817740814156656\n",
            "Test accuracy: 0.9918\n",
            "Epoch 151 loss: 0.02229656584832424\n",
            "Epoch 152 loss: 0.021875236061776975\n",
            "Epoch 153 loss: 0.02337252517947669\n",
            "Epoch 154 loss: 0.022763304575197478\n",
            "Epoch 155 loss: 0.021341555661073566\n",
            "Test accuracy: 0.9919\n",
            "Epoch 156 loss: 0.022902593961723252\n",
            "Epoch 157 loss: 0.021185311463861062\n",
            "Epoch 158 loss: 0.02122321932105971\n",
            "Epoch 159 loss: 0.020396255764058794\n",
            "Epoch 160 loss: 0.020742509280021654\n",
            "Test accuracy: 0.9918\n",
            "Epoch 161 loss: 0.019384483223347102\n",
            "Epoch 162 loss: 0.020336050161424635\n",
            "Epoch 163 loss: 0.020940047022931254\n",
            "Epoch 164 loss: 0.019868980783141102\n",
            "Epoch 165 loss: 0.02075653235990999\n",
            "Test accuracy: 0.9924\n",
            "Epoch 166 loss: 0.02038460265810806\n",
            "Epoch 167 loss: 0.020355909891766128\n",
            "Epoch 168 loss: 0.019943787947236034\n",
            "Epoch 169 loss: 0.02113843160080237\n",
            "Epoch 170 loss: 0.020630008811111263\n",
            "Test accuracy: 0.9926\n",
            "Epoch 171 loss: 0.018910288472157488\n",
            "Epoch 172 loss: 0.02042249932044696\n",
            "Epoch 173 loss: 0.020443080805181497\n",
            "Epoch 174 loss: 0.019852531069990444\n",
            "Epoch 175 loss: 0.019842820650369707\n",
            "Test accuracy: 0.993\n",
            "Epoch 176 loss: 0.020007170646437512\n",
            "Epoch 177 loss: 0.018998993242887764\n",
            "Epoch 178 loss: 0.019164210014012473\n",
            "Epoch 179 loss: 0.01839132219759513\n",
            "Epoch 180 loss: 0.01850015281397762\n",
            "Test accuracy: 0.9925\n",
            "Epoch 181 loss: 0.01876249388460991\n",
            "Epoch 182 loss: 0.017911496279917078\n",
            "Epoch 183 loss: 0.01907454486596269\n",
            "Epoch 184 loss: 0.019908623485915315\n",
            "Epoch 185 loss: 0.018538276427241762\n",
            "Test accuracy: 0.9924\n",
            "Epoch 186 loss: 0.019107203046827993\n",
            "Epoch 187 loss: 0.01823548999675083\n",
            "Epoch 188 loss: 0.017351921847976768\n",
            "Epoch 189 loss: 0.01704887478458591\n",
            "Epoch 190 loss: 0.018553736594953397\n",
            "Test accuracy: 0.9926\n",
            "Epoch 191 loss: 0.01831529715256309\n",
            "Epoch 192 loss: 0.017460425342393438\n",
            "Epoch 193 loss: 0.017667402005218255\n",
            "Epoch 194 loss: 0.017448469794897448\n",
            "Epoch 195 loss: 0.01672538879763351\n",
            "Test accuracy: 0.9924\n",
            "Epoch 196 loss: 0.01793368377334087\n",
            "Epoch 197 loss: 0.016548424774258615\n",
            "Epoch 198 loss: 0.017558902380307877\n",
            "Epoch 199 loss: 0.01748384349494934\n",
            "Epoch 200 loss: 0.017427563342873428\n",
            "Test accuracy: 0.9924\n",
            "Epoch 201 loss: 0.017291667317087426\n",
            "Epoch 202 loss: 0.01762703093968189\n",
            "Epoch 203 loss: 0.015685608884421928\n",
            "Epoch 204 loss: 0.017055114708046045\n",
            "Epoch 205 loss: 0.015761569105046513\n",
            "Test accuracy: 0.9928\n",
            "Epoch 206 loss: 0.01504434158564853\n",
            "Epoch 207 loss: 0.01728559534863702\n",
            "Epoch 208 loss: 0.015374664601416631\n",
            "Epoch 209 loss: 0.016643561314849026\n",
            "Epoch 210 loss: 0.01622769225009981\n",
            "Test accuracy: 0.9928\n",
            "Epoch 211 loss: 0.016621283571532957\n",
            "Epoch 212 loss: 0.01587166523726941\n",
            "Epoch 213 loss: 0.015481928297974316\n",
            "Epoch 214 loss: 0.015162166376198048\n",
            "Epoch 215 loss: 0.015722112404232196\n",
            "Test accuracy: 0.9928\n",
            "Epoch 216 loss: 0.015308378718489075\n",
            "Epoch 217 loss: 0.016132225238244986\n",
            "Epoch 218 loss: 0.01464891534896621\n",
            "Epoch 219 loss: 0.015351577032010443\n",
            "Epoch 220 loss: 0.016848846314239628\n",
            "Test accuracy: 0.9926\n",
            "Epoch 221 loss: 0.015082651953786802\n",
            "Epoch 222 loss: 0.015055448105901297\n",
            "Epoch 223 loss: 0.014372205658036886\n",
            "Epoch 224 loss: 0.015552483200999612\n",
            "Epoch 225 loss: 0.015668109519844518\n",
            "Test accuracy: 0.9927\n",
            "Epoch 226 loss: 0.016799906160864138\n",
            "Epoch 227 loss: 0.014286838127545881\n",
            "Epoch 228 loss: 0.01409783773930934\n",
            "Epoch 229 loss: 0.01540210475670551\n",
            "Epoch 230 loss: 0.01424175634672905\n",
            "Test accuracy: 0.9925\n",
            "Epoch 231 loss: 0.014307209297700393\n",
            "Epoch 232 loss: 0.014404449453134327\n",
            "Epoch 233 loss: 0.01405531214593576\n",
            "Epoch 234 loss: 0.014232135607303976\n",
            "Epoch 235 loss: 0.015416149397443698\n",
            "Test accuracy: 0.9921\n",
            "Epoch 236 loss: 0.013591398406070846\n",
            "Epoch 237 loss: 0.014552118567149662\n",
            "Epoch 238 loss: 0.014568128523780212\n",
            "Epoch 239 loss: 0.015143872899458066\n",
            "Epoch 240 loss: 0.015190588271886761\n",
            "Test accuracy: 0.9924\n",
            "Epoch 241 loss: 0.014548760180712552\n",
            "Epoch 242 loss: 0.014397172176012043\n",
            "Epoch 243 loss: 0.013026973497167702\n",
            "Epoch 244 loss: 0.013635164525333558\n",
            "Epoch 245 loss: 0.013667285580238673\n",
            "Test accuracy: 0.9924\n",
            "Epoch 246 loss: 0.014623629461317612\n",
            "Epoch 247 loss: 0.013247072570962208\n",
            "Epoch 248 loss: 0.013903906358394034\n",
            "Epoch 249 loss: 0.015248984152360309\n",
            "Epoch 250 loss: 0.014161178765735074\n",
            "Test accuracy: 0.9922\n",
            "Epoch 251 loss: 0.01224411216672204\n",
            "Epoch 252 loss: 0.013088138224353502\n",
            "Epoch 253 loss: 0.01341235238861646\n",
            "Epoch 254 loss: 0.013166031057659441\n",
            "Epoch 255 loss: 0.01278980727036504\n",
            "Test accuracy: 0.9923\n",
            "Epoch 256 loss: 0.01257566462689415\n",
            "Epoch 257 loss: 0.012779883847302342\n",
            "Epoch 258 loss: 0.012767147527623382\n",
            "Epoch 259 loss: 0.013878397198351992\n",
            "Epoch 260 loss: 0.013994007649683038\n",
            "Test accuracy: 0.9928\n",
            "Epoch 261 loss: 0.012487488009704789\n",
            "Epoch 262 loss: 0.012178928288327071\n",
            "Epoch 263 loss: 0.013419511540892668\n",
            "Epoch 264 loss: 0.013352571514314477\n",
            "Epoch 265 loss: 0.012709662238442361\n",
            "Test accuracy: 0.9924\n",
            "Epoch 266 loss: 0.012935903538838833\n",
            "Epoch 267 loss: 0.012532001703112288\n",
            "Epoch 268 loss: 0.01235467275295075\n",
            "Epoch 269 loss: 0.011875181136545348\n",
            "Epoch 270 loss: 0.012859750000984847\n",
            "Test accuracy: 0.9929\n",
            "Epoch 271 loss: 0.012336407583440219\n",
            "Epoch 272 loss: 0.012526768525486003\n",
            "Epoch 273 loss: 0.01284523270003111\n",
            "Epoch 274 loss: 0.011340829756183883\n",
            "Epoch 275 loss: 0.012088227997389265\n",
            "Test accuracy: 0.9933\n",
            "Epoch 276 loss: 0.012756911164956607\n",
            "Epoch 277 loss: 0.013198405582229915\n",
            "Epoch 278 loss: 0.01279349364777173\n",
            "Epoch 279 loss: 0.011568494478060899\n",
            "Epoch 280 loss: 0.011799403328553035\n",
            "Test accuracy: 0.9929\n",
            "Epoch 281 loss: 0.01206400706863422\n",
            "Epoch 282 loss: 0.012128115896056635\n",
            "Epoch 283 loss: 0.011938686582485963\n",
            "Epoch 284 loss: 0.011370541498391156\n",
            "Epoch 285 loss: 0.012634421865708246\n",
            "Test accuracy: 0.9924\n",
            "Epoch 286 loss: 0.011206501027672683\n",
            "Epoch 287 loss: 0.011526719032845615\n",
            "Epoch 288 loss: 0.011451822411922266\n",
            "Epoch 289 loss: 0.011614539002621002\n",
            "Epoch 290 loss: 0.011614715186738794\n",
            "Test accuracy: 0.9927\n",
            "Epoch 291 loss: 0.01208368432174411\n",
            "Epoch 292 loss: 0.011754802578048947\n",
            "Epoch 293 loss: 0.012113964873805904\n",
            "Epoch 294 loss: 0.010893350291966414\n",
            "Epoch 295 loss: 0.012692742146158677\n",
            "Test accuracy: 0.9932\n",
            "Epoch 296 loss: 0.012121979619310832\n",
            "Epoch 297 loss: 0.011859397727761295\n",
            "Epoch 298 loss: 0.012949978681477303\n",
            "Epoch 299 loss: 0.011282412965104594\n",
            "Epoch 300 loss: 0.011281837265073348\n",
            "Test accuracy: 0.9922\n",
            "Epoch 301 loss: 0.010534901716642137\n",
            "Epoch 302 loss: 0.011517698018120847\n",
            "Epoch 303 loss: 0.011158127317433186\n",
            "Epoch 304 loss: 0.011423578871161997\n",
            "Epoch 305 loss: 0.010807735338968573\n",
            "Test accuracy: 0.9926\n",
            "Epoch 306 loss: 0.011773633410973447\n",
            "Epoch 307 loss: 0.010588415816392395\n",
            "Epoch 308 loss: 0.010864494806441947\n",
            "Epoch 309 loss: 0.01199455403631974\n",
            "Epoch 310 loss: 0.011167418698706766\n",
            "Test accuracy: 0.9926\n",
            "Epoch 311 loss: 0.00987345946061962\n",
            "Epoch 312 loss: 0.01060320829446646\n",
            "Epoch 313 loss: 0.010913738243612968\n",
            "Epoch 314 loss: 0.010317146306860446\n",
            "Epoch 315 loss: 0.010518515984849708\n",
            "Test accuracy: 0.9928\n",
            "Epoch 316 loss: 0.011344399657718508\n",
            "Epoch 317 loss: 0.011049889943214071\n",
            "Epoch 318 loss: 0.01067300837441434\n",
            "Epoch 319 loss: 0.01135029785650552\n",
            "Epoch 320 loss: 0.011411496172567879\n",
            "Test accuracy: 0.9931\n",
            "Epoch 321 loss: 0.01092913247464197\n",
            "Epoch 322 loss: 0.011411232850588818\n",
            "Epoch 323 loss: 0.009920314646670784\n",
            "Epoch 324 loss: 0.010235037996437647\n",
            "Epoch 325 loss: 0.01035923824205236\n",
            "Test accuracy: 0.993\n",
            "Epoch 326 loss: 0.011136845912818345\n",
            "Epoch 327 loss: 0.009254495455719122\n",
            "Epoch 328 loss: 0.010090695884598018\n",
            "Epoch 329 loss: 0.010238147476739658\n",
            "Epoch 330 loss: 0.010917603565673001\n",
            "Test accuracy: 0.9927\n",
            "Epoch 331 loss: 0.010570653222515177\n",
            "Epoch 332 loss: 0.00892001625025339\n",
            "Epoch 333 loss: 0.009036430140811035\n",
            "Epoch 334 loss: 0.010804076284861568\n",
            "Epoch 335 loss: 0.009873665043655267\n",
            "Test accuracy: 0.9929\n",
            "Epoch 336 loss: 0.010515519137395474\n",
            "Epoch 337 loss: 0.008829538461140124\n",
            "Epoch 338 loss: 0.009954047978735345\n",
            "Epoch 339 loss: 0.009750050699383245\n",
            "Epoch 340 loss: 0.010092973188774636\n",
            "Test accuracy: 0.9923\n",
            "Epoch 341 loss: 0.010535170640906881\n",
            "Epoch 342 loss: 0.009707097333969116\n",
            "Epoch 343 loss: 0.010010587155453063\n",
            "Epoch 344 loss: 0.009970788181507349\n",
            "Epoch 345 loss: 0.009971891013683993\n",
            "Test accuracy: 0.993\n",
            "Epoch 346 loss: 0.009369500656872849\n",
            "Epoch 347 loss: 0.009477695147668237\n",
            "Epoch 348 loss: 0.010997988511267718\n",
            "Epoch 349 loss: 0.009349251928798078\n",
            "Epoch 350 loss: 0.010142157472929965\n",
            "Test accuracy: 0.9934\n",
            "Epoch 351 loss: 0.00918784747593356\n",
            "Epoch 352 loss: 0.010934814532777157\n",
            "Epoch 353 loss: 0.009583175614168287\n",
            "Epoch 354 loss: 0.010846840487587064\n",
            "Epoch 355 loss: 0.008824634963475197\n",
            "Test accuracy: 0.9935\n",
            "Epoch 356 loss: 0.010668143637381297\n",
            "Epoch 357 loss: 0.010001486854571631\n",
            "Epoch 358 loss: 0.00931221770482815\n",
            "Epoch 359 loss: 0.009149371922691427\n",
            "Epoch 360 loss: 0.009371513663797326\n",
            "Test accuracy: 0.9927\n",
            "Epoch 361 loss: 0.009612441816244837\n",
            "Epoch 362 loss: 0.009350629887690814\n",
            "Epoch 363 loss: 0.00839658801541053\n",
            "Epoch 364 loss: 0.009314706341047568\n",
            "Epoch 365 loss: 0.009490623004526756\n",
            "Test accuracy: 0.993\n",
            "Epoch 366 loss: 0.009459870628751829\n",
            "Epoch 367 loss: 0.009014795915299861\n",
            "Epoch 368 loss: 0.008789320771595704\n",
            "Epoch 369 loss: 0.010013517562252123\n",
            "Epoch 370 loss: 0.009147943208021531\n",
            "Test accuracy: 0.9929\n",
            "Epoch 371 loss: 0.009145353279123802\n",
            "Epoch 372 loss: 0.009561505883954205\n",
            "Epoch 373 loss: 0.008765992829666735\n",
            "Epoch 374 loss: 0.009834706503240122\n",
            "Epoch 375 loss: 0.00942981691357356\n",
            "Test accuracy: 0.9926\n",
            "Epoch 376 loss: 0.009321310805319536\n",
            "Epoch 377 loss: 0.00960442915138454\n",
            "Epoch 378 loss: 0.009144946255492541\n",
            "Epoch 379 loss: 0.008961907753089532\n",
            "Epoch 380 loss: 0.008702533117804884\n",
            "Test accuracy: 0.9926\n",
            "Epoch 381 loss: 0.00924611958898609\n",
            "Epoch 382 loss: 0.009100425599257339\n",
            "Epoch 383 loss: 0.008356831964510711\n",
            "Epoch 384 loss: 0.008575500086922176\n",
            "Epoch 385 loss: 0.008210678911283374\n",
            "Test accuracy: 0.9926\n",
            "Epoch 386 loss: 0.007385136202114065\n",
            "Epoch 387 loss: 0.00898845874557226\n",
            "Epoch 388 loss: 0.009050325164039282\n",
            "Epoch 389 loss: 0.008186677003419945\n",
            "Epoch 390 loss: 0.008634598940168558\n",
            "Test accuracy: 0.9925\n",
            "Epoch 391 loss: 0.008492760231399086\n",
            "Epoch 392 loss: 0.007655300616685337\n",
            "Epoch 393 loss: 0.009141605876678198\n",
            "Epoch 394 loss: 0.007968074606114334\n",
            "Epoch 395 loss: 0.008479904193609982\n",
            "Test accuracy: 0.9924\n",
            "Epoch 396 loss: 0.009137331056214972\n",
            "Epoch 397 loss: 0.0077762473482425876\n",
            "Epoch 398 loss: 0.007973793782880838\n",
            "Epoch 399 loss: 0.008893634442033263\n",
            "Epoch 400 loss: 0.008058882818616523\n",
            "Test accuracy: 0.9929\n",
            "Epoch 401 loss: 0.00836610264015616\n",
            "Epoch 402 loss: 0.008947633506399693\n",
            "Epoch 403 loss: 0.008214251564654204\n",
            "Epoch 404 loss: 0.00947696119036177\n",
            "Epoch 405 loss: 0.008705160535802729\n",
            "Test accuracy: 0.993\n",
            "Epoch 406 loss: 0.008410162273307901\n",
            "Epoch 407 loss: 0.008440548812615316\n",
            "Epoch 408 loss: 0.0077082511284691435\n",
            "Epoch 409 loss: 0.008963182549002762\n",
            "Epoch 410 loss: 0.008436978659509415\n",
            "Test accuracy: 0.9933\n",
            "Epoch 411 loss: 0.009197122583248049\n",
            "Epoch 412 loss: 0.008775302969227815\n",
            "Epoch 413 loss: 0.007950941169008201\n",
            "Epoch 414 loss: 0.008021026739490348\n",
            "Epoch 415 loss: 0.009460966116493084\n",
            "Test accuracy: 0.9929\n",
            "Epoch 416 loss: 0.008060202311156247\n",
            "Epoch 417 loss: 0.009000808750127834\n",
            "Epoch 418 loss: 0.007253300821028107\n",
            "Epoch 419 loss: 0.007790864288277523\n",
            "Epoch 420 loss: 0.008278955560627339\n",
            "Test accuracy: 0.9933\n",
            "Epoch 421 loss: 0.008019239624598816\n",
            "Epoch 422 loss: 0.007996065428292851\n",
            "Epoch 423 loss: 0.008524583076775474\n",
            "Epoch 424 loss: 0.00720806377787302\n",
            "Epoch 425 loss: 0.007950079254073792\n",
            "Test accuracy: 0.9932\n",
            "Epoch 426 loss: 0.008460299527935455\n",
            "Epoch 427 loss: 0.007270147656100487\n",
            "Epoch 428 loss: 0.007990210754067095\n",
            "Epoch 429 loss: 0.00688408499803339\n",
            "Epoch 430 loss: 0.007534267294165956\n",
            "Test accuracy: 0.9928\n",
            "Epoch 431 loss: 0.007484841113702194\n",
            "Epoch 432 loss: 0.007279059470639311\n",
            "Epoch 433 loss: 0.007101932645486573\n",
            "Epoch 434 loss: 0.007060060799976355\n",
            "Epoch 435 loss: 0.007842610323586801\n",
            "Test accuracy: 0.9933\n",
            "Epoch 436 loss: 0.00822842625388466\n",
            "Epoch 437 loss: 0.007562430109107321\n",
            "Epoch 438 loss: 0.007197040992803926\n",
            "Epoch 439 loss: 0.0088681894589742\n",
            "Epoch 440 loss: 0.007703901452302406\n",
            "Test accuracy: 0.9936\n",
            "Epoch 441 loss: 0.008130122541016757\n",
            "Epoch 442 loss: 0.007162531880173596\n",
            "Epoch 443 loss: 0.007583282120640142\n",
            "Epoch 444 loss: 0.006833617185714217\n",
            "Epoch 445 loss: 0.007626730262303165\n",
            "Test accuracy: 0.9932\n",
            "Epoch 446 loss: 0.007402267699750865\n",
            "Epoch 447 loss: 0.008113693053073532\n",
            "Epoch 448 loss: 0.00820533464917653\n",
            "Epoch 449 loss: 0.00785117414339215\n",
            "Epoch 450 loss: 0.007601359069093514\n",
            "Test accuracy: 0.9926\n",
            "Epoch 451 loss: 0.006984849086137804\n",
            "Epoch 452 loss: 0.007560974355621476\n",
            "Epoch 453 loss: 0.006753502120350441\n",
            "Epoch 454 loss: 0.007285648022391978\n",
            "Epoch 455 loss: 0.007434252400782385\n",
            "Test accuracy: 0.993\n",
            "Epoch 456 loss: 0.008197641160982877\n",
            "Epoch 457 loss: 0.007081975731792659\n",
            "Epoch 458 loss: 0.006960006665313311\n",
            "Epoch 459 loss: 0.007135654350586381\n",
            "Epoch 460 loss: 0.007537902132586743\n",
            "Test accuracy: 0.9929\n",
            "Epoch 461 loss: 0.0071250340302801765\n",
            "Epoch 462 loss: 0.007559663991771216\n",
            "Epoch 463 loss: 0.006286430890374479\n",
            "Epoch 464 loss: 0.006123485217081085\n",
            "Epoch 465 loss: 0.007298178346400066\n",
            "Test accuracy: 0.993\n",
            "Epoch 466 loss: 0.0076487379054108115\n",
            "Epoch 467 loss: 0.007532144906653673\n",
            "Epoch 468 loss: 0.007766137879044535\n",
            "Epoch 469 loss: 0.006703850027507428\n",
            "Epoch 470 loss: 0.006370590911022507\n",
            "Test accuracy: 0.9929\n",
            "Epoch 471 loss: 0.0073203961601626235\n",
            "Epoch 472 loss: 0.006841588232112474\n",
            "Epoch 473 loss: 0.00666508689344459\n",
            "Epoch 474 loss: 0.007320047813980225\n",
            "Epoch 475 loss: 0.007595805581418926\n",
            "Test accuracy: 0.9929\n",
            "Epoch 476 loss: 0.006980383763263406\n",
            "Epoch 477 loss: 0.007380409951601401\n",
            "Epoch 478 loss: 0.006379469998247746\n",
            "Epoch 479 loss: 0.007821958338825537\n",
            "Epoch 480 loss: 0.007273443808200395\n",
            "Test accuracy: 0.993\n",
            "Epoch 481 loss: 0.007718229802718301\n",
            "Epoch 482 loss: 0.0077096022128755265\n",
            "Epoch 483 loss: 0.007542876448143889\n",
            "Epoch 484 loss: 0.00673221381653762\n",
            "Epoch 485 loss: 0.00710481900392912\n",
            "Test accuracy: 0.9933\n",
            "Epoch 486 loss: 0.006863813476431231\n",
            "Epoch 487 loss: 0.006808732819502153\n",
            "Epoch 488 loss: 0.006852661072407344\n",
            "Epoch 489 loss: 0.006234356502570504\n",
            "Epoch 490 loss: 0.006314689578041797\n",
            "Test accuracy: 0.993\n",
            "Epoch 491 loss: 0.006271054312544556\n",
            "Epoch 492 loss: 0.007165090454943479\n",
            "Epoch 493 loss: 0.006235883117809871\n",
            "Epoch 494 loss: 0.007127335510869907\n",
            "Epoch 495 loss: 0.006220986751487604\n",
            "Test accuracy: 0.9928\n",
            "Epoch 496 loss: 0.006998692472911016\n",
            "Epoch 497 loss: 0.006728633731214351\n",
            "Epoch 498 loss: 0.0064607796738739495\n",
            "Epoch 499 loss: 0.006721230910972817\n",
            "Epoch 500 loss: 0.007316278115789778\n",
            "Test accuracy: 0.993\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1373"
            ]
          },
          "metadata": {},
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "mlp_test_accuracy = [accuracy_array.split(',') for accuracy_array in open(\"mlp_test_accuracy.txt\",\"r\").read().split('\\n')]\n",
        "mlp_test_accuracy = np.array(mlp_test_accuracy, dtype='f')\n",
        "lenet_test_accuracy = [accuracy_array.split(',') for accuracy_array in open(\"lenet_test_accuracy.txt\",\"r\").read().split('\\n')]\n",
        "lenet_test_accuracy = np.array(lenet_test_accuracy, dtype='f')\n",
        "\n",
        "labels = ['MLP no dropout', \n",
        "          'MLP 50% dropout in hidden layers', \n",
        "          'MLP 50% dropout in hidden layers + 20% in input layer',\n",
        "          'LeNet no dropout',\n",
        "          'LeNet 50% dropout']\n",
        "\n",
        "plt.figure(figsize=(8, 7))\n",
        "for i, r in enumerate(mlp_test_accuracy.tolist() + lenet_test_accuracy.tolist()):\n",
        "    plt.plot(np.array(range(1, len(r)+1))*5, r, '.-', label=labels[i], alpha=0.6);\n",
        "plt.legend(loc=4);\n",
        "plt.xlabel('Epochs');\n",
        "plt.ylabel('Accuracy in test set');\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 442
        },
        "id": "qGkk-o1N5IOH",
        "outputId": "482bfd2c-9966-4c1e-ed09-46142b31f7b9"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 576x504 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfgAAAGpCAYAAABh+iQtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOy9d5wdV333/55ye9lepSvtrmXJkm25yTbuBeMKdkILEGMgITxJICFPEjDNYNNJI/k9DxAbcMOYasDkZ4y7LVu9d2m1TXu3397vnfr8caUrrXYl7UorraTXef+1szNz5nvOPXM+5/s935mRbNtGIBAIBALB2YU82wYIBAKBQCCYeYTACwQCgUBwFiIEXiAQCASCsxAh8AKBQCAQnIUIgRcIBAKB4CxEnW0DZor6+nq7ra1tts0QCAQCgeCUsWHDhqht2w2T7TtrBL6trY3169fPthkCgUAgEJwyJEnad6R9IkQvEAgEAsFZiBB4gUAgEAjOQoTACwQCgUBwFiIEXiAQCASCsxAh8AKBQCAQnIUIgRcIBAKB4CxECLxAIBAIBGchQuAFAoFAIDgLEQIvEAgEAsFZiBB4gUAgEAjOQoTACwQCgUBwFiIEXiAQCASCsxAh8AKBQCAQnIUIgRcIBAKB4CxECLxAIBCcRqQiBfZtj5GKFGbblNOG5Fiero1jok2myVnzPXiBQCA4E0lFCiRGcvirXeQzGmv/pxeHS0Z1KFz7/oVUNXhm28RZw7Zt9m2PseLXe9FLJt6gi3f85RJqmnyzbdoZgRB4gUBwWpAcy5MczVPT7DsuUUtFCiRH81Q3ec8YURzuTvL6U7vRShZgE6hzk00U8de4kWSZ5Gj+jKhLcixPaqxwxLY/MIk58Nse67dKRQoM7ImTGisQHcximTZVDV5SkTzrnuvjwhvmoBXMWf2tz4T+JgReIBDMOqP70rzy2E5MwyJQ5+amexdPa9BMRQq8+ctOtIKB06Ny3Qx5vidrELctm/DuOFtfHUAvWVTVu9FLJvMW19JnxMgmivhsN9VN3hm75snANCw6146w/rk+FIdMoM7N9R9YNK6thrqSvPaTXRi6hcOp0H5RPYOdSRSHjNM1MUqRHMvzwiPbKWQ1FFVh6U1zCcsS9n6Rt0ybl368E7dfxVvlmrHfejoMdiZY/rM96CULT9DBLR9ZQk3L6RdVEAIvEJwgZ8JM/nTFtm1GulNsfHEfumbi9jpIjebZvWqYy+9qQ1amliYUH8pW1mdNw54RzzcxmuOVx3dh6hYOl8JV716Ax6+SihSP+7dORQqM9aWJDmQp5nWaOoLYtg02OFwK517ezIJlTWx9NUwxZyBJk5cx2/0tOZqjZ1OEZKRINllEUiSQIDGcp39HjAtvnItt2fTvjLP19QFMw6aqwUMuoTHYmSSbKOJ0q2hOmcRIrlIPQzfZ9FI/haxWEczqJi9tS+srdY6GM8QHs+iaRTZRGnf+oUwlSjCddkxFCgx3p0hH84z0ptFKFv5qJ5l4kdXPdtN+SQMuj0pti/+I5Z1olGq6CIEXCE6A+FCWlx7diWFYBCfxXmaKwwejY22fDhzLpuhAhh3Lhyjmdepa/ZiGjW3Z2JQHwpW/6aZhXoDmjqqj1kkrGgx1JTE0E1mWKGY1ArWu47Y5PpLDKJp0bRyrCFEhq7Hqd11kYiVUVcZX7eSmD0+MMiTH8sQGstSHAhP2pSIFXn9qN8lIHkmSuPo9CzjnkgbS0eKEdrrq3QtY/1wfO1cMsezONlSHAkB4Z4w3f7kX07RxuBQuviWEN+jE0Ezq5ky85kwx2ptmuCeF6pDJp0vsWT2KaZg43SqX3jGf7vVjaEWTvFUq/xa6SXwoh6FbtHQEYf8kprrJw9K3h9j0Yj+5RJFiVmdobxJPwEkmVhZQvWgSqHUjSxKyIlXa5dC6BRs8ZBMlSnmdbKI0wd7wrjjLf74HbAnVJXPprfNw+50kRnL766Cxe/UIAKrzYDvm0zrVDR4CdW4ysSKxoSyKU6aQ1tj51hB6yURWZM6/tqU8+bKgusmHN+hkzbM9SJKE2+fg2vefizfoJDVWoKrRgyxL9G6NsvXVAXzVTlwe9ZTkVwiBF5xVnEqhS47mWftcH4WshtOjkhzNM9qXmvHrJsfyvPLELky9PLh0XFJPz6YolmmN25YkcHlVrvuziZOMUz0BSI7lee0nu0CSJg3DDnenePnxnViGha/axZV3d7D4mtaKjfGhLG/9ai/dm8YI1nm4+b6JYpqKFBjtTTHclUSSJG744EIKGZ3wzjiJkTy1rf5p2RwfzvLaU3vIp0pYls28JXVUN3qRFQnbdtHUUcW+QrQyAdn4wj4WXtmEXjRRVJnoYJbtrw1gWTbB+ok2R/rTpCIF3F4HLq+Kw6kgSdIE8QJwOBUWX93C5pf72fbaAC6fg3SkwFg4g6Fb+8VIY+eKITKxIiDhDTq55SOLZzRUbNs2u1YMse4PfdiWjSTL1M/1oagSDfOq0IomHp+Ta9+/cH/UxE14d4KNL+wDwF/j5sq7Ozj38uZx/S9Y5yE5mscwLHq3RHjhh9vBBiS44UMLqW70HbG/VjV4uO79C0mM5EiM5omEM+zbHmP+BXVYpkXvlig7VwxhmTZuv4N8usS25YNkYsX9dZCoa/UhyRIuj0o+dbAdK/vn+IkNZivb5aUSidpWP7ZlUx8KsmDZwTolR/NEB7NgQzZRYvWz3eRTGooqY+oWDfMD+/uJhL/GTSmnn5L8CiHwgrOGVKTAq0/uopDR8FW7uPHPzzspN1ByNMfu1SOkIwU8fpXqJh+WaZFLlBjYlcAbdGHq1oyEBgsZjQ3P95GJFXC4VfSixt71YxTS2qTb2aTG9uUDzFtSRzGr43Ar5NMaW18NY9s2bp+Tm+5dVEngOhmCb1s2O98aIhUp4PSo6CVlXBg1MZJj4wtlwWhZUE0pp5MaKzD/grrKMcnRPIE6N3rRJB3J071xlEtva6tc48CkJ5soIctw832LmbOwBgBJkhjYnaBlQTXeoHOcbcmxfDnZq8lLsN5DcizPwO4EpYLBWE+aTKyAN+hCUSXaL6qnpvmg0ADEwhkMw0J1yBTzOi8/thMJCZvy+rDqlFEcCuloOUmsqmEOAKZpMdKbBgmcXhWHSznm+np1k5eGeQHWP19uK0WVueDGOcgS2BZ4/A7mnV9L59pRkCQysQKrn+2m45IGnC6VmpZyHdPRAonRfKXO6WiB5FjhmAlvlmXTuXaErk0RnC6F+lCAQkaj/aIGejdHyoI1iYedihQJ1nkI1rnJp7UJvy0w7njLsBjtSaM6ZJweFezJJz2HcmD//Atsdq0cpmdLhHSswEh3Csuymbu4FulAOwUchJbU0rVulECtm0JWp+OSRno3R7BM+2A7rhvFG3CRT5fw17goZnUCtW6KOZ1zlzVVjlcc8qRRBbfXgWXa1Laq1M3x0bs5VpkgNM4PMu/8Wlb/tptSTq9EJk42QuAFZwWmabHjzUFSkTxOd9mb7lw3wrI72pAmW8icBgcygB1uhdhAji2vhDENE4/fya1/eT6yWhZLt8/B7lVDvPrELjxBJy6vyvUfWIgkSSRH8/iqnSSGc6x7rg+9ZOL0qFxxdwdt59eRTZbGDbLJsTy9myPEhnIgQaDOg+qQUVQ3F90cYsurYUzDrmxvfjWMljfQVbM80fnJrsrgEqhzo5VMHE6FVCTPGz/rrIj/kUKFkw36U5mYmLpV9obiRTwBJ6ZuUchojHSn8FY5GepMMrYvQ6DOg2nYRxzsqpu8FQ/X0C0i4SzbXh8gUOfGtmz2rh8lEyvgq3bhcCoYmlU5t/3iBsb6M3RvHOPCG+cCZU+0e1OEVc90YZrWBC9NUWUWXt6Mvj9KoihSRQAPresBT7W6yUtsMEM6UsDhUrBMm3Mva6RvWxRDt9CKBsNdKULn1RKoc7Nn9QhGyaz0h6lOrHzVLpxuFX+NqywezX46LmocN+no3xHHMm2qm3x4Ak5WP9tzRE/04DaoTpn2pfUMdSWRZQmH+2B/jQ1miPRnyWc0Oi5qoG9blGJWR1Fl5iysYc7CmiP2heomL879E8upCFnj/CDVjR5M065MGKaKJEksvqqFXKrExhf2IVHu7x0XNdBxUcO4dhrYGaeUN1AdE+twoB0NzcTtc7DgsiZyiRJaYfLjJ4sqHNo3oJyPcKBOHRc3TDjmVETSJNu2T/pFTgXLli2z169fP9tmCE4xqUiB4a4kY/1pCply2MvlVSlkdWpbfDTODzJnYTWFjH5cN1UqUuDlx3aQT2tYlk1VnYdcRqOu1YdeNFl6c4j5F9RVju/eOMaq33UjyRJ60aCm2UcuWcK2wdAtAnUucokS/tryGl91sxev30lkIAM2yLLEvAvq6N0SqWSE3/KxJbg8jimvwccHs2x4YR+egJNSXmfhFU3s2xbDNG1M3SJQ72ZwTwLVqaAoMlfe3U77RQ3j6vzaU7soZHVkSWLh25oB6FwzguqQj7gMEA1n2PxqGEu3WHLdHAK1bhIjOfSiyb7tMYa6klimhcujcttfXYDiUKaUBFXV6KFva5Qtr4Yr3vKchdWko8X9dZAmTFL2bY/RsyXCxW8P4a9x07l2hO7NEdKRAlUNHvIZjdpmH/GRHFX1XrSiwUU3hyrh1qn0lVSkwFu/7MQybeT9NkA5+uCtctC9IUIupeHxq2QSJc57W8u4vjLV/nf4NY42GUuM5Njwxz6cbnVcHb0B57htj89JNlHE5XeQiR6IDhlUN3nJJUtYpo1lwTXvPocFy5qOKyHtZB5/OL1bI6z/wz5qW7zkU9qE+3Iq15jpvJZTtSwmSdIG27aXTbZPePCCM5ZUpMBLj+4gmyihqDI3fGghVQ3eiigUMjo73hxk2+sD+Kqc+5NfppfY0rluhEy8iK/aDcA5yxrp3x5DL5qTeif1oQDVjR60oonLo5aTdeJFnG4F1VH2mAY7k9imTW2rj4tuDtG/s1ye061SKhgMdSYwNIv6uQEsy6aQ1mmcFzxiiHOybY/fgWWUxTS0uI7Q4rpx3kUxo5FNapTyOuHdcZxeFaNk4XArdK4d3T9RclAqGox0p8rn5HRUp1JZBph7Xg3J0QKKeiDxagTLtAnWuwnUusfZZNs2o31pgnVuZEUiEy9NCNsezqHnVzd58fid5cQmSWLBZU3jwueHlzN3cQ3DXUm2vT5ALqWhqjLnXtZI96YItmnjr3Kx+JpWtrzcj140Jg01H4sjeWSVOlvwwo92YOomLp+DqsbpD/JT8fqOFCo+tI6TbVc1eLjolnlsemkfetHE7VWpavSSS2q4vCqKQ8bhVie9xlTsPpnHH05tix9vwEE+deSowVTD/jNl04mePxMIgRdMmdMpU9u2bXauGCKbKFbC17Z1mCg0QiZeYLg7SSlvYFnTe3xqcE+C+FAOt6+cFKUoEvOW1DFvSd1RQ3XX/dmicWJaSB/0wDoubqTj4sZx5wfrPSRH8limTaDWxUW3zKsMwtMNWR6w4WjCA1RslGTo3hjh5cd2oTpkDN2kcX4QX7W7Mim5+t0LAHjzF3soHbIM0LVh7LDEK5mmdj+lnDGhnZvaq6hu9FTaYbp1qmn24atyVs6fLHx+KIoi09QeZPkvOpFliWC9h9Zza2g9d3yYNVDrPqE+fTQbChkdf40LRZWxbbv8IpjG6a+7nuik4/A6Hm0bIL/fgz9V68QzwWyEv88EhMALpkQqUuD1p8sZxv5qF9d/cGYeBzs86elYCUCpSIH4cJboQJZ0tIAn4MThUo4ohE1tVdQ0eUlFi+TTGulYYcJbtw7fTkUK9GyOMLYvTeuCaq68u2PCW7qm4wkcS2ynMihPl+l4K3rJZKQ7CZKE2+Fg4ZVN1M8JTLj+oROX+GCWQlojWF8Odx9IvCrljEmF4UQH4OM5X1ZkArXll8XkEuUch6Mle8001U1enPvX5xVVPmVieSxP9FjbZ6pQng4e8+mGEHjBlEiM5MjECkiyRCpaFtkTuZlsa3/S0++6MXVzXAIQgOqQ6bi4gcHOJEigKBKLrmxm95oR8ikN07C48l0dk4rvoRzwqA88UjO0N8n25YPlR59Mm/pQgKHOJAAOl0zHpeUQfC5VwuVRueJd7VQ3eo/L8zrUhmO11UyHB6dD/dwAwfqDSU71+5+nniw6MS4U7HNQzOqoU0i8muz86TLd86ubvLg8KrlEaVa80TPVqxRCefYgBF4wJbS8gWlYON0KmmaRHMljX2hPO0M9OZYvfylrLE9y/5vH6lr95LMaLo+K2+fA7XWQTZQqb7w6kAC0bfkghXT5mXOP34Un4JyS+B4YsOZfYLP+D30MdyVxuFUsw6KY1XF5VZwelVyqVBF3f7Ubh0chHSme9R+2OB4hmsoywGxzOgisEEvBbCIEXlDhSGvsxZzOaF+a865soakjSC5ZYqw/w8CeBKHzaseVEQmnycZL1Lb6x4XYPQEHscEsG/64D0MzUZ0Kl7xjHr1boxOSnsqP/JTfeLX5xX4M3UKudrHk2lZ2vjUElD386XpkkiSx8IpmhrtT2JaN6pDHrXfXNHkrb9kCjmv9+0zleIToTBCvM8FGgeBkIQReAJTF/Y2f7UErGLj9jsoHHGzbZu+6UWzLZunb5+LxO7FtG3v5IN0bxrBNm2Jex9AsYoNZejZHyuLpVJh/YR3De1NoRaPyEREoP/dq6Bb+GjfXHWPt+cAbrw5sN84PnvDa9A0fXDStawoEAsGZiBB4AVBeY09Hy2vshYxG75YIbUvr6d8RY7g7xeKrW/D4y28FkySJxVe3suI3Xbz+0z1YllV5W5PH78Ttd5CNF4n2Z8inS3iDLmzs8pr67gSmbh3xsaTpbh8Pp+IaAoFAMNsIgT+Lmc6LG0r719jdXgeGbTOwJ8H25YNYho3ikAjWd4w7XnHINHcECe+KU1PnxdQt5l9QV3mdY1VD+RnbzS/3Y+9/5KZ9aQPtSxuEdywQCASnACHwZymjvWmW/2IPhmZV3o7Wvz2GrMp4A+Nf+FJZY39bC03tQaqbvOzbHiMxnEN1ynj8jkmf4W1qq6K6ofxss8OlTJpJPdnjXkLYBQKB4OQjBP4sIhrO0Lc9RjGnEx/OkUuWcPsclPLlt6MV8waqU8Y0rMrHP2zbLn+swoaLbg7h9jsAkC6UCO+KY+rWERPappJJLcLdAoFAMDsIgT9L6N8ZK6+HmxYOl8Lia1rABgkI1EqVcPmBbyhHwhlCS2qJDWSJDWVZcGljRdzh4CcZjxVOFwIuEAgEpydC4M8ChrtTbHl1AEmGOefUUMzq1LUGOOeSpknD5YWcxkhXijXP9pAayxOo9zB3Uc2EcoV4CwQCwZmLEPgzmORYnj2rR0iO5WkMBQCbYvbg5zePlh3ucqu8+Yu9WJaNVjRJx4onRcy1cBittxdnezvOUGjGyz8eTkebBAKBYKY5qQIvSdLtwH8BCvAj27a/fdj++cCjQAMQB+61bXtg/77vAHftP/Rrtm3/4mTaeqaRGM7xwo+3U8zquP1Orry7g/OubplyhrqsyATq3HgCTvRi+eMgnmJ0RoWv1N/P8Gc+C6qKEgzQ9IUvzLqgFrZvZ+ShryJ7vchez2lh02SISYhAIDhRTprAS5KkAN8D3gEMAOskSfq9bds7Dzns34Anbdt+QpKkm4FvAR+WJOku4FLgYsAFvC5J0vO2badPlr1nEulogXXP9e4PxfuxgdRYYcKHNI4mEtVNXpxuBb1Y/jiI10wy+E9fwMpmcYTm0vylL52wsORXr0YfGUH2+bDyeUrd3bMiVlo4THHHTvTRUXIrV2CMjaHU1IAkldvnFNtU6u+nuHUrjrlzcbS2AqCHwxR27EB2OtGjUTLP/QG5qgq1pnpGJiGl3l5K3d24Fy2aUlmT9Z1D/+eYM4fi7t0YkQiujo4plVns6kIPh3EtWCAmLQLBKeBkevBXAF22bfcASJL0c+Ae4FCBXwL84/6/XwN+d8j/l9u2bQCGJElbgduBX55Ee097kmN59qwZITVWwOlRK4+tTfZKVS0cZviBB8C0kP2+CSJxaAZ8VYOL4uPfwxgdRfb50IdHZkT49EgEyeEARcFMJsmvWYtSXY2VTp80z/SACDna2lCrqihs3UrkP/8LIxYDWabq7ndh5fPog0NYhQJKY+NxX2OqddDCYUo9PSh+P8Xde4g/8Th2SQNZxnPJJQAUNm0CywJFwb1oEZauQzaDKcuT/hbTsSG3di3DDzyAres4Wlpo/fa3j1ieo60NK51m9DvfwS4UkBSVmg9+EID4009j5fNgmjg72inu3IXkdKI2NdH6ja8f0Q4rnyf57LPEfvRjJEXBOW8ezV/58gn//mdilGMqNp+J9RKcnpxMgZ8DhA/ZHgCuPOyYLcC7KYfx/xQISJJUt///X5Ek6d8BL3AT4ycGAEiS9AngEwDz5s2baftPKZO9hObQGz1hVPHy4zvQSyZuv4Nb/+J8zMgosV1h6pfMmxCSz61Zi9a3D9nvQ9b1SUXCU4wijXZSfHk7xugYamMjtq5jplLYhnFC9TFiMcyxCHV/9XEcLS3YhkHm5VcY/Kd/RvZ4kH0+Gj71SZT6eoyxsYoXePjgdqzB7tD9ZibDyENfxdpvv3vpUqxMBiMWwxmaC7KC99JLqX7Pe8itWUN+7Tqyr7yCJMsYIyNTGlBL/f0M/fNnsLUSakMjzV9+YFKxLO7ahezzY8aixB57HCubwTZMHKG5SKoD14IFGLE4nosvBkAfGsQZCmHEEwRueTu2aaD17UMfHS1Pkg4hv3kzIw8+hG0Y5Xb8u0+h1NVNaMdSdzfG6CiZV17FNi3Uunr0wSHiTz5J9fvfjzE8XG63eIKRr38dM53GNgwcLS3o4XAl8pJ67rnybzo6ilJTjW1ZSE4Xak0NtmWh9/cTffiR/ZOnAs72NpyhEKWeHjIvvIA2MIAZiyO7XSArlPp6yW/YcMy2Ltehp1KeFg5T6uxE8fsxojFijz6KXSohe9zUfvyvcNTXYaYzuBaeO2XxPFb/KvX3o/X0jrNB6+ubcn/VwmGKnZ0ogQBGNErsRz/GLhZRAgGaHnoQz3nnVc5R58xB6+oi+v0fIHk8R4zeTOWax3sPTXVCcTyT3JN5/FSYbrvMho0zzWwn2f0z8H8lSfoosBwYBEzbtl+UJOlyYCUQAVYB5uEn27b9CPAIwLJly+xTZfRMk4oUeOXJnWRiJTx+Bzffdx6+5D4G//c/gttNuukCBs67G0O3aWoLohcNhv/nVVy/+j84vF7yra1oXzw4EJjJJIXNm5FcLjBMjNFRjFgc2z749TctHGb4Kw+iDwxgGwaNn/kMngvOp9TVRX7tOvJr1uJdtgwlGDyq7Ufq5LnVq0GWCN55J0ogAICt6xS2bkVSFIzRUaIPP4I+OAiShOzxEHjHO8ivXo0NSA6VqnvuIfXss0g2SH4/zV8cP9hp4TAjX/0aZiJR9sZrajBGRlAbGrA0Dfd55+G+4Hzijz0Oto3kUCt2OkMhvJdcQvThhxn89D+gNjchu91HDYfbuk7iJz+pRDpKXV0kn/kNwbvuxBgeRqmtpdS5l+gPH8HOF0CWccyZg5lK4WhsxDZNgnfcQW75cmxNR62tIXDTjQDkV63ETCSRnQ68V1yB94orKO7cRX7tWjKvvgaKihGNoA+PkF+zBjOVQgkGy+343w+Pb8db30F+1WqMeBxb1wneVU5lsS0TVQJtYJCBv/0kst9fFpu6OoxIBLWxEatUxHvJxRRUFSQJyeGg8f7PAjD2L/8CpoXkUKn54AeIP/YYlqYju1xY2SxD93+uHIWQZVyLFlHaswcsCzkYpP6Tf0v62Wexcjn0UpHcqtWodXVg2xMGXKW2ltKePUR/+CPsQmFCeQfbNYlaU4sRi5H46U/LbWDbyIEADX/3Kfw33IAxNnYwMpHLMfqtb2Nr2vj+pSjInvH5GLZtk3nlVcb+9V+xi8XxNgCyx4P/7W+nsHYttiSNK88uaWCZuC+5hNwby7FLpYM2JxIo1dXoo6NE/uO7uM9fQn7lKsxsFqtYxNHSgplKlSfl2SzFPXvG9cdSfz8jX/4KVrE4/pq6cdRtbBvZH5j8Hvra17CyOSSPh+avfBnpwNLVYcszpa4ukCSKe/aQ+tWvK9Gblq8+dNRJiFUsMvqNb2Dr5Qlp8wNfKh8zidgqDQ1oezqJPvwwVqmEWl9Py7e/hbuj44jlT0VcS/v2MfKVr2AVS+N/e1k5rlycA2OPXSwi+XwT2vV4bDwZSLZ9cnRRkqSrgAdt275t//bnAWzb/tYRjvcDu23bnjvJvqeBp2zb/sORrrds2TJ7/fr1M2L7qWbv+lFW/qYbl0ellNOonxdgbnY72ivPMVy/DN1SqGmvJ+2fj5lMQKHAkuIqHPt2IqnlOVrDP/0TwbffjK1pxH78Y4x4guBdd2LG4xT3dJa9tXPOwdHUiK1pZFetJr9yJXIwiBIIUP/X/wv/9dcDYESjRH/w3zjmzqX2I/chyfKkdhd272Hkyw8g+/zIblflJrEKBcb+/T9wL15M9XveXTleC4cZ/eY3sXUDJAnPZZeRffllZJ8XfXQUpaoa45A1e6WuDjMaLW9rGlV3vwv/dddjxGNgGORWriL7xhvIXi+2YeC/8QZK3d1IkozkUCv2HO1Giz/9NLFHfojs9YJtU/2+9+G7/jqMoaFxx1v5PImf/YzCjp3oAwPIHg9mMona2kpp714kRcHWNBxz5mBEIjjnz8fKZgneeSfZt94Cw6jYBEw6gE5mo5lOM/bd/yT75ptgGKDIVP3Jn1Dcug1se/J2DFaVve2qKiSvl4a//Ruc7e2V8vPr1hP9wfeRHE5s0yRw882U9u4FmJaNh24XO/cS+8EPUGpqMGMxnOecg9bdXY4QpFLU3vfhig1KQwPpZ58l/eJLyD4fkqriv+EGsm+8gZXPV9rRjEZwzA1hxuOV8pT6esxMhqo77yS7YkW5TVQV39veRvqPzyN7fWj9/bja21Fqaiju2gW6jm0YqM3N4yITh/YvLIvq970X33XXkd+wAX3fPkrdPWj79uGcN2+cDfL+iZVcVYUxPDx5f83ncew/z9HSgpnNELzzTnIrVoJhlCd7t91Kbs1aChs3olRVYdY7yFkAACAASURBVAPBd76TwoYNGJExzGQK//XXE7ztVvSxCHaxQG71GgobNx7xmkfd1jSq3nkX/uuuw0imwLbIrVhB9rXXkb1erHweZ3s7xugoKAqSWhZCgNRvf4uZyVQmZMbYKEgyViZD4NZbqbr7XRiRSHnSGYsRe/gRrEIB29BRG5vGtbtr4UL04eHyBFJRKr+9XShgaRqO1laMsTHU2lr0kRHcF15I8B3vQA74UVtaMWNRov/f/8EqFZFdbuo//fd4L74YIxYb1z9L/f3k12/AzuXIrVtLccvWSdvFtkyq3/0efNdfhzk2dtRIR3H3HmzDILdyBbk33yqPG5ZF7cc/Ts1731M5ttTby/CXv4yVySIHg7R87au45s+fihxMG0mSNti2vWyyfSfTg18HnCtJUjtlz/wDwIcOM6weiNu2bQGfp5xRfyBBr9q27ZgkSUuBpcCLJ9HWWcO2bKLhLJIELp+KJ+igqS3Ivjd9ROa+r7wma2Rojm+mbv0z5NRaAmqB1r/6AOk/FDGiMYxYjMK6tThaW0g98xuMRIL6T3wC96KFAPhvuonE0z8j9qMfVTwg71VvQ21qKmeTe9w429srNqn19QTvupP4Uz8l+t8PU/Wud07o8EY0SuyRh8uh/QYLCFaWAfIbNmJrGr6rrxp3jjMUoukLX6jchADFLZuxdQNXezs1H/0osUd+CJoGjY1Uv++9JH/16/K6bz5PcdduUr/93ThhU+vqkL1eJJ+P2g9/GJgoTAc89snwX3cdmZdfwUzEsXJ5cmvXEv/pU8gOB5LHS91f/AVyMEj698+ColD3sY+hVFcdFMuNm4j170P2lG/0wG23kl+5Els3UKqC+G+4Hv8N109q0+FtM5mNSjCI9/Jl5FaswDF/Hkgy3osuovpP/uTY7WjbyC7nuKhF5XovnIOtachuFzUf+uAR2+1oNh6+rTbUY+sGalP5t4s/9hhmKjUhcgKgDw2RffMtJEnCTKXI7Y9KqLW1YI9vx0PLs3UDtboK/4034L/xhnFtkF+zutwGCxZQ97GPkl3+JtbGjai1tVjFIp6lS0GRkSx7XP8y02msXI7cuvXEf/o0mCaSy0XtRz9Cdnl5YnW4Dc62+dR89KPE94fcDy0Py0LyeKj7y784xOZqAjfeSODGG8e1s/fqqxl56KsggexwUHXH7VTdcTtaby+S2036j39k6AtfrORn+G+8EUdra3lif+g1TfOo21YhD/kCxT2dpJ79/cF76JKLUetqwelCCQZxhEIYw8NIDgdmKkV+wwYAzGwG55xWLMMkeMftZF99FatYwnI60QcHGfj7T5dtlKTyJDcWQ62vwyrKuM9fApKEbejIfj9KbQ1aXx+y2z3+t6+pQfJ6CdxyS+W3dLbNR62rY+y7363Y7Jgzp5y8G/CjJ0eJP/Y4yUCA4tat5eifLONeupTi5s3Yug6qQuCWWzBj8bKzcmi7ZDKY+Rz5jRtJ/PznSKqK7PVS99GPotTXow8NIrs9GPEYyV/+qjwWSRKey5eh1teX2yaVIrdyJfrwMEp1FRgG+fUb0PvLkxq9v5+xf/t3grffVo4uFgpTTkw9UU6awNu2bUiS9CngBcqPyT1q2/YOSZK+Cqy3bfv3wI3AtyRJsimH6D+5/3QH8Ob+cHKa8uNzJ7YofJrSvzNOMa9z9bsX4HAplTX4tWveIB1opK7ZgyY3Ql0NgZEu6toCmFkJpaqqIpY2kH7uDwx86u/2D0ZNyN6Da/KSJOEMzUVtbEAJBLE1jarbbx/n1R3e2ZT6erSuLgobN5J54QVavv41PBdeCIDW30/ip08je7w4mpvRR0eRc3nUOXOwLYv8mjU429pwtLRMqO/honCo4DtDIZytreO2PUuXHvQSd+4k9sMIcjCIrZWoetc7cXV0HFOYjoYzFKLloQcrYdz8unUYI8NIThdGNEriF7+ohL+doRBKddWEOmTmzauERCcbxKdr0+F4LrwQ1zkdlWtMJtjHasfD69z8pS/OqI2HT96coRCOlpYj2uBeuBDnvBC2buCY00rtxz5WEcMjteNk5R2tDRytrWg93WWBraul5oMfoAYm7V+OtjYKGzcSHx3B0dyCpWk4586l+YtHr9PR+utUbHbNm0fLg1854jFmJktx85b9yyclAjfeQN3HPnrUax5tu7hrVzliFQiU76F77hl3DwGMRsbKv0trC02f/Uz5f/sjb6q/HC3xve1tB8vcsQN9cBC1rg4zm8X/9pvJr1kDhonqUKm99144pN0PLW/S3/7tNxN4+82V40s9PZQ6O5GcTqxsFt81V1PYuKm85AjU3vdhCtt3UNq7F0d1NUYiUU4Q9XhwLVyIVSzgv+Yaau+994jtkt+0ifjoKJLLiTEWIfHrX5fv+/0OkbO9DSwLZ3s7tmFQddddlXZT586luHkL0e9/f/+EQiV4262Y6VQ5SlYq4V68mOwbb5Bftx61oQG1rvaUPKJ70kL0p5ozLUQf3drDwIZ+RrI+Qhe3suTa1sr6uJlO0/ud77HDcxVysBpZkbjyugCFH/xr5SY4vHMkf/tbIt/7Hs7WOeB0UnffhyshdxgfHp/s/MPJLl9O7PEnQAKtuwfXwoVlj9w0Ke7Zg2POHGrvvRcrnye7fDn5TZvwLF6C2tpK6je/oebPP0TgxhtntM2mW4cTvQaqiu/aa8j84Q8458/HiESpPaxdD5xzstfaTof1vJlmppOepnKNYx17svvXdJlpm6ZS3rEekZzs+MPLhIkRoaNd40TLP/yYwycNx2q3w8/3XnsdmeeeQ21owEynCN51F9lXXjliednly4k99jhKVRVmJk3dRz4ywYFKv/wysUd+iOvcBRijY5OOJcfD0UL0QuBngejWHl79rzfJ2H5kh8Kd/+sCmi5dUNlf2L6D5C9/ieN9HyUnBSte/bFugpFvfHPcWu+JJH2MXy8H9wUXlEN7pons8zHnu/+Be9GigzZv207s8cfLj3vJMq4FCyZNPDlRTrWYAqfdoC84eZyOE6mZtmm2J1Inq/wTnTAe676HI09apjpxOhljiRD404zdP3+ddS8MYUoOnFKJS98xl/M+cGNlf/r558mvW0/TF7+ApChTLvdkDgRaby+xHz9aTsbRNOo+ct+E2Wfs0ceIP/EE7iVLsHV9xmaos83pOOgLBIKTy8l4tO9kjCWzlWQnOAI1C5rRX4wg2RaqWaR+yfhn+LX+MI65c6cl7nD0ZLLj4fDyZK8HW9eRnY5xSXkHCLzjFnKrVmHremWt+GxgpttVIBCc/kz3vp/K8ad6LBECPwvYhRI1niK1aoYWxxi15x18MtDWNPThIfzXXjuLFk5ksiSqyY5p/vIDwtsVCASC0wAh8LPAwMZ9eH0qV3z8VpJP/QStqwv3kiUAaIODYNk4Qqffm/lOxxmqQCAQCCZn8jeYCE4aer5EJJylqS2Ie0EHssdTfhnHgf3h8tt9naEJ7/sRCAQCgWDKCIE/xQyu7sQyLOZc3o4ky7jOO4/inj2Vd79r+/pRGxrKb0gSCAQCgeA4EQJ/ihncHMbjNKm7+FwA3EsWYxdLaH192LaNPhDGOU+EuAUCgUBwYog1+FNINl4gNZSm/dwq5P1fCHN1dCA5neWvjwWCWIXiabn+LhAIBIIzCyHwM8RUnm8cWNMNpsGcy8+p/E9yOHCdey7FXbtxNDcDCA9eIBAIBCeMEPgZoPLpQF2b8NnJAyTH8nSuGSagavgvWDRun3vJYoo7dpBbtQrZ50OpqzuV5gsEAoHgLESswc8AWm8v+sAARjSKEYlS6ukZtz8VKfDqk7uIRgwiUguZ1PhP27sWLkRSFYxoDEdobuWd9AKBQCAQHC/Cg58BHHNDWFoJbLByeQqbNpU/UhCN4mxvZ18XpIbTOI0cqquK5GieqoaDX3uTXS7k6hqK27bhufjiWayJQCAQCM4WhMDPBBJ4LroY72WXotTUkH1jOYN//2lkv4+oOoeh5qsg50Ax0lijBXxWCjgYhtfCYfKrVqEPDZHSdbyXLxMvixEIBALBCSEEfgbQ+vpQAgGq3/1uZI8HLIvYjl4GPBdTkPw0GsN0ZHZhtJ6LJz+KOzkf6Dh4fm8vkteL99JLMTOZcrKeEHiBQCAQnABC4GcArW8fjuamsrgD+rwlbJv7XoqWG5dcYsF752A804Od75n0IyzO9nZkpwMzkzmrPtIiEAgEgtlDCPwJYus6ejiM94rLK//LyVXYtc340PFUN6A1n0vLUT7UMpUPuQgEAoFAMB2EwJ8g+uAgtmHgbGur/K+6yYstq9hOB6rPQ3WTF2dD3VGFW3ykRSAQCAQziRD4E6TU2wuShHP+/Mr//NUuGtsC1DR7WXRly7iMeYFAIBAITgXiOfgT5PD1d4B8RsPpUWm7sF6Iu0AgEAhmBSHwJ4BtGOjh8ISkuHxaA8AbdM6GWQKBQCAQCIE/EfSBgQnr7wCF/QLvEQIvEAgEgllCCPwJUOrrK6+/zxv/9bd8RsPtdaAoonkFAoFAMDsIBToBtL6+8vq71zvu//m0JsLzAoFAIJhVhMAfJ7ZhoPeHJ4TnbdumIAReIBAIBLOMEPjjpLL+fliCnVY0MQxLrL8LBAKBYFYRAn+c5NatQx8ZAXl8Ex5IsPMGhMALBILTn3AmzJsDbxLOhGfbFMEMIwT+ONDCYeJPPIHW10fku99FCx+8McQjcgKB4EwhnAnz4MoH+dd1/8o313zzuEW+J9nD8oHlRzy/L9V31P2Ck4N4k91xUOzci5XN4Wxvx9aNcV9/y6c1FFnC5RNNKxCcdcR7IdYFdQug9sz/KNTu+G4GMgO4VBdjuTH6Un2EAtN7ZXZfqo9PvvJJ/E4/Td4m7r/i/nFlhDNhPv3apzEsg/nB+Xz+ys9P+xqC40N48MeBpMggSdil0oSvvxUyGp6gE0mSZtFCgUAw48R74bd/Da99E/74ufL28ZSx96XjO3eGsW2bvYm9+zcgo2fwO/zTLmf96Hp0S8e0TDRLoy/VN27/htEN5PQciqwwmB2kNzX7dZ8t9ib2ntLlEOFmHgdWJoP3iisI3n4brgULxn0kJp/W8FW7ZtE6geAM5XT3jnf9HlL94AqCoZVtnY6d8V54/nOgZcHphzu+Pav1XD+6nkghwqcu+RQe1cPr4dfZFt3GxY0XT8tBMW0TWZJJa2n8Tj9tVW3j9idLSVRJxaW4SJVSRAvRCWWEM2H6Un20VbURCoSOuX1KmOH+uGVsC59d/lnqPfXUuGsmRDpOBkLgp4lt25Q69+K5aCmBm24at8+ybApZnYZ5gVmyTiA4Q4n3wu//DowieOvg9tkVvwkMrIfhraC6oZQGX0N54J8OsS7IDJXrqGWnP0HYz0yI3Vh+jBf7XmRB9QLe2fFOJEki6AzyfN/zdCY6WVS7aMplpUtpbm+/neHsMC3+lnE2FYwCg5lBPrj4g7RXtbM3sZdt0W0srl3M4rrFAHQlunho1UPkjTyKpHDD3Bt4Y+ANTNsctw3Q4G3gc1d87uSL/HH0x2P9Ln/s+yOWbdFR3UEkHzmu5ZDpIgR+mhiRCGYqhf+G6yfsK2Z1bNsWCXYCwXTpfpVwsoc+l5u2vEloKuI3BQ+rJ9nDQHaA9qr24xtM472w90UYWAutl8C1/wivPAjn3jZ9cTZKUIiDJIOeB9uctjnhdJh/fOMfUSSFek/9cXmBvaleHtn6CE7ZyT3n3FPx1i9ruoz1o+t5cd+LnFN9Dqp8bHnI63mGc8PcGLoRR4uDF/e9yEhuhOZSAWJdbLFyGLbBbW230exr5qrWq3hixxM8tfMpOqo7MCyDrZGtDOeGCTgDZLQMK4dWksxH8ckKGctk5dBKEqUEHtVDupQ+JcJIZA8k94HDC6rnmJOxcDrM59/6PKlSirmBuXzhyi+MszFZTBIvxKl11xLJR1BldUKk42QgBH6alPaW16xc5547YV8+XQLAIx6RE5xsphs+PJ3D37Fuwn1v8FW/SgmdoB3lfrPAUYfweC88/xkopMBTO2m4e93IOr701pdwq25a/a0TBt1jEu+FZz8FiV5w+uD6z0LDImi/HgqJKRVR8epkN6Ge1+H890DoCuhfBV2vQvNFEGyZskkrh1aSKCbwOXzk9fyUxC6cCdOb6sWjeuhMdPLUzqfQLZ1QIESilMDvLK+7K7LCrfNv5eGtD/Potke5o+OOY5bdm+rFxqajqoNady2v9r/Khu4/ctfGX2Njs0HJM3fRXTT7mgFwyA6umXMNz3Q+w5qRNThkB+/seCcls4Qqq6iyyn2hW3ly1TcwTANVcXLfJX/GE+EX6Ep2kdWzxy2Mx/Kwx+3Xc2BbUMqUJ2R1C454flpL89iOxxjJjeB1eBnJjUz4XVYOrSTgCvDgNQ8SzUdP2VKDEPhpUtq7F7WxEaWqasK+yiNyVULgBSeReC/8+i9AdYG76tjhw3gv4d99gj4jS5u3mdCd/zEjIn/C66TxXti3AnrfYoVLoS9QiweJrFaga+//T6j9JnAdYblraBNEu0BxgJaD6N5xdbJtm2e7nsXGBqAn1cPKwZVc7ZtH3/A62louJxS6inB41bjtcUT3QipcDsd7aiDZT9jtpU+1aRvbS6iYBnfwiG2S1bN8/c0vkS0m8OhF/iq4hNqlf0oKnXMu/RChTT+DdT8k3HEdfZFtk9twCLZt05/px6k4KRpFooUo84PzJ/1d9sT34Hf4iRViPL7zcbJaFtM2meufi1Nxsrhu8aQTBJfqoivZxebIZtaNruPLV335qL9ld6obt1KeQMmSzPk1C9m65UluiXcz5Ksh6pa4Rx0/VkbzUVr8LdS6a8loGZY2LOXuBXcf7Dvdb9GSytPncNBm5AjFB2hZeC//0/lr+qziERMBj9b/elO93L/8fhyyY9L1795UL1948wsgQZ27lvsNH6Hz/xScAYjuIZwJ88CuR0kWk3hUD39z0d9QqxXZMvAWu80Mhuqm2ddMySyRKqVQJKVSdkbLsHFsIxc1XMTi2sVQe8TmnHGEwE8Dq1RC27cP39smvwkLaQ2nS8HhVCbdLzhDOJ29XYDtz0B2pLw2OJXwYecf+EdrmIIqMU/bx+cH1xA6wXqFM2EeWvlQxfO655x7+H3371FlFafiPHboON4L//MPEO9mvVNl44IrcJsBnIqLtBRnRW4A30v3U6pfSNucK8tifGAAl1yEel4re1iSDKVU2cu2bdgfbt4W3UbJLNHkbUKVVSL5MZbvfJpnEj3INpjbf8QFVeewOtOLW1Jo3fUzPnfzf4wXWIe7HEaXHaC6CLv9fHPNN9FKaTz5Qe4Pv0Xo3DvHtcm31n6LRDFBXs9Tq3gZTfZQbRjEJPiJGWdozUNYtoUsyVxcdS4M72DLvl/hUhxU7XqKB2/8N5y17ZMK1a74LvJGnn+49B8YyY2wLbqNkdwI84IHP3bVnezmn9/4ZzJaBlmSafW1kiqlaPI2Ydomt7XdxpuDb5LX85OGiftSfdS4apAlmb5UH8/3PM/Hl34cWZr4wJVt23Qnu2mvakdO7IPeN7lsYAVb9Bzb/NX06Uncuofz59047ry2qja8NmSTfThcwUo9Q4EQjO2G3uWEbIWQUg1GGnpeJ7T2x9zrreK/nBobuv/ALS1Xj7tHN41u4osrvkjQGZywdGFaJj/d9VPixTheh5ecnmNPfE9l/0huhIe3PEykEMHn8OHUNfpMD6GrPwuNi+G1b7Jt+9MM5gfxO/1EChEe3fx9huJ7sGwbr6Tw9Wu/QU3j+XQnu1k5uJL1o+tZ1rwMp+Jk9fBqLNvimjnXHPWeOhkIgZ8GWk8PmNak4XmAfFoX4fkznXgv/OYT5YHdWwd3/MvpJfKpAYjsBtUL+VhZ4I+W7GXqPNf/CglZwovEoK3Rmx87evibo3tDlm3xTOcz9KX78Dq85PU8P9vzM2K5URpUL353zbFDxwNrCSd7+F3ARb+qcomnmb++8nMMZgdxKk5e3PQIXxpZjm/sTRy7n+CG9jt5I74du5TGo5f43zWXUn3b1xhI99FWLBAa2wmdf4RFd6CZGi/3v8zCmoX8fdvd9O3+LXOReDbTxR4JvIpKyTIYSvXhtHRMVaJfKrF3cNV4gc+MwLxrYMk90LSETfFt9CR78Dq81AB9Q+vGCXxvqpeB9ABIoJs6izytOEyQbZkmWeXymvN4KbePgDNArBhjXv1iKGTpHosgKzLDaHxj43dJyhLVrmq8Dm9FqCzb4vXw6zR4Grhl/i1ISDy9+2le6X+FRbWLqHJVkdbS/Gjbj8gUk5zjrqMgydzadisrh1ZiWAaqrHL1nKu5es7VR/xt26ra8Dq8KJJSfowuuZcndzzJFc1XECvGDp4T7yU6uI50rIvrDQVeuw/yUeaqbpqX3sEqp4fk8Eau0G0cxfS4a4R0g/uHB+jT07SpfkKpUdB02P0cjGyG+kXwtr+G7Fi5b3c+D2N7CBYzLCrl2LT+v7kx85+oihMCzVh3/AvP7H0GzdTIG+WcgN5Ub6Xdftv1W1KlFM3eZnRLJ1lKsmJoBW7Fzc7YTnpSPbhVN3P8cxjLjxEpjtHScAM0XwiShLn4nex68wEUSafOXUeLr4XLDImXdYN6ZDTJJBXZydJzbicUCNFe1c7jOx7njYE3uKb1GtaPrOeC+guoddeWo06J3lPmPAiBnwalvXuRXC6c8yYfuPIZjbo5vlNs1SwT3QuJvtPX250um38K6YHyY0yZUVj9fei4qewZNpw3u3U0NNj4JARb4Zr/DWu+D5ICVXOPeMqe9Q/TY6aoCrSiIpMoxogMrobsB8DfOOk54UyYB1Y8QLQQpcZVwzeu/UbFS9RNnd/s/Q2D2UFqVT8e26bBVcu73R38fHQXI3KCdHaEqlJu0nL7Un20Kl66O3/H9/yQU0xqJJm/a7uNedUddFR3AJDbt4KNI+sBm6Sls6L3eRKWhteGhCTxX+YoQzu+hyqrNHubeahxGaFtv4Khzaz0OMkkdvFeRzOh175HyCyBM8B7rvxLdu18DBMLt+Lmvgs/xpObfkDc0Ik6HKzVYjTHdhIrxGjzhwgNbYa2a2DxXexN7GXN8BoUWSGv59Eli5ZsbFzUYDg3TNEsUuuupT5Qz5+f+2fQuYI+VaHNMw/OeRfrO3+KYRk0eZu4Z8E9UH8Zu1/4awytRNDpZV71OQzEtmJjY9pmZaK0LbqNSCHC+xa+r+JN39V+F9/f8n2e2vh/CVkSm0pR5HyUUDaOlIkSlFSua0pyXdP19GX6aZt7TUXQjzT5CgVC3H/F/eUJQLCNWDHGrzp/xc93PkWT6iPgCnJ/3RWE1j1Bt50Hl0yHswvyUahdgCQrLAvM5+fZLjLBJlqpg22/hJo28NWBXoC1jxBKhAk5/ZAdhJcegOQA2Eb5EcTrPgON5x00auEd0P06lNIss3zscjrZKWVYaskQ6WTt1scpySVa/a3k9BwZLcOu2C5CFvx69y8Zkyz+dMm9hIIh+lJ9+B1+Xuh7ga+u/iqWbRF0BvnO9d/BrbpZt+d3rNn7LFuqGjgHkIDXjSRZl4/7jSBqYBFttgKDG9gE6JKEwzJoG+suLzdpeebXLeDSxkt5qe8l1o2sI/H/2Dvz8MjKKv9/7q19r1T2pZJK0vvedAMNNLQgCOKCOuq4jeKM4jgq8xvHEXAcRR1lHMdxw2VGRQUdERFZZN+37gaa3tcknX1PVZLaq27d5ffHTSepTjqdhg6gvJ/n6aeTuu+teypV9X7vOe95z8mNsbl2M/Ttgts/Ct5KcAVfkZ0iQuDniWEY5FtbcTQ3IVln/tlURUPJqa+vGvRtj8E9V5uerq/qVdnadFr3x7Y+DKPt4C4zBd6aNr2I+//FDAV7q+CdP4HS5uLzThbSn0fIf16v4+Cdpj2b/gHKl5gZ3Vu/Dx1PwaI3zhg+3PkUd3Tcy+LyNXzyzKvpSfZwdHgf+9vuZ/mjX2TFkrdD5UrTpmk2PtD7CL3JXnx2HwPpAX64+4dcufJKZEnmloO3oGgK767YRE3rC3Qq40QKBcL+FGvSBi/aLeyySjxw9B6y7hIUTTGTuXJjfGfnd4hnR8llRqiWnOS9ZSxzVaBaHHTJOvXTbF9cew61Lb9HNXSsGHy47GxuHnyGgtWGDqz1hHkya47tTnbzR18D/zByiHTPNp51W1jpb6TepoPFDtVrIZ8kHGziixd9p2jNvdpXR+f272MJb+JeucA/P/nPhBwhPLrBF1UPUugSHtn/Cw7HDtMUbOJDKz/EgegBXjh6P8+lemlK9CEH6tgf3c/h0cN8aPmHiAQiZtZ+tBOqNxJu3GIm5oUauSZYV/w++8Jcc/Z1dO7+JZEV74Elb6J/+7/Tk+hhsDBIXsuj6ipP9jxJtafaXMOdIOgMsspZyY27f4ChaViBb0jllKV1Oj0BIqk44YP3wHg3YQMIPGx+fk/yHQ0XVMLZHLhVwsHFDFpL2ZkeIaMP4dANOod7CRsG7YFKQlqWkqVvMz+DugoWGyVly9nX+Udsso3/8ehcozkIP/tdCDXB4B4zcc1bZeaPGDrUbIDcQ+YcYhhm3sN0gQ81wpu/CbE2GkPNhI78hh3qE6zRXIwbAzzev5V1FWu5YNnf0jG8h3Gbk2f6tvMPu35MASiXbISX/c3UMgAQV+K8OPwiNZ4a8lqewfQg59duJpxVKPMt5nEtyaLoXgL2AM/2P8sZNedw4dafQ88+U/XP+gTXnHUVncN7zLyWjq1wxyfMhE9fNcvP+wd+FN1HQS/gt/vJF7Lwwk9BK5ivJzX8krdJngpC4OeJOjyCFk/g3bJl1uPHEuwWsotcdyzD0WiK5jIv9aXuk59wms8vQtdh9/+ZoexC5hX7wE7n2DrwWH6MSnelWQKzoJ58/fx4wR3tgAN3wNAB01u/6N+mjo+0QHLAnLwSvfDYv8PKd5oTkcMH6WHYeiNgmHfmb/l28XVj7fTc+XE61SQRTw3hy78zw64jo0f4tyc/36JS7wAAIABJREFUjw8Zl6uEa877ygyR79nzazr3/ppI4yWEy5eYD5Y2Q/lyaHsUGs4z14wnaG29l//dfgMee4D3nvtvBNwhGgONnFtzLr9KDvDrI3eyqW876yUX4cgb6Ol8gg5Zp8cVoKV6yeRapt/uJ+gIctP+m9g7tBNdU6i1+anp7yCcGCLsrQSLAcvfTvjoY4TjPWxJJfhxapgvPX0dsq6jyTI1vnr6410EClm8SGxe8nZ2JzsmQ8fHrwWHw+dwzYX/PSXGniqq7++lU8sQsbhh6Xs43PIbFM383vWMtfFfbgt5XyW5QpZLVn/EFI4HrjUFxWKD0kWEQ41FYfjw4ssJj/XBWCex2jPZMbSDZCHJUHqE62UHw7u+TU7L4bP7uGrtVTQHm1keWk7EWcq9W/+D2/b+HFf1Orb3b2dpaCkfWP4Bc3uZrsFzP4eqNXDGhye9/OkiM2nDqr8mPNYD8QHw1PLFTV+kZbSF/bH9PNv/LIf6tnJwZC8fXPK+GQVogslh7KqKV7LgNSBVs5b19j7CugrltbDoYth3u+kdx7th2w9hw0fN/fizfUdGO+DefzZFVitA7RmszI5Qo6p0Wy3osoVI82WoPS/Sqfazzh6EZW8x/018Z/oyvVR5qqjz1hHNRuksbSC8/RemmFtd8Pbvg7926jsGMHLQvN7E+zSDUCOEGpGADZE38nA+zmD52Tya7YfYAd4yHiV49/8jbOhgaCSDIXYZKk2GBZkCnT3PEPZUTV6zOdhMpc1HPtmHdSIPgKOPQ9dWNq99H+1SmtsO30ZWy1LmLOPS4HJwTCQL2j1QsZxwZAvhyIQeWFzmzgg1C4leBoZ2U+oqJVPI4Lf76Tz0B8JqDoJhc6480es8zQiBnyfprVspDAwguWcXxkxyYZvMdMcy/Oud+8AAu1Xmy29beUoifex8VdNx262nfP4Mup4BdAg2mGvB6WFQMrNe97TdVBzHgegBepI9OKwO+lP9dPRuJ7zz9rkrhXU8Dfd8BnQD7G5Y9W44/Cczac0ZhAu/ODmZTOIqMScfqxNsLnjwuqkEr2AD5OLm4yNHzBD6qr+CRD/oKj17fs2/GIPEbTL1SjdfnCXB7Q87b2R0rB0ZC5ZEP529zxFePiUCe3b8D9ftuRFZguDR3/HlujNwli4yPcHwesIjh6D9CVh6GRgGe3f+jH/d831yGNTrConYEQJuU9SsspXN/sXcbrPwgg1ceoY3Dj/DC7YMSRmyWp6P2M/lM2dfS1eii0ggQq23llt2fI/D6YdZXtBISGN0htYRDjWbHrLFBksuNf9FWylJDXNWy+28kIniw0IWnY26B1tqFNkwsNndXFp9Lpcuf++cUYtw+JxiMX7zf5v74yeEabo33NG7nRtiN1DQdMpkG4lALYFQoxlVOtkNX+MWGNrPOt1Cc7CZjJIikB6j0RdhpBCjKdCEJEn0p/ppDprRm431F3Jw1y+4qfsBGHoWq2zlo6s/OrV3vHeHuZVu9Xsmxf2ESBIsugRe/AUM7CZcewZhX5gt4S38atsN/O7IbdiR+N3z3ybirZ36m+g6jYkoDbqEZrObN0or3wvThAyAtkfMz68PSPTB7z4Ankpwl8yMuo20wFiX+Z7aLFC+nHDNB7h+2/f4k5GgXSowVrsWLXIehYO/onnp+6fOn/g/YrPitXmJZqOmTd4607N1lUzUAMjO/I7N532aYF35Ou5uu5vvjO5AN3T+au3fEuxvgaGD4C6HQoYL6jaztf9hNFXBqqtEOp+H/fcCEugq4ZozuKbzBToljYguEx77HAzuB0lG3vt7zjrnKm49ciuartHgb2Co9kLCgTDoJ7gJqVwFwYjpBOTGiWgaQUcQj82DTVWIDB4xb7bqz3lFE3iFwM8DpaeH2M9+hlEoEP3BD6j8wheKytMCRHtSpEZzFHIqBE5/qdpdPWO0DqWoDjrxYOVoNHVSweyOZdjXP45S0HmmLUrrUAqPw4qm57hzdx8XLStnJKVMiu+8xVhJw5EHoHYDfes/R6xrDw2ZQwTaHweL1Zz0SxdxRCnlmj/sw26R8Tlnv6l4qSH2glZg78heJEnCJttIKklisSNTlcKy4+ZkFWqcfF1LHOPUbv+RGX3wV5tJVIfugfQQBMJg88DoUShtmrrQ8SIRbYVoizlh5ROw4p3mBFrImBP68EG49XaQLBT0Ar+pqCVqteLRdfop0CEbRQluR2KHGRvej1/XicqgqBki6Yk91oUcsd238KuDv0STwGfzMKjn+MbO7zFukXBZXQQdQa4LLSJ88C6MsS5eHNnNL0d3kQdW20rI6AqdAy8UCeWw00O17ETCYFjWeN5XylAyg0fTCWka9a5y6v31U9nZhsFF0T5eUAskrHaskoXI4rdA3aaZk9XE/0uTvdS2tKJKEn5D44q8zhUFN50ltUSUAuFcCmbxZufkOFGY7g13+jqpLVuOU1NRLHY6JdX8Ox8vJLNRthj8tYQHDnDdmdfRefQBInknrPtbho78+oRRhobS5TgSR7Ba3QSdJQylh1hSssT8fLU9YnqpFSvm99qq15oRoLZHzII6koQ1n2Jx907KdJ0IVkbJFb+XrQ8S1lSu3fA5OtV48Ta7E4ln93YYPgS5MUCfGXVLD4GWM8XfXQqr3w2hRsKlzXwi2sLN8QPcF9vN4uBi5EA9kfqZmeFF6/iBiBlV8/9xXh76fIjlYrSOtxLPx3Fb3dR4aqCxEo7ca17D6Se84WNcu+ytZgQo2Ey47QlTwO1e8wZjvIuw1UvYVwXpKNi85ne6dBHk4sSiBymxl2MlQEHN0ympyOdcT7RrP2UNq6g93tZQo+lQRFsheoRwrI1rIpvp9JUROfQAYX8QVr7LjLK9glFOIfDzQOnoQM9msYfDM7rHAcRHsux7shc1r/Hs7W1sfu8SAuWu02qDphnIssRgPIfTJlPmmRkpOCZktQEXLcNJvvdIK8mciixLnNccoirgRDcM0nmD3d3j3PpCD1ZZwiLB+vogu7rHkSSJmqCTb7xzzQwxPvb8q0YfwZ1Ocr9jIz++cxi7pZZydy0/qH4A972fJ+moICv7+IX97xiMe/E6rWQUlbbhZNFzdie6+dyTn0OW5MmtLfJIaupL1LT8+JdIdyxD20iStsyj5LQc/3r2v2JgNs3YN7ibFcoYS3UZjAwM7OJxdQVff+AoDfYEf124E199KfFQPZ0UiJQ1ET77M+Y6tiTPe/LJWINkkxlcTi/uxRfD4ounJtDe5+l5up0jDiv7JAexQCUVmg9yCaK5UWLTGnHktTz37ruJZhx8WqrkQSNNq5aho/0RwrKLeNuD3KIO4Qk0UB/vwKIb+HHQULqUZ6J70AyNkcwIt2BnU+9TPDP8JEnJwrKaDaixw2R0BaskE6k+s+jmLVJ3Nu6ulaj5BA0OPx9e+/fctOtG1HwCZy5JJNpuhrWP7UE/ch/h9CjXGEE6rV4iFjfhuk1zTsrh1X/NNV3P0qlniFj9hM/7LGy7kXChYO4AOM3hyUgggtNZgqqr2E+1SpgkmV78nv8jnM8Qzingr4fas7kmUFt0A9oZTXN0OMXiSh9NtZuo73qAvNWN0+YhEojQHcsw1PIcy2L9+M79uxne+wlvoiXJ9PB2/8a8SQzWw/Yf02gLELJ5GFUVrFqBiD6xXW20HVoehLozCa//0Jy7IrqNSo7qHpoNL/X1m8wbmtF28wbX0CZtWmqLUtO/i7GVH6Hb1lgsZKFG5FAj78xv4id7fsJjXc+iFKzs7uvm7PrFM16XoZSiph0YLi+Uumd46MePP5lzMf14V7YTn92HRbLgsXnoSnRRX3f+jGsYRiVqoBmj1AtlK8xjumYuE5z7aTJPfo9sOofLWYl7w5Ww7UYzGmexoVqX0za4DdUYQsLCbxIZjvQOglRG8NAoX39HHK/DVmzzse/Dooth968p2XsfNkXH67DCmz8DNueCRjRnQwj8PLBFIgBo8Ti26qqi7nEA40MZ1LyGJ+hA1wzGhzKnXeDHsgXetKKSJZVednaPcc/eAbKqTjqvUh9yE0srfPP+w8QzCtmCTm3QRV7VWVLppaDpXL6mhuYy7+SHa1fPGMNPteO2WxnLKCRyKlaLjEWWaB9J86d9/XxyS/Pkmt/kEkFykHen7mK8bAOPJyQ03aAgGbREVX5VqODdaZV0apQxVNY0j9BVXkEsqTCaKfBcxyilXjujmQLNZV7u6rqL0dwoPrsPRVN48cjTND38U6y6wsjzXnjPd4tEviuW5hu/v5W48SSjDpVPnvMxzq87H4Bzqjfxy85nuLmskk21F7DSUYl0+EWiu/6bpbkI783cT0r2cU/Vu3lo7Fc4NBW7M8A1lYsJv/U78w6bdRuVfCf3QULZLkZp4Gq9guFMPztiOTY6Cng9Ib7llRgihw58vvFyltRvpiPewZHdv2J/77OcOd5OXbCJRzseIDVyiL+uPJva9Vfy8Wgrd8YP8Xjno6S23sDjdhmHw8OnNv47qqFNrkcTrGPw+W+SVJI4LU5Gc1Fu8FrQkanEytX1l3LF2k9Mjjfca/nCH/eSymt4HBa+dsUq/mbFdezoP8zGmmWcVb0Yo/Bv7Og/zNnBEsJH/wh7boUzP2Z6lK0PweJLkb2fw999ALlh1ayTdBGhRuTN/4m/a785vnE5fUbZiT2g2f7WpzAZhn1h/mbxP06+pmOe/byFpHYDHL7HXK4Z74Ell4EkTUYJNN3gnj39fOvBw6i6QbnXwbfftZpr7WE68RFZ8kF0JcR3bnuAK8Zv4Xm7l8UXLilKHOyOZfjSXfsZzxZw2mSuf9tKllX7p2wKLafeFTJzW9JRMAzCF/4b12SG6ex/nkh6nPBIK+y8xcy7cIfMJaY56Iqlue6OfWQUDbfdwtfesQrnOV9ltGMX9ak9aLvv5fvDY3QaFbw/eyvN5V7+I7ORjGHH3TrOV96eYGnVlI11QRd+Yylbe+9D1j1cPfBl3tv4aZ49bKDqBlZZ4vLV1dy3bwAAj8PK165YhSyZNxmNuoex7jFuuO8wOVUrGq8b4HFY+OoVq7DJMkejKZrKPKTyKl+75yC5gobLYeEfLi7HYrgwNBnZ6pi8mTt2I1NXcBE7GuNbDx0BDDx2K197xyqs532DaNd+guEVDEvV/CL5fqrVHsakCFf7zsC26Xr6WvfQLlXzwB4XlrF3UBaMk8uUojpCWC0ZnDYLA+M5vnLPQUbTCi67Ba/DylevWIndYpn8bCmutdg6vo1bT9Evh+g8MEDGY+XmbZ04rPLpWSadB0Lg54G9uhrXunU4mpsJvP1tM8Lz3hIHum6gKhoOl5Vg5el90/KqRmc0zTnNpVy+upo3Lq/k+4+28tnf7UaWTJGtDbroG8tS4rFhs1i4aFk5z3eOoenmmvuxCW36B6rS70TVdXxONx/b3MRPn24nV9DQDYND/Un++6EWImUeZAmeaYuSHmzjSuNu7ORRF13KtSsb+daDR8irGlmXjZirmd5oBY1SP3YZtGVr+WbjSo5GUygFnYcPDfKp/9tFiduGy9tHeV0HHpuHpJLEY/UQaxngzHwvqmwnr6Z4dtvT1GhljKTyZPIauw4/zLD0fcasBi7Vwt7dg/iMIQBWq/vZXIDfSxpP9T4NuoeNljN5v3sPH8ndy/NumT7Jy672m0jbYlS5Irg1i7kNqe78ImGfS1geOTTE1oSCzWknH8/Tdscd9Mi/BSnHTYd1lpTWM+J045ft+Jwl2AN1GEopWtrBRUs/w592fpEfP/0NnP5ziI4+ypslD7XrPmxOToaHtZFNtI+18p1UF5pspUHXUcfaMWrebnojbi/1PvekmG2oWcr+9qc52neQSk0mYTF4UTF4x8T6ta4b3Ph4G0dH0rjtVgbjOa67Yx/DiRyy7OMuuZ/zmhWePRoF/Gz1WfnvjZdS2Xar6Unm4hhNb+Bpxxa+fk8LmlaOvHuY85p1th6NIksSboeVz1+6lJDHzsGBBB6HlWgqzy+fHUGdNt68RgWeg6P80yUjlPscDCZyReJ7cCCOz2ljJJnnp0+3o6g6bruFz7xxMeVeO6NpheZy3+T4lqEkAbc5/sdPDJIt+LlL6ufy1WYFu/v2DaDrxkQUq4ytR6PYLDIhj51/f8dqgKn3umEzqR3/Ry45RqHJRzWmQD5yaJjOaJru0QyyLFHjc9I7luUX92/l82MHOUPLUjj8PLusa7l6/Dm8UpZBpZztO3dhbDiD9miacImLe/cN0DKcxGWzMhhXueH+w1QHnLzYNYbLLlPitvPtZc0Etn4DVdeQShpxSjKGey1qsBmj2Q0HfkThsRtQdQ1Ci3AlB8zPznE3MTu6RknnNba3R2kfSeNxmO/9Fybee6RqLFoJH7Pcx0Xp32JY7Fi0HD+If4TOjIHHoTGcyPON+0wbd3SNki/oaAYEQ8PoBLDqVeQY474je0mPL8bjsJLOq9y5u49oUsHjsNI/bl5zKJlD1Qw03aAm6KJ7NDPr+GIbQdUNqv1T4wcTeX7wIPSnL8KwDRMzqviDLQu0cO/eAXIFDc2Auon50OOw0jeW4wt/3MdQPIeilaE9N0RtMEF3MojHUUY6pjJ4xz4GEzmyhXIsksa7zghgGBEArEF5cn5UdZ2Ay0pTuZfHjwyjGzCcyPOFO/YznMwhSaDpBpc59nORFiBrqcOhpdi3Zwd7XTL94znW1wfJKOq8lllfLkLg54GeV7B4vbg3bJgh7gBIUNUUoG5pCXXLQi/Je59LVNpH0qi6wZJKM2Ra6Xdy7qJSnusYJei2kVU0NjWFsFslZEnCZpG5fHUNl6+uOeFz1pe6+fLbVhYdrwm6Ju+a9/SM858PHkGdWBp4WzjHlcaPqdEHGLOWs7rJTm21v+g5AH54h87G1OOEC5002RNUTrupUDWdgc7DlOWPsMvSTii7hn9Z9zF+vOcn5JJurAkVq2SgSxIBI01oaBv/e0cMXe7HJitYfHsZt6hUaDIuXcESf5Iv3V2FW0txlX4rrSEHw7YAhlTAkEcZKOvg9z7YbXOgylYchkbEKGFrepxEph0ntbCivOhvn1c1vnz3AQBcNsvkXbZhGDzRMsLWrhbUktvRLRk0X46kLYShJJH1IKqq4JeacQdtWCQLdosdu17Fl+7aTyJXQNMNLnFHeDK5ldTwQQJGjkXqRdz3UIKdXV3IsoTbbmFjQxOS9gyNWUjLGt/aM8DO+7YjSxIum4W3rqnmydYRUrkAv1N6CQXKGTQaGLdkyBluDo3Ws3okxcH+BAcH4vSO5vA7bQTdNkIeG0urfMRSedwOK4lsgSODSRRVx26VOTqS5tZWg6v69kE+ge4IcI/3w9zf3006pxLy2ifPyRXMc/rGsnz7oRZ6x7OTYloXdJFWVErcU+PzBR2X3cJAPMeNj7VNjrfIEsuqfBweTKJNO38soxB02RlK5PnfJ9tPOj6aUij12knlVLpjZsKnour4XbZJG7IFHVWHlqEUNzxwiMF4Dk03hedNZVne27oVBSs9f/gS3916Nc/EfGi6gdtu4aotTTx0YAhN12kq97DU1k532oKMG5+RIuztxCnrxORKcrqdvra9XLnX9JyTuQKLyn14HVaCLhvlXjtvX1vNwf4EeVU3hSKZ4nZ1nDdnLfRYIjiHsuy7+2F+PrQIVTNf93v9Lt6StjAg1eAYTNP/yBP8OrYUVdPRDIP14RKeaBlGUc3xFy+voK7EhdUiUeqxs7zKTyyt4HVYyRVkknUXc8Gh63DoecZtpfzd5gj/s9egoOno+pSNmm5QHXCRUVTOql/GgwNPoRsJ7JKDT6w/m3t3FlB1Havs4iPnNvCrrZ3kFJ2g20ZTuYfBRI5Sr51cQWdjpASbVQIDrJZj47tQVA1tho0q6+oDyLK5O0332llZ42e8tQKnXE0iW5h8rwuaPmnjuYtKcXTJ5AoaAZeVcImb/vEsFX5H0XxZUI0iGyOlbiQk1jeU8O4N4Vnnx2NzXe9Ylqyiok3YNNZq3qRkFZVA3XKsGSchsug2J1dcvIXL/Q38x/2HSeVU7FZ58nkWEiHw88AomBnykn32DPlkNIfDZWXppiqsttnL1M4l4MdCd6MZhTKPnevfvqpozJHBJA6rTGTaY6trgjSE3OYdpdPGuzeEZ3wggTnvEI/36Kf/3h5NU1fiwm03P7CX18RYOpImTxm1/gpcWi+wfMZzfOpdl3B0ZBNL+26lpOd+aFxp7m8FzrB2YDf+i1u9GqtVK87Bt/DFI0Nk5BUY7hdZ7h/HWfd+Bj1rqA64COz7LUb2FoatEhqw3lZBvaZgBRyywlvkGNW2B2iw9CMXMjzjuwKbchd2q4FhlHBJ7QdxsI+ueCcRrCQkg3XhTRw+tI6U4ynU+Nn86KFxRjNDqLqBqhlU+p10j2bwOq3ousGDBwa4YEk5f9jZS8/YOHJgO37rKBacWC1Ozqk8k4eP7qaARqHgQU9sYkP4MgqWISqcYf60K8+RoSRuu5W0ovK0tQGfvBW3nMWpW+itXEaDx03rUBKPw8pQIscLfY3ErY0UJgTbyC/G67Dic9oYSuR46OAgI0kFn9OK3Wrhgsal7Oj7GOOFPsiW0trn4G93v0BB15EliesuX87nLl1CezQ9Oan0jWVRdZ1yr4OPn296J3lVx2krIMUOsy/uRJEDWHMFMoMtvHP9W7h7dx+aYRSdU9B0qgJwRn0Jjx4aojrgJJFVefv6Gh47NDzjGgVNp9LvYGNDCQ8dHCLothNLK7gdFgIuG5V+B8mcylvWVPFkSxRNN8ef3Rji/v2D+F22ovHlPjvJnMqbVlaw9eioOd7n5P9dbG4l/Mo9B2bYkFE0Mg4LqqozMJ4zJ+WCSn60l5illLijBqcyToR+9rlWUR9yk1FUKv1Orp92Q9vXaSXfW4JD1hmhFOmsT1PTfjOOXI4Gp5Ou2jVou1RyBR2rLHPZqio2NZUWfUfPjGTojGVIKyoexUI7Ybr1ChxkyBkW9ucr8DqslPscxNIKY55GoqNVeGWNjGbhvn4vvZnspLC0Dadw2iwsrnCTUzXesKyCv9vcVCRMPWMZVF2n1GPnisYEvvFGUoaDGotGc2CcuivOm9VGM9rn5MozV3JpunpyOeTs+sVsChfPb4vKfUXXHIjnUHWdErfM+86s531n1s85frqNHzo7AmdTdLxrwp5yr2PGe+1zOnnX+jretb6u6JyRZP6E8+V0G62yPGvE8/jfrz/OuemOTdn8tjeciWV9HdGu/VRMyyf62hWrxBr8aw0jb3aJkxyzC3wilsUdcMwp7p+7fQ+jaYVwiYuvHCfgR6Mp+saz6AaAUhS6MQyDlqEkzeUerJapetCzeeDHHj8dNJd58ditqBMh/qoSP3ZDwe4JgPPE5VEnvwSRj8KT34Rnv2dm4+bGUQf38INQlowsU++0cH5wmBfj5dS5lyNbnuCQluesc69h7US4/KnYi/R37cVjSNgMiS31l7Nu0WZzbblqPZ623YSf/T4SOuOWMj66Zgmx3R8krQ/gkau5pHETkn0x26N7SOQTWB1+1i+6gOe7R+kyOrEEkzSWuOlpzVDitpMvaKysNZuHpBSVTF5l29FRfvb8NhTHThzOOBsrgpQZIQL2AE6rk79d934ua3gnO/oPs6ZyMfu7rNzybDe67kGWx9gQDlLqteOyWagPuXjfijXctutmVF3Fgp33LFlERVWYg/0JVF1nSaWPj5yziu8/bSUvDVJjqeEz52+aDA8urvBy5XkR/ufJo4CEwyrPmKyeaRuh4+kUlV4nLqsFt91CQ6mHhtKpKosnit40l3k52mJDf/SPuNBQrXYWr1jP+etrOaO+ZE6PZn9fHEXT8TqtnL+onPMXlc85fndPHFXXqfY7+eBZDdNCoDYuXFrJhUsri8a/0Dk26/gSt503rajmTSuqZ3wf5nqdAP965z403cBpc3PJus2oD/4Jn55At9nZdOYmduxWyRZUbJbZJv0V/LD8U1Qo3Qzb6/nUyotxr1qNeyKf42yjkvv7DpBVNFx2C5uaSmcVjek2Ghh89/c61YVeYs563nPRZsYnXme138llm1fwx8e0yWteeeF5/OjJNjTNwGV3c9X5U6Fkp9Uyq1BNv16FVAZHvLi0AlicULqI+tDcNprPt5iz6xcXjZlLDE82V53q+NmOv5RzTnZ8Lk5qc+nyGYnCx5+z0AiBnwfHBF52zNz+ZhgGiViOstoTh1uORlMMJ3NYJJmBeG7G2osEpBUNDANV04lMm4hHknnGMgXesHRmWdGF/LAc/6WubL3ZLAKz9HKzAcPJkqScfrN6172fNfeM2zxsb9pIZqSHZQUVVVJw1vppjnuR1QxNuQxHS91k/VW4MP+u/W4Ji9VKQLfgla2sW7S5eG90PkeuvIE0DmrsFpo9UW5424XHfUndXHPeV4oyoavfluGu1kG6szt439IQQ4n8xJ27g4+ea76uo9EU4RIbN+35AzuNW7BYDHTDw8bQR3jjkuJmIGEfk5OdqgxT5XfgsltRVI13nlHLooop76R+9FmarTW06CpLLHYW2QswywQaKXvjCYWpvtRNQ8hzwslqM+U8OuE9H/NGZnt/TzjBLlnND/d/Zkq4mlee/Bxe/iR+/Gs8/vxTHT8fm7/+jtVFz9Hn/y7Rrv2UT3hdXw6fOPJWX+o2I1ZFx6d2FtRT7OWd6Lt6vE3/9J5L53ydNcHia9aVuE/6dznx9eZXK+DlzjWnev7Jxs92/KWc83JsPN3nLwRC4OeBPunBzxT4XLpAIa/hL3POOHaMkNtmrolJGtkC6KarDpgJdM91jPKmFZVESt282DXOWEahEVPkjwwlAVhaeYK2maeR4/ekT35gx3tg5JDZdGPxJfN/Qkk295cH6yEdZcTpxeYuQ5Ed2HMJNhSinPHWFcT33IkrWs5vykrZM7KHTdWb2Na/jTHZ4LMb/glPdnz2Vpqli3D6S3FO2197vPcBM6uH1Ze6+YjvQr734l6GCof48ttWguemAAAgAElEQVSKQ5Id8Q7a8o/wRGsv/YUBbLIDl16PJqXxu5VZq5Edo7nMS4nbbnpQThuLKnzFX3xpEYvcQRYdtyf4ZEJ0st+nc6LoznyZXbjmd97LmcRPdQI+HRPq8c9R21Tsdb0aovFyPwunfM1T2IMu+PNCCPw8MJQCMPsafCKaA8BXeuLEumhKYUN9CRevqGBvzzhbj8bYGAkRcNl4+OAQ8WyBT1zQRH3ITSLXxhNHRlgfDiJJEkcGk1T6HQTctlM3/BTanvYke7jhmS+Rz43PLJfa9rC5dzSy+dSuX7rIrDedjhKVJUZk+Lt1f0+lu5JIfJBw+zPQejMMPQF1Z1FbEuDFoRep99XzWM9jLA8t59Il75lRnnOS+VYqmwW/3c/S0FJ2De/iDeE3UF9qRkj2juzluqevI6/l8dq8fHzNx7FIfyKRz+N3BDmzbumcz3tScX0ZNp8Kf4neiEAgODWEwM8DQ5nw4GcR+GQsh0WW8ARnr15nGAZ7esZZUxfginW1nNNcyo2PtXH7i71csrySrUdjnN0Ymlwf3bKknFtf6OFAf4JFFV46Y2nOay47daNHO+CuT5nhcWfgpI1gWrufoWt4NzZdpzThmCqXmhyEgb2m5247xd0B08Ts+VwflnQvl0Uuw2v3Qq0B4/2w/UbTRlVhY81V/KbvMb75/DcJuUK8temtJxb36dd4iSK5sWojh0YPcSh2iDXla+hKdPGzfT9D1VXWla8jVUjht/u5fvMXTqna3knFUXhMAoHgFUAI/DyYaw0+Ec3iDTmR5dmFqHcsy0hK4fwl5QBU+Jy8dU0Nt2zrYtvRKOU+B5eurJocv7o2wCOHhnjiyPDEnkpYUvUSwvPRVjO07vCAbJuzEUxBK7Cv+wkKWgEFCVnNEunbC2Vr4MVfmjXmGy+Y12U74530JHumxDDUSMZXye6d32V12WpT3MGs3FW2FGzuiXanEoH0KPui+1B1lXpfPbFcDLdt4bzIRn8jpc5SHul+hL0je9kf3U+Vu4q8lidVSE2WJ50rJC8QCASvVYTAzwNDmX2bnK4bpEZz1CwpOeG5u3vGscoSq2oCk49V+h20DCWJZwvUl7oZTuQnPT5ZltiypJw/7OzjoQNDOKwyDaGXIHIOP6CbpRd1DULNsw5TdZXb9v2c+HgH12RlWlwe2o0s2tAe+L+7QcuDpxzSUXqU8RmebE+yh8Oxw+S1PC1jLTzQ+QAem4cabw3XnnUtYV+YXcO7KOgFNlVvKr54+VKzheTEenSv3UGZq4xSZyl5LT/ZC3uhkCSJel89N+6+Ed3Q8dg8/NeW/8JpdZ6+FrQCgUDwKiEEfh7o+TyS1TKjD3x6PI+mGydMsNN0g7294yyr9uGyT22ha4+mKfc5WFLpJZErzMiqXxcO8sihYdpH0pR4bPSP5059PVTNQvhss53oyCGIHiluogJ0dT/Dr/ffxHi8iw94F7N+89e5IBPjR9HneCAxwFWxHmRnAJxBevqe46sDjzKcGcYiWbikwUy2e6DzAdKFNLIksyy0DLfNjWZo9CR6aB1rpcZTw3ODz9Hob6TSU1ls43Hr0RGblWD/E+S1/KzNPRYCt80sbBH2hbFb7GZf6LrzhbALBII/e4TAzwMjryDZZlt/zwLgK51d4NuGU6TyGuvCwaLHm8u8OKwyiVxh1m1MVovM0kovd+/pp8xr5yv3HDj1usXRFtNr3/IvsOvXZmOKsiVmD3Ggu/tZPvv4PxLXFKp0KFv//yB8Jjbg4tIwt+/7BbtCNWwwzHagWwujdMbNJg95LU9PsgcwIwDNwWYKWoELwxfyaPejxHIxYtkYLwy+QFJJ0jLWwsZlG2e3c9p6dBiKu1C9AiK7NLSURSWL0HTtFbupEAgEglcCIfDzwFDys26RS0Rz2BwWnJ7ZM9z39IzjsllmbHGbzzYmv9NGmdfO8mo/g/HsqdUt1jWIHYW6M83fV73b7B617QfQcD4GBrcd+DlxLc9iw4JhsdOZ6JjsSLWidAX15St53OpiZfkmDlpgZ/9TOK1OQs4QTquTT677JADffP6bqLqK2+ZmY9VGNlZtpDPeiYTE/R3388PdP8QiWbjtyG00BhpPKtqv9Hp32Bfm2rOuFSF5gUDwF4cQ+Hmg5/OzVrFLxHL4S12zZnq3DSd58MAgm5pDRRXojnGyTOvFlT7KvQ4G49kTFithpAXGOs32j9MT6Ma7zLXzsolKUzYnNF8Ef/x7jJYHedAhMxQop0yyYsg203OtPnPydEmSuDRyKd+NfZevjDxjZpVXrOOqtVfRn+ovEsLZPO5j/4/mRtk1soslJUvIqbkFX1N/qYgkOoFA8JeIEPh5YCjKjAx6taCRieepqJ+Z4d4dy3DtH/bRO5Ylo6hcuqL6JRW7mNPLHzoEv30fuIJmrffp2+CirYBEj9NNZ+/TpvgWsvR4y7jLrtNl5Hhj0+WsrNtM5+COWYvIaIZGe7yd8fw4Jc4SPr3u0zQEGmgOFifrzSWO6yvX0xRoIqfmRPhbIBAIXmGEwM8DI68gu4rX2ZOxHAbgmyXB7mg0xXi2MNG6VX7JbQHn9PJb7gNdAb0AqlK8DS7aQo/Lz3XbvkpSSWKRLZxVspzt9iQpNEokKyvrNhOuP5dw/bmzPn1nvJOQM0QkECGlpOhOdtMQaDgl+8O+8Cu+pi4QCAQCEyHw88BQFKRAoOixgfY4ydEcGDPHN5V5UFQNGctkk4rTa5ABiUGw2CE7BnbfVPMXVYGxTp5x2xnMDFLiLCFTyNCaHURxh2i0+ZBtXjpljbnkNhKIYLfYSSmpl+V9i/C3QCAQvDoIgZ8HxyfZxUey7Hm0BzWvsf3Oo2x+75KiHvBuu5W1dSWsqvXz5lWnHp4/KdFW03u/+Ktw+F7wlEJJxDw22k5CzXJQK+Cyugg5Q1S6K/nwyg9z8wGzi9l8BFt43wKBQPDnjRD4eaDn80j2qUz58aEMhbyGL+hE1wzGhzJFAt8ZS+N1WnnnGbVU+E7chOYl0/Us2Dyw7HLwVcKe35qiX74EY+QIdyqDuCrW8/W1HyOej091UfNUn5JgC+9bIBAI/nyZmd59GpEk6TJJko5IktQmSdK1sxxvkCTpUUmS9kqS9IQkSXXTjv2nJEkHJEk6JEnS96WTFiVfGAzDMNfgp3nw/lInhm6gKCqyRSJYWeyhd0TTeB0Wyr2z16d/WeTiMLgXwmeZ3chqN4DdCx1PArC15wk6rBbe3PxW1pSvKSraEvaFRREXgUAgeJ2wYAIvSZIF+CHwZmAF8H5JklYcN+y/gJsNw1gDfBW4YeLcc4HzgDXAKuBMYMtC2TonhQIYRlGI3mKXqWoKsPyc6hnheYDOaJqGUs/JG6W8FLqfM5uzNJxHT7KHpwe201O1HIb282LrvfwmtosqfwPryted/msLBAKB4M+GhQzRnwW0GYbRDiBJ0q3AFcDBaWNWAJ+d+Plx4M6Jnw3ACdgBCbABQwto6wnRJ+vQTwl8Jq5gd1lpPqMCT6DYS49nCoxlCpz7UjrAndQY3QzPly3lqJrg6seuRtEUZGBpQePwc9uRDB0jP0Jvqld46gKBQPA6ZiFD9LVAz7Tfeycem84e4F0TP78T8EmSVGoYxjZMwR+Y+PegYRiHjr+AJElXSZK0Q5KkHSMjI6f9BcBUoxl5WqGbTEJBliRcvpnFbzpjaQAiZQvQBW34AOTGIXIeD3Y+SF7L0xBowO8I4nCXEdDhDMOOYeh0xjtP//UFAoFA8GfDgq7Bz4PPAVskSdqFGYLvAzRJkhYBy4E6zJuCiyRJOv/4kw3D+F/DMDYahrGxvLx8QQw81ip2eie59Hgel98+a4vYzlgah1WmJnCKvdNnY7QDWh82/zcMOHg3ZMfIyDb6Un347X6skpVKTyXvibyFynyaMSWFNdZGxBD5kwKBQPB6ZiFVoA+KtlrXTTw2iWEY/Ux48JIkeYG/MgxjXJKkjwPbDcNITRy7HzgHeHoB7Z2VSYGftgafSSh4S2ZPoOuIpqkPuU/YH37exNrhzk9CPgm6arZW7d8JrhBbH/5nHHWr+NI5XyKpJM2s+MHDVBtBOn0VRJQC4Vzq5V1fIBAIBH/WLKQH/wKwWJKkRkmS7MD7gLunD5AkqUySpGM2XAfcNPFzN6Znb5UkyYbp3c8I0b8S6Mf1gtc0nWxSwR2YKfAZRWUokaexzPPyL7zzZoh3A7rZPAbAU0GyejXPaUlWWwOsq1g3lRVfuoiwt5bzCxC2uqcK3wgEAoHgdcmCefCGYaiSJH0aeBCwADcZhnFAkqSvAjsMw7gbeANwgyRJBvAU8KmJ028HLgL2YSbcPWAYxj0LZetcGPniJLtsQsEAPIFZ1t+jGQAaXm5hm+7nTHH3lIO7zNwOd9ZVsO1Gno63ossyW5rfWnzOcb3Vi5rPCAQCgeB1x4Iu1BqGcR9w33GPfWnaz7djivnx52nAJxbStvliKGaI/liSXTpuCv7x2fMAXbE0VlkiHHoZAj/aDvtug9qNcOEXYfTopGCPu/zs3P0j1tdsIlQ9yza4ab3VBQKBQPD6RmRinYTj1+Az8TySJOHyzewB3xFLU1viwjZLe1hGO+b2rkc7oH8XtD8OngrY8BGwe6C0aXLIH6M7GbBa+au6TafnxQkEAoHgL5ZXO4v+NY9+nMCn4wourw35OBHPqxp9Y1kipbOsv492wH3/Ak9+C+7/vPn78cfv/Sw89jVofxKWXGaK+zQe6nyI3x7+LYOZQX6464f0JHsQCAQCgeBECA/+JBhKAclqRZJNQc/E83iCM8PzPaNZdIPZE+xibZAcAE2BRB9s/wksugiUjJkh3/4EjBw2u8K5y8297seubxg81fsUd7TdgcPiYGXpSoYzw3TGO0UhG4FAIBCcECHwJ8FQ8pMZ9Lqmk0kWKAv7Zox7oWOUgXiWWTfHlS4CrWD+c4dMsb/3s2bJWUmGyAXgqzYbyNinMuA74538vuX3DGeGObvqbHZIOxjODL+s9q0CgUAgeH0gBP4kGIoyGZ7PpgoYhoH7uAz67liGm7Z2oBR0vvnAYb78tpXFLWJLIlBzhtn5bcOVMHIEkoPg9EMhC6veZYr6tDX6jngH//T4P5EqpCh3lXP1+qu5uOFi0b5VIBAIBPNCCPxJ0PN5pIkM+swJMuiPRlPkChrVfheqrnM0mioW+PQI2Byw9PKpBDt3ienRO7xTiXfTku8e6XqEVCHFitIVKJpCZ6JTdIITCAQCwbwRAn8SjLyCPBGiT4/nkQC3v9iDbwi50XVIKyo+p5PmMm/xk4xPJMQFJrrhnmTPuqqr9Kf78dg8KJoiQvICgUAgOGWEwJ8EI59HdpveeDqex+mxYbEWZ9B7HVbOqC9hbTjAm1ZUFXvvAPEekK3gq5p6bI4963tH9gLw+TM/j27oIiQvEAgEglNGCPxJMJQ8UkkJYNagny2DPppS8DqtXLqyavYiN/Ee8NeAbDnp9XRD55m+Z6jx1HBuzbkL01NeIBAIBH/xiH3wJ8FMsrNj6AaZhDIjwQ4gmjL3ypd5Z2lAYxgQ74VA/byudzB2kLH8GJtrNwtxFwgEAsFLRgj8SdDzCrLDQTZVQNeNWUvURlN5vA4LLvssHno6Cmpuav19DgzD4Onepyl3lbMstOx0mC8QCASC1ylC4OfAMAzTg7fbySRML/34BDuAWEqhdDbvHczwPEDw5GvoLWMtDGeHhfcuEAgEgpeNEPg5MAoFMAwku4P0uLlF7kQh+lnD8zCVYOetmv34BD2JHm45eAuyJLOqbNXLtl0gEAgEr29Ekt0cGNN6wUd7k+TTKulxhUC5a3JMrqCRyKmUemcKP2Cuv/uqwVL8p+5J9LBtYBt22U4sF+POtjvJqlmqPFX0pfpE1rxAIBAIXhZC4OfgWCe5VN7C4W0DgMQzt7Ww+b1LJkV+NG3eBJTPlWB3XGvXnmQP1z59LUOZIWRJpinYhITEouAiDAxRZ14gEAgELxsRop+DYwKfSMuoBQNP0I6uGYwPZSbHHMugn9WDz4xCITMjwa4j3kEsF6PSXUmDr4FLGy6l3m9m2dtkmyhqIxAIBIKXjfDg5+BYq9hgpRuJLPm0irfEQbByaq/7pMB7ZvHgT5RgZ5j73SUkHFYHG6s2srFqo6gzLxAIBILThhD4OTi2Bu8v91DZVCBY6WbZpuqiNfhoSiHgsmG3zhIMifeY3eJ8NUUP96Z6Oa/mPN4QfgNNwaZJQRfCLhAIBILThRD4OTgm8IbFit1lpbopUCTucCyDfq4Eu5qiBLvx3Dht421cUHcBW8JbFsx2gUAgELy+EWvwc3BsDR6rKeDyLF56NKmcuILdeM+M9fcXh18E4IzKM06vsQKBQCAQTEMI/Bwc8+B12fTALZbi4jMZRSVb0GYX+OwYFNJFAq/qKruGdrGkZAkBR2DhDBcIBALB6x4h8HNwLMnOOIEHH02aNwCzZtD3PGeG6KdVpDs8epi0mmZj5cYFslggEAgEApOTCrwkSf84n8f+EjHyCpLViq6bv1usxR78yImazIx2wJPfhOGD8Oz3zN+BHYM7KHGU0BxsXnDbBQKBQPD6Zj4e/EdmeezK02zHaxJDySM5HOiaqfCypfjPFUvlkSUIeY7z4GNtUMiCuwx0DWJt7Bnew/bB7dT76kWdeYFAIBAsOCfMopck6f3AB4BGSZLunnbIB4wutGGvBfR8HslhR1MNAOTj1uCjKYWQx45FPk6wSxcBEihp8NfQ7XBz/bbrSSkpHuRBNlRtEFviBAKBQLCgzLVNbiswAJQB3572eBLYu5BGvVYwFLNV7DEP3nL8GnwqT+nx3jtAqBEWvdEU+PP+kTs77yGpJFkeWk5BL4hStAKBQCBYcE4o8IZhdAFdwDmSJDUAiw3DeESSJBfgwhT6v2iMvIJkd6DO4sEbhkEslaep3DP7yVYHlC9jj5agJ9lDyBmioBewylZRilYgEAgEC85JC91IkvRx4CogBDQDdcBPgDcurGmvPoaiIHs86OpMDz6RU1E0Y/YStQBKhl49xz1H72FV2So+ve7TdCe7RSlagUAgELwizKeS3aeAs4DnAAzDaJUkqWJBrXqNYCh5pFAJmjbTgz9Wg77cN0uIXtc5lB3ip/1dhCpW8p4l78Ftc9MQaHhF7BYIBAKBYD4CnzcMQzmW+S1JkhUwFtSq1wh6Po9kt0958NOy6GOpiT3ws3jwPWMtfClzmHGLjUjCSywXw21zzxgnEAgEAsFCMZ9tck9KkvQFwCVJ0iXA74F7Ftas1wZG/liS3YQHP20f/OGBBEPxHIlcYcZ5naOtpA2VKlcpFslCZ7zzlTJZIBAIBAJgfgJ/LTAC7AM+AdwHfHEhjXotYBgGhqIgORxomo5Flib3r3fHMvz6uS7aRlJ89Z6DdMcyReeGHUFUA3KGKpLqBAKBQPCqcNIQvWEYOvBT4KeSJIWAOsMw/uJD9Mfq0Et2B7pqFJWpPRpNkS/oVPgcqLrO0WiK+tKpELxdU1lr87O2/k1sWfIOkVQnEAgEglec+ZSqfUKSJP+EuL+IKfTfWXjTXl2OdZKTHXbTg58Wnm8u82JgkMgVsMoyzWXeonMHk314JStvjrxZiLtAIBAIXhXmk2QXMAwjIUnSx4CbDcP4siRJf/GFbqY8eDt60igqU1tf6mZDQwk1ARd/fWZ9kfcOMJgZxIFMyFfzitosEAgEAsEx5rMGb5UkqRp4L/CnBbbnNcMxD15yONBUvWiLXEHTcdqsbIyEZog7wEB6iCqLC8nmesXsFQgEAoFgOvMR+K8CDwJthmG8IElSE9C6sGa9+ujT1+A1o6jITbagAeC0WWaeZ+gM5WJU2fxFrWIFAoFAIHglmU+S3e8xt8Yd+70d+KuFNOq1wPQQvaaqRR58TjEF3m2fKfCjuVEKWo4qR+UrY6hAIBAIBLMwHw/+dcn0JDtd02f14F2zCPxgehB0lSpX2StjqEAgEAgEsyAE/gQUr8EbRR58ZsKDd80Soh9ID2DRdcqFwAsEAoHgVWQ+2+Qa5/PYXxp6fiJE7zjxGvyJPPgKLFjs3hnHBAKBQCB4pZiPB/+HWR67/XQb8lqjaA1e04vK1GZP4MEbhsFAeoAqrGAXtecFAoFA8OpxwiQ7SZKWASuBgCRJ75p2yA84F9qwVxtDMRvNSJKErhpFjWZOJPAJJUFWyVCNBWwn6BMvEAgEAsErwFxZ9Evh/7d35+FRVNnDx7+nl+zsILIEEgVZlLAKqICAyjCKKIoDDiCLygiCoq/b6Iyig7vjKIgy6AjqKPoTF9RRRlZRR2UzLAIqSGQVA8iStbur7vtHLyaQbjqYppNwPs/TT7qqq6pPKsvpe+vWPQwAagOXllh/GLg+lkFVBiZQSc4Y42/Bl7gGX+i1SHQ5cDhK3wbnH2Dn5VRHIiRogldKKRU/YRO8MWYeME9EzjHGfHECY6oUbI8HSUzAtoOV5Eq34Mu6Re6n/J8Q26KhI0m76JVSSsVVNNfgBwXmoneLyCIRyRWR4TGPLM5MsQdHQgK2z5/gnUe04MONoK/rSiFBHNpFr5RSKq6iSfD9jDGH8HfX5wAtgNtjGVRl4O+i909TC0e04L0Wp1i74fsFsH9raP1P+T/RKKGmf0Fb8EoppeIommIz7sDXS4A3jTEH5SSYgtV4inHUqIltBVvwvyZ496Ef6f3D32C7G5LrQP9HKKjRkIOeg5yd1CSwkSZ4pZRS8RNNgn9fRDYBhcA4EWkAFMU2rPizPR6ciQm/tuBLdNGnHt5KSnEupDYEywv7NvOT0/9BoJEzUGBGB9kppZSKo2N20Rtj7gLOBboYY7xAAXBZrAOLN1PswZGYiG0Fu+j9Cd4Ywy5nUxwC5OWC0w31WrDu53Xszt+N5ckHhwucCXGMXiml1MkumpnsUoDxwHOBVY2BLrEMqjLw7d1L8eYtFO/YDfzaRe+1DLnuxhxq0BnqnwH9HmS728XLG19m68GtPL39I7aL0UpySiml4iqaLvpZwCr8rXiAnfiry1Xb2vDFP/5IwYoVOGvXpuC7PdgdB4da8IUeCzEWDncy1KwNSbXIObCJAm8BDVMa4svfT454SY/z96CUUurkFs0o+tONMY8BXgBjTAFQrZunnq05uOrVI/G007B9Bvvw4dBc9IVeC7ddjCt4Tb5wP43TGmMZiyJfES5jyEjWUrFKKaXiK5oWvEdEkgEDICKnA8UxjSrOEk8/DVeD+lgHD2JcDXHUqBHqoi/0WrhNMa7gLHYF+3DUbkT7Bu3p0rAL525fS3pqozhGr5RSSkWX4CcD84F0EXkVOA8YHcug4i0hPZ2Gd9+NZ+tWbEdD9v9oSnXRu+xinMHiM/l72e2CtIQ0Lj39UlK2rddb5JRSSsXdMRO8MeZjEVkFdMffNX+zMWZvzCOLs4T0dBLS09m/YR/8mFuiBe8jwS7G5Qhc3Sjczy6Xh9qJtUlxJYO3QG+RU0opFXfHTPAissgYcwHwnzLWVXvBqWqD98EXemxcwS76hDQo2McuOUSj1EZgecD26Sx2Siml4i5SudgkIAWoLyJ1+HVgXU2gyQmIrVKwLBuHQ5DANfcCj48EU4zTIVCrKQW/bOUXVyqdG3YGT75/J52HXimlVJxFasH/CZiE/773Vfya4A8Bz8Q4rkrjqFrwXotU8fpvc6/ZhF27V4KdRKO0Rv7uedAWvFJKqbiLVC72aeBpEZlojJl2AmOqVGzLDg2wg0CpWKfXv1CrCbvsIvAV+bvof9nuX68teKWUUnEWzVS1J21yB7AsE7oHHgIteIcXnImQUp9ddhH1HEkku5LBk+ffSAfZKaWUirNoJro5qdk+u1ShmUKvRTIecCdDqj/BN3Yk+V/0Fvq/ahe9UkqpOItpgheR/iLyrYhsFpG7yni9uYgsEpG1IrJURJoG1vcRkewSjyIRuTyWsYZj+UzpBO+xSBEvuJM5bCwOY2gcvNKhg+yUUkpVEtFMdIOINAGal9zeGLPsGPs4genARcAOYIWIvGeM2VBisyeAl40xL4lIX+BhYIQxZgnQIXCcusBm4OOov6sKZFl26S56j0WSeMCVwq783eBKorEdeNGb7++6d0Z1WpVSSqmYieY++EeBIcAGwAqsNkDEBA90BTYbY34IHOd1/GVmSyb4tsCtgedLgHfLOM5g4KPAHPgnnG0ZXG4n4C8VW+i1SMQD7nrszt+NuBM51evxb+wp8HfdK6WUUnEWTVPzcqCVMaa88883AbaXWN4BdDtimzXAFcDTwCCghojUM8bsK7HNUODJst5ARMYCYwGaNWtWzvCiY/vs0LS0xT4b20Ci8YAriZ15O2mQ3ICEwgNgTGAWu7SYxKGUUkqVRzTX4H8A3DF6/9uA80Xka+B8/KVog70EiEgjoB3w37J2NsbMNMZ0McZ0adCgQUwCtCxTYhY7f2iJFGNcSezO203jtKZge6H4sP8avA6wU0opVQlE04IvALJFZBElqsgZY246xn47oVRZ9KaBdSHGmF34W/CISBpwpTHmQIlN/gC8Y4zxRhFnTPhH0f9aSQ5jcBsPB8WQ78unca02sHcnFOzzJ/iajeMVqlJKKRUSTYJ/L/AorxVASxHJxJ/YhwJ/LLmBiNQH9htjbODPwItHHOPqwPq4KXkffKHXwoGFC4udlv+WuMZ1z4Atn0PBfi00o5RSqtKIpprcS8dzYGOMT0Qm4O9edwIvGmO+EZEHgJXGmPeA3sDDIhIctHdjcH8RycDfA/DJ8bx/RTDGlLoPvtBj4Q5UkltXuJufC37GG0zoBXv9LXgtFauUUqoSiFRs5v+MMX8QkXX4R82XYozJOtbBjTEfAh8ese7eEs/nAnPD7JtDnIvaGNtgoFQL3m0X85OziDk7P8HrdPHE109zp9NN+qGdgNEWvFJKqUohUgv+5sDXAScikMrIsgKlYl0lWvCmiO0UUmx8NExuhM/2keNMIP1A4IYBTfBKKaUqgcZH3U4AACAASURBVEjFZnYHvv544sKpXGyffwYbZ4lBdgmmmNOdqRgbDnsPk5aQRkaNTMjd7N9Ju+iVUkpVAjrlWgSW7+gWfJrDx6nORNrVaMvp9c/kktMvIX3n2l8TvN4mp5RSqhLQYjMR2NbRLfgaTg/5xiItsQbdGncjvUY6pNT/dSedh14ppVQlcMwELyKXishJ+UHgyBZ8gccixeEl3/jA4SLFFWitp9T7dSdtwSullKoEokncQ4DvReQxEWkd64AqkyNb8EVefyW5PGOBOEkLTkubUvfXnbQFr5RSqhI4ZoI3xgwHOgJbgNki8oWIjBWRGjGPLs5CLXhnsAXvI1m8FDhdIJDqCiTzpNogDnAlg+Ok7OxQSilVyUSVjYwxh/Dfr/460Ah/YZjVIjIxhrHFXagFH7wP3mOTLB7yHP6EnxpsrTscgAPy9sD+rfEIVSmllColmmvwA0XkHWAp/qIzXY0xvwfaA/8vtuHFl13iPnhjDEU+iyQ85DscJDgScDsDNXj2b4Vt/4Of1sL8uzTJK6WUirtobpO7EviHMaZU/XdjTIGIXBubsCoHq8R98EVeG2MgSbwUSInWO8C+zZDaAGo1gcN7/Mt1M+MUtVJKKRVdgp8M7A4uiEgy0NAYk2OMWRSrwCoDu8Qo+rxAqdgEU0T+kQm+XgtwJ/uTu9PtX1ZKKaXiKJoE/yZwbollK7Du7JhEVIlYgWvwDqeDAo+/Um6i8ZCHoU7JBF83E/o/4m+512uhrXellFJxF02CdxljPMEFY4xHRBJiGFOlYVsGhwgOh1Dk9bfg3cZDgbFoeuTtcHUzNbErpZSqNKIZRZ8rIgODCyJyGbA3diFVHpbPLjXJDcbgsorIxy7dRa+UUkpVMtG04G8AXhWRZwABtgPXxDSqSsK2DI7gNLUeC5fx4HFYGHFogldKKVWpHTPBG2O2AN1FJC2wnBfzqCoJy2fjDExyU+i1cNnFFDltcLg0wSullKrUoqomJyKXAGcCSSL+hGeMeSCGcVUKtmVwuH6dpjZZPBRigSNBE7xSSqlKLZqJbmbgn49+Iv4u+quA5jGOq1KwfDbOEtfg05y+QILXFrxSSqnKLZpBducaY64BfjHG3A+cA5wR27AqB9tX4hq81yLN4SXfWOBwaoJXSilVqUWT4IsCXwtEpDHgxT8ffbVnWb+24As9FqlOH3nGhzjcJLuS4xydUkopFV401+DfF5HawOPAasAAz8c0qkrC9tk4AvPN7z5QCAcOYLs8pCbUwCFaNU4ppVTlFTHBi4gDWGSMOQC8JSIfAEnGmIMnJLo4syyD0yVs21fAwo0/cx57SUo5TEKzlHiHppRSSkUUsRlqjLGB6SWWi0+W5A6/3ge/ZW8eXtumfpKhUGwKPO54h6aUUkpFFE0/8yIRuVKC98edRPxd9MJp9VMRBF9xPoUOIb123XiHppRSSkUUzTX4PwG3Aj4RKcJ/q5wxxtSMaWRxZowJdNE7OLVWEp2a1aGzSabQlUx6nTrxDk8ppZSKKJqZ7GqciEAqG2MbjDE4nEKR1yYtyUWzBMDjItWlt8gppZSq3I6Z4EWkV1nrjTHLKj6cysO2/LXgnS4HxYFKcpYpAIeTtIS0eIamlFJKHVM0XfS3l3ieBHQFVgF9YxJRJWGFErxQ7PPXhffa+SBOUlw6il4ppVTlFk0X/aUll0UkHXgqZhFVEnYgqTucDgp8/ha8xyoAt0tb8EoppSq945mtZQfQpqIDqWws369d9EVef7Ivtv1d9HoNXimlVGUXzTX4afhnrwP/B4IO+Ge0q9ZsK9iCF4q8FmJsiu0icKSQ4tYueqWUUpVbNNfgV5Z47gPmGGM+j1E8lUapQXaFNi7joQCbBGciCc6EOEenlFJKRRZNgp8LFBljLAARcYpIijGmILahxZflK92Cd9vFFIlFql5/V0opVQVENZMdULJ0WjKwMDbhVB7BFrzD6aDYZ5MixRRgkZpwUk4LoJRSqoqJJsEnGWPygguB59X+InSwBe90+VvwqQ4f+cZHakK1nsBPKaVUNRFNgs8XkU7BBRHpDBTGLqTKITTIzhVowTs85BuL1ERN8EoppSq/aK7BTwLeFJFd+OehPxUYEtOoKoHQbXKBa/Bp4vW34BNrxzkypZRS6tiimehmhYi0BloFVn1rjPHGNqz4KznIrthnU4t8DJCaVCu+gSmllFJROGYXvYjcCKQaY9YbY9YDaSIyPvahxZdtGUQEcQTug5d8AFITtVSsUkqpyi+aa/DXG2MOBBeMMb8A18cupMrB8tk4nYKIv5qcMXkgDlKSdBS9Ukqpyi+aBO8UEQkuiIgTqPYzvdiWweHyn55in4Ux+f5Kcm69D14ppVTlF80gu/nAGyLyz8DynwLrqjXb8rfgjTEU+2x/qVhxkerWeeiVUkpVftEk+DuBscC4wPIC4PmYRVRJWD4TGmBnDHjtAsThJNmVfOydlVJKqTg7Zhe9McY2xswwxgw2xgwGNgDTYh9afNmWHboHHsBnCklxJeOQ4ynAp5RSSp1YUWUrEekoIo+JSA7wALApplFVApbP4HQ6KPb6a8F7Pb+QWpwP+7fGOTKllFLq2MImeBE5Q0TuE5FN+Fvs2wExxvQxxpwkLXh/F32d4p38cug79hzMYftHt2qSV0opVelFasFvAvoCA4wxPQJJ3ToxYcWf5TM4XQ5/JbnidSxJMGS7hUd9u9i+86t4h6eUUkpFFCnBXwHsBpaIyPMicgH+qWpPCrbPDg2yy3G78QrUsww+EXJc0YxNVEoppeInbII3xrxrjBkKtAaW4J+T/hQReU5E+p2oAOPFsn5twZPQFnEmcjApDdcpbclo2i3e4SmllFIRRTMXfT7wGvCaiNQBrsJ/69zHMY4trgoPe9i3K48iZzI1qEOWuzYZp3bksrNvJr1GerzDU0oppSIqV19zYJramYFHtXUwt5Bd3x0gMcVF4XfgPrUYd4KLrFotNLkrpZSqEvSm7jLs/ykf27ZJqZmA5TMkFXpBwO1KindoSimlVFQ0wZehTsMU6pyaim0bjIC7hr86rstZ7afgV0opVU3ocPAy1D4lhd7DWnNgTwGf/nSAwrxcKNIWvFJKqapDW/Bh1GqQTPOz6uFNduB2BlvwmuCVUkpVDZrgj6HIa5MgHgCcLu2iV0opVTVogj+GYp+FW/wteLdWklNKKVVFaII/hiKvjcsR6KLXBK+UUqqK0AR/DEVeC6f4AHDpIDullFJVhCb4CIwxeCwbJ4EWvFtb8EoppaoGTfARFPtsjAFHYJCdy5US54iUUkqp6GiCj6DYZwOEWvBuV2I8w1FKKaWipgk+gmKvFXjmBXHicrrjGo9SSikVrZgmeBHpLyLfishmEbmrjNebi8giEVkrIktFpGmJ15qJyMcislFENohIRixjLUuR19+CF+MFceBy6MR/SimlqoaYJXgRcQLTgd8DbYGrRaTtEZs9AbxsjMkCHgAeLvHay8Djxpg2QFfg51jFGk6xL9CCl0CCF03wSimlqoZYtuC7ApuNMT8YYzzA68BlR2zTFlgceL4k+Hrgg4DLGLMAwBiTZ4wpiGGsZQpeg8d4cTicOB3OEx2CUkopdVximeCbANtLLO8IrCtpDXBF4PkgoIaI1APOAA6IyNsi8rWIPB7oEShFRMaKyEoRWZmbm1vh30BR8Bq88Wn3vFJKqSol3oPsbgPOF5GvgfOBnYCFv8pdz8DrZwOnAaOO3NkYM9MY08UY06VBgwYVHlzwGrzBqwleKaVUlRLLBL8TSC+x3DSwLsQYs8sYc4UxpiNwT2DdAfyt/exA974PeBfoFMNYyxS8Bm8bHy6HjqBXSilVdcQywa8AWopIpogkAEOB90puICL1RSQYw5+BF0vsW1tEgs3yvsCGGMZapiKvTaLLgWV7dYCdUkqpKiVmCT7Q8p4A/BfYCPyfMeYbEXlARAYGNusNfCsi3wENgQcD+1r4u+cXicg6QIDnYxVrOEVei0SXA5/t1XvglVJKVSkxbZYaYz4EPjxi3b0lns8F5obZdwGQFcv4jqXYZ5PoduIr9uJOrBHPUJRSSqlyifcgu0qt2KcteKWUUlWTJvgIirw2SW4nXtuLy5EQ73CUUkqpqGmCj6DIa5HoBJ+xcDk1wSullKo6NMFHUOyzSXEafMZogldKKVWlaIKPoMhrkey08GLj1gSvlFKqCtEEH4YxBo9lk+ywsDA4NcErpZSqQjTBh1HsszEGkhwWXmNwuxLjHZJSSikVNU3wYRQH5qFPctj4MLicmuCVUkpVHZrgwwjOQ58gXmwMbldSnCNSSimloqcJPoxgJTkXHv9XbcErpZSqQjTBh1EUaME7JZDgtQWvlFKqCtEEH0bwGrwmeKWUUlWRJvgwgtfgxWgXvVJKqapHE3wYwWvwYhcD4HIlxzMcpZRSqlw0wYdR5A224L2AdtErpZSqWjTBh1Hss0l0ObDsIgBcbm3BK6WUqjo0wYdR5LVIdDvwWUUgDtwOrQevlFKq6tAEH0aRzyLR5cTn84A4cDlc8Q5JKaWUipom+DCKvf4uep9VrAleKaVUlaMJPoxin02S26kJXimlVJWkCT6MIq9FktuB11cM4tRr8EoppaoUTfBhhK7BW15wOHA6nPEOSSmllIqa9juHsfdwMQ6BVE+BjqJXSilV5WiCL8OP+/L54of91ElxYzt/xnOq4BRtwSullKo6tIu+DJt/zqNuipvM+qnYxkexD0Qk3mEppZRSUdMEX4aWp9SgYc0kDhd5EbGokajT1CqllKpatIu+DM3qpXDfpWeyZW8eP69JYmeiVpJTSilVtWgLPoxm9VLo0+oUEpw2Lh1gp5RSqorRBB+JMXhtLy5nQrwjUUoppcpFE3wktg8vNm5N8EoppaoYTfCRWB4sDE7toldKKVXFaIKPxPLiNUZb8EoppaocTfCRWF582LicOopeKaVU1aIJPhLLg88YXC5twSullKpa9D74SCwPPoy24KsIr9fLjh07KCoqincoSilVoZKSkmjatClud/RjwjTBR2J5NcFXITt27KBGjRpkZGTo1MJKqWrDGMO+ffvYsWMHmZmZUe+nXfSRWB58xsbt0gRfFRQVFVGvXj1N7kqpakVEqFevXrl7JzXBR2J58GJwOTTBVxWa3JVS1dHx/G/TBB+B8fnvg3e5tNiMUkqpqkUTfAQ+qxgAl3bRqyiJCMOHDw8t+3w+GjRowIABAwCYPXs2EyZMOGq/jIwM2rVrR1ZWFv369eOnn36KaZy9e/dm5cqVMX2PcJYuXcr//ve/uLy3UicTTfAReH0FANqCr8a27Stgybc/s21fQYUcLzU1lfXr11NYWAjAggULaNKkSVT7LlmyhLVr19KlSxceeuihComnvCzLivl7aIJX6sTQUfQR+HzBFnxynCNR5fXB2l3sPhB5QMovBR4WbNiDZRucDuGitg2pkxJ+zoNGtZMYkNX4mO998cUX85///IfBgwczZ84crr76aj799NOoY+/VqxdTp049an1GRgYjR47k/fffx+v18uabb9K6dWv279/PmDFj+OGHH0hJSWHmzJlkZWWV2rewsJDRo0ezZs0aWrduHfoAApCWlsaf/vQnFi5cyPTp01m+fDkvvvgiANdddx2TJk0iJyeH/v3707lzZ1avXs2ZZ57Jyy+/TEpKCosWLeK2227D5/Nx9tln89xzz5GYmEhGRgYrV66kfv36rFy5kttuu43Zs2czY8YMnE4n//73v5k2bRo9e/aM+twopaKnLfgIfL4iQHDrRDfV0r48D5ZtqJXsxrIN+/I8FXLcoUOH8vrrr1NUVMTatWvp1q1bufb/4IMPaNeuXZmv1a9fn9WrVzNu3DieeOIJAO677z46duzI2rVreeihh7jmmmuO2u+5554jJSWFjRs3cv/997Nq1arQa/n5+XTr1o01a9aQnJzMrFmz+Oqrr/jyyy95/vnn+frrrwH49ttvGT9+PBs3bqRmzZo8++yzFBUVMWrUKN544w3WrVuHz+fjueeeC/u9ZWRkcMMNN3DLLbeQnZ2tyV2pGNIWfAQ+XxGIE5foaapqomlpb9tXwP3vf4PPtqmbmsCNfVrQrF7Kb37vrKwscnJymDNnDhdffHHU+/Xp0wen00lWVhZTpkwpc5srrrgCgM6dO/P2228D8Nlnn/HWW28B0LdvX/bt28ehQ4eoWbNmaL9ly5Zx0003heIr2cJ3Op1ceeWVoWMNGjSI1NTU0Pt9+umnDBw4kPT0dM477zwAhg8fztSpU7nooovIzMzkjDPOAGDkyJFMnz6dSZMmRf19K6ViQzNXBD6rGBwOXA49TdVRs3op3HfpmWzZm8fp9dMqJLkHDRw4kNtuu42lS5eyb9++qPZZsmQJ9evXj7hNYqJ/wKfT6cTn8/3mOME/Q5bT6TzmdkfepnOs23ZcLhe2bQPo7IJKxYF20Ufg8xWDOHBrudhqq1m9FPq0OqVCkzvAmDFjuO+++8J2tVeknj178uqrrwL+AWz169cv1XoH/3X91157DYD169ezdu3asMd69913KSgoID8/n3feeSfUjb5t2za++OILAF577TV69OhBq1atyMnJYfPmzQC88sornH/++YC/Oz54KSDYwwBQo0YNDh8+XFHfvlIqDE3wEfgsf4J3Oo7dulGqpKZNm4a6xI80e/ZsmjZtGnrs2LHjN73X5MmTWbVqFVlZWdx111289NJLR20zbtw48vLyaNOmDffeey+dO3cu81idOnVi1KhRdO3alW7dunHdddfRsWNHAFq1asX06dNp06YNv/zyC+PGjSMpKYlZs2Zx1VVX0a5dOxwOBzfccAPgHxtw880306VLl1I9BJdeeinvvPMOHTp0KNfgQ6VU+YgxJt4xVIguXbqYir6v99ulU3h9/9dc328ajdOOfU1XxdfGjRtp06ZNvMOolnJychgwYADr16+PdyhKnbTK+h8nIquMMV3K2l5b8BEEW/B6DV4ppVRVowk+Ap/tAYdTE7w66WVkZGjrXakqRhN8BD7Lqy14pZRSVZIm+Ah8lsef4PU+eKWUUlWMJvgIfLZHb5NTSilVJWmCjyDUgtcueqWUUlWMJvgIvJYXp8N9zBm7lAqKZbnYUaNGkZmZSYcOHejQoQPZ2dkAGGO46aabaNGiBVlZWaxevRrwzx3fuXNnsrKyQhPU+Hw+LrzwQgoKjl09Lycnh7POOqv8J6GCPPXUU2HjvO6669iwYUPUx5o8eXJo7v4jnXvuuWWuHzVqFHPnzj1q/dKlS0M/z98qnmV7VfWnCT4c28JnLFzaPV+97d8K3y/wf60AsS4X+/jjj5OdnU12djYdOnQA4KOPPuL777/n+++/Z+bMmYwbNw6Af/7znzz99NN8+OGHoeT23HPPMXz4cFJSjn/mvoqaIvdYIiX4F154gbZt21bI+1Tn0rUnovyvqry07zkcy4sXg9upCb5KWv82HNoZeZuC/bDpP2AsECe0vgRS6obfvmYTOOuKY751rMrFhjNv3jyuueYaRITu3btz4MABdu/ejdvtpqCggIKCAtxuNwcOHOD9999n/vz5YY+1atUqxowZA0C/fv1C62fPns3bb79NXl4elmXxzjvvlFmidvLkyWzZsoXNmzezd+9e7rjjDq6//nqMMdxxxx189NFHiAh/+ctfGDJkCEuXLuWJJ57ggw8+AGDChAl06dKFQ4cOsWvXLvr06UP9+vVZsmRJqTh79+7NE088QZcuXUhLS+Pmm2/mgw8+IDk5mXnz5tGwYcOjvrcNGzbQu3dvtm3bxqRJk0IzDaalpZGXl4cxhokTJ7JgwQLS09NJSPi1iuT8+fOZNGkSKSkp9OjRI7Q+Pz+fiRMnsn79erxeL5MnT+ayyy5j9uzZvPfeexQUFLBlyxYGDRrEY489FvHnOG7cOFasWEFhYSGDBw/m/vvvZ/HixUydOpV3330X8H9gfPbZZ3nnnXf4+OOPue+++yguLub0009n1qxZpKWlkZGRwZAhQ1iwYAF33HEHP//8MzNmzMDlctG2bVtef/31iHGo6kNb8OFYHnzYuBxaKrbayt/rT+7Jtf1f8/dWyGFjWS72nnvuISsri1tuuYXi4mIAdu7cSXp6emibpk2bsnPnTm688UYeeughRo4cyd13383f/vY37r77bhyO8H/2o0ePZtq0aaxZs+ao11avXs3cuXP55JNPIpaoXbt2LYsXL+aLL77ggQceYNeuXbz99ttkZ2ezZs0aFi5cyO23387u3bvDxnHTTTfRuHFjlixZclRyP1J+fj7du3dnzZo19OrVi+eff77M7TZt2sR///tfli9fzv3334/X6y31+jvvvMO3337Lhg0bePnll0Mt+6KiIq6//nref/99Vq1aVeryyYMPPkjfvn1Zvnw5S5Ys4fbbbyc/Px+A7OzsUBndN954g+3bt0f8Ph588EFWrlzJ2rVr+eSTT1i7di19+vRh06ZN5ObmAjBr1izGjBnD3r17mTJlCgsXLmT16tV06dKFJ598MnSsevXqsXr1aoYOHcojjzzC119/zdq1a5kxY0bEGFT1oi34cCwvPmNwOTXBV0lRtLTZvxXm/wyWF1LqQ89boW7mb37rWJWLffjhhzn11FPxeDyMHTuWRx99lHvvvTfs8Zo1a8bSpUsB2Lx5Mzt27KBNmzaMGDECj8fD3/72t1CZV4ADBw5w4MABevXqBcCIESP46KOPQq9fdNFF1K3r7+EIV6IW4LLLLiM5OZnk5GT69OnD8uXL+eyzz7j66qtxOp00bNiQ888/nxUrVhxVFOd4JCQkhK6Jd+7cmQULFpS53SWXXEJiYiKJiYmccsop7Nmzh6ZNm4ZeX7ZsWSjGxo0b07dvX8D/wSAzM5OWLVsC/lK5M2fOBODjjz/mvffeC10CKSoqYtu2bQBccMEF1KpVC4C2bdvy448/lvogdqT/+7//Y+bMmfh8Pnbv3s2GDRvIyspixIgR/Pvf/2b06NF88cUXvPzyy8yfP58NGzaEyvd6PB7OOeec0LGGDBkSep6VlcWwYcO4/PLLufzyy8txZlVVpwk+HMuDD4NLu+irr7qZ0P8R2LcZ6rWokOQeFItysY0aNQL8JWNHjx4dSipNmjQp1TrcsWPHUdf977nnHqZMmcLUqVO57rrryMjI4O677w5VoYtGsEb8sZSnrGzJkrJwfGVl3e5fB8JGKqMbLLV7rO3KwxjDW2+9RatWrUqt/+qrr8r1flu3buWJJ55gxYoV1KlTh1GjRoXOxejRo7n00ktJSkriqquuwuVyYYzhoosuYs6cOWUer+TP6j//+Q/Lli3j/fff58EHH2TdunW4XPqv/2SgXfThhBK8tuCrtbqZ0PKiCk3uEJtyscEubWMM7777bmiE+8CBA3n55ZcxxvDll19Sq1at0IcBgE8++YTGjRvTsmVLCgoKcDgcOByOowaw1a5dm9q1a/PZZ58BREz+kUrUzps3j6KiIvbt28fSpUs5++yz6dmzJ2+88QaWZZGbm8uyZcvo2rUrzZs3Z8OGDRQXF3PgwAEWLVoUeo8TXVa2V69eoRh3794dujTQunVrcnJy2LJlC0CppPq73/2OadOmESza9fXXXx/Xex86dIjU1FRq1arFnj17SvWcNG7cmMaNGzNlyhRGjx4NQPfu3fn8889DZXrz8/P57rvvjjqubdts376dPn368Oijj3Lw4EHy8vKOK0ZV9ejHuHAsLz5j43ImHntbpY5wrHKxwUFTAF9++WVUxxw2bBi5ubkYY+jQoUPoeurFF1/Mhx9+SIsWLUhJSWHWrFmhfYwxTJkyhTfeeAOAsWPHMmzYMHw+H88999xR7xG8xisipQbZHWny5MmMGTOGrKwsUlJSSpWozcrKok+fPuzdu5e//vWvNG7cmEGDBvHFF1/Qvn17RITHHnuMU089FYA//OEPnHXWWWRmZoZK0wZj7d+/f+hafKwNGjSIxYsX07ZtW5o1axbq8k5KSmLmzJlccsklpKSk0LNnz9AHj7/+9a9MmjSJrKwsbNsmMzMzNGCwPNq3b0/Hjh1p3bo16enpoa73oODPPlhJrEGDBsyePZurr746NBZjypQppS65gH8U/fDhwzl48GDodsratWuXOz5VNcW0XKyI9AeeBpzAC8aYR454vTnwItAA2A8MN8bsCLxmAesCm24zxgyM9F4VXi52zwae/+TPpJx+IcM6T6y446qY0XKx8Td58mTS0tK47bbb4h1KtTJhwgQ6duzItddeG+9QVByVt1xszFrwIuIEpgMXATuAFSLynjGm5OwUTwAvG2NeEpG+wMPAiMBrhcaYDrGK75hsLz4Mbm3BK6XiqHPnzqSmpvL3v/893qGoKiaWXfRdgc3GmB8AROR14DKgZIJvC9waeL4EeJfKwvLi1S56pcpl8uTJ8Q6h2lm1alW8Q1BVVCwH2TUBSt74uSOwrqQ1QPB+pkFADRGpF1hOEpGVIvKliJR5b4eIjA1sszJ4n2iFCQ6yc2mCV0opVfXEexT9bcD5IvI1cD6wEwjOrdg8cF3hj8BTInL6kTsbY2YaY7oYY7o0aNCgYiOzPIH74DXBK6WUqnpi2UW/Eyg5q0PTwLoQY8wuAi14EUkDrjTGHAi8tjPw9QcRWQp0BLbEMN7SLC9ebNzu5BP2lkoppVRFiWULfgXQUkQyRSQBGAq8V3IDEakvIsEY/ox/RD0iUkdEEoPbAOdR+tp9zBlfMRbg1KlqlVJKVUExS/DGGB8wAfgvsBH4P2PMNyLygIgEb3nrDXwrIt8BDYEHA+vbACtFZA3+wXePHDH6PuYsqwgjDlxOnSpARU/LxVYcLRdbOQ0bNoxWrVpx1llnMWbMmNCc/hX5e3jvvfeycOHCqGNauXJl2HknfoulS5eGrTYY7m+5MonpNXhjzIfGmDOMMacbYx4MrLvXGPNe4PlcY0zLwDbXGWOKrGBTSwAAGC9JREFUA+v/Z4xpZ4xpH/j6r1jGWRafzwPiwK3lYqu17Ye38+mOT9l+OHIhkGhpudiKo+Vif7vyloudPHkys2fPjrjNsGHD2LRpE+vWraOwsJAXXngBqNjfwwceeIALL7ww6ri7dOlSrgqM0YqU4GOtIv7O4j3IrtLy+YpAHLgc2oKviubnzOelb16K+Hh69dNMWDSBx1Y8xoRFE3h69dMRt5+fE77MaknBcrFAqFxsefTq1Ss0BWk0ylsutmTltyOtWrWK9u3b0759e6ZPnx5aP3v2bAYOHEjfvn254IIL2L9/P5dffjlZWVl0796dtWvXAv4EMWLECM455xxatmwZquxmjOH222/nrLPOol27dqGZ9Y5sDU+YMIHZs2czderUULnYPn36HBVnyZZvWloa99xzD+3bt6d79+7s2bOnzO8tWC72tNNOK5UM0tLSQjFOmDCBVq1aceGFF/Lzzz+Htpk/fz6tW7emU6dOvP3226H1+fn5jBkzhq5du9KxY0fmzZsXOl9XXHEF/fv3p2XLltxxxx1hz3nQuHHj6NKlC2eeeSb33XcfAIsXLy5VIGbBggUMGjQI8Be6Oeecc+jUqRNXXXVVaArajIwM7rzzTjp16sSbb77J1KlTadu2LVlZWQwdOvSYcRzLxRdfjIggInTt2pUdO3YAFft7WLL3JCMjg/vuu49OnTrRrl07Nm3adNT2JX+PgrMsHvmzzsnJoXXr1gwbNow2bdowePDg0AfIjIwM9u71V5NcuXIlvXv3JicnhxkzZvCPf/yDDh06RCz5/P7779OtWzc6duzIhRdeyJ49e7Btm5YtW4YqAdq2TYsWLcjNzSU3N5crr7ySs88+m7PPPpvPP/88FPuIESM477zzGDFiRNj3i5Ym+DB8viJwOLUFX43tL9qPbWxqJtTENjb7i/ZXyHG1XKyWi4XqXy7W6/Xyyiuv0L9/f6Bifw+PVL9+fVavXs24cePCXmopKdzP+ttvv2X8+PFs3LiRmjVr8uyzz4Y9RkZGBjfccAO33HIL2dnZ9OzZM+y2PXr04Msvv+Trr79m6NChPPbYYzgcDoYPHx6q2bBw4ULat29PgwYNuPnmm7nllltYsWIFb731Ftddd13oWBs2bGDhwoVhCwmVhzZPw/BZxdqCr8L6Z/Q/5jbbD2/n0cJH8dk+6iTV4bp215FeI3w5z2hpuVgtFwtVp1zsunXrQq3Fn376iYSEBJ566ikAFi1aRL169Y7aB2D8+PH06tUrYuKD8v8eluWKK/zTpXTu3LlUD0o4Zf2sgVLz/A8fPpypU6dWyLTKO3bsYMiQIezevRuPx0Nmpr941ZgxY7jsssuYNGkSL774YqhY0MKFC0uNITl06FCoB2bgwIEkJ1fM3VuavcLwWR5N8NVceo107ux6JzkHc8iolVEhyT1Iy8WGXy5Jy8X6xbNcbLt27UIDNidPnkxGRgajRo2K+H3ff//95Obm8s9//jO0Lpa/h8FzGe3PLdy5D/e7WfL38Hh+BydOnMitt97KwIEDWbp0aWhGx/T0dBo2bMjixYtZvnx56Pu0bZsvv/ySpKSko44V7d9ZNLSLPgyvFbgGL5rgq7P0Gun0bNqzQpM7aLlYLRdbPlWpXOwLL7zAf//7X+bMmVOqm72ifg9jadu2baGR/K+99ho9evQA/N3xwSmBgz1TEP3v4MGDB0MfZkpWVgT/HR/Dhw/nqquuwul0AtCvXz+mTZsW2ib4AauiafYKQ1vw6rfQcrFaLrY8qlK52BtuuIHmzZuHzs8VV1zBvffeW6G/h7HSqlUrpk+fzpgxY2jbtm1opP99993Htddey1//+ld69+4d2v7SSy9l8ODBzJs3j2nTpoW9HDF58mSuuuoq6tSpQ9++fdm6dWvotYEDBzJ69OjQhzOAqVOncuONN5KVlYXP56NXr14VOj4iKKblYk+kii4X+/3823jNu5trez9K0xpNj72DijstFxt/Wi42NrRc7G+Xk5PDgAEDWL9+/Ql935UrV3LLLbdEHIUfrUpTLraq89keQFvwSqn40nKxVdcjjzzCc889V66xLhVJs1cYXssDDk3wSpWHlouteFoutmJkZGSc8Nb7XXfdxV133XVC37MkHWQXhlV4AAoP4D64K96hKKWUUuWmCb4s+7fi3bMODu7AtegB2L/12PsopZRSlYgm+LLkfocvqQ4k1cZl+WBf9NOGKqWUUpWBJviyNDgDX3JN8BXhdiZAvRbxjkgppZQqF03wZambibfjCBwNz8Lx+0ehbma8I1JVRLBwSTQmT55MSkpKqaIm0ewfrtJcRStZgONEe/fdd8tVDlYpdTRN8GH4UuviqpOhyb2a82zfTt6yZXiOUQgkVurXr1/u259OVIIvizGm1NSysaIJXqnfTu8BC8Nn+/QWuSrs0Ecf4d39U8RtrAMHOLxoEVgWOJ3UuOACnBFm+XI3OpWav/99uWPZsmULN954I7m5uaSkpPD888/TunVrwD+l7ezZs7nzzjtDhVyC/v3vfzN16lQ8Hg/dunXj2Wef5Z577qGwsJAOHTpw5plnHnV/bVpaGjfffDMffPABycnJzJs3j4YNG5KTkxOqQtagQQNmzZpFs2bNSu27b98+rr76anbu3Mk555wTmn41JyeH3/3ud3Tr1o1Vq1bx4Ycf8swzz/DRRx8hIvzlL39hyJAhLF26lHvvvZcaNWqwefNm+vTpw7PPPovD4WDOnDk89NBDGGO45JJLePTRR0PxBqdOnTt3Lh988AFjx47lvffe45NPPmHKlCm89dZbnH766eU+70qd7LQFH4Ym+OrPt38/WBaOWrXAsvzLMTB27FimTZvGqlWreOKJJxg/fnzotbS0NMaMGcPTTz9dap+NGzfyxhtv8Pnnn5OdnY3T6eTVV1/lkUceITk5mezs7DInzwhXOnXixImMHDmStWvXMmzYsDKn0b3//vvp0aMH33zzDYMGDQpVRQP4/vvvGT9+PN988w0rV64MW/p1+fLlTJs2jQ0bNrBlyxbefvttdu3axZ133snixYvJzs5mxYoVpabqPdK5557LwIEDefzxx8nOztbkrtRx0gwWhib4qi2alrZn+3b27NuL8fpw1a1D/T+NJSFCOc/jkZeXx//+9z+uuuqq0Lrg3OFBN910Ex06dCg1veuiRYtYtWoVZ599NgCFhYWccsopx3y/cKVTv/jii1CZzREjRnDHHXccte+yZctC21xyySXUqVMn9Frz5s3p3r07QMTSr127duW0004D4Oqrr+azzz7D7XbTu3dvGjRoAPjnVV+2bFmZpUuVUhVHM1gYmuCrv4T0dBrefTeerVtJyMys8OQO/mpetWvXjlgtqnbt2vzxj39k+vTpoXXGGEaOHMnDDz9crveLtnRqecWiVOyRrx9PmU6lVHjaRR+Gz/hwO9zxDkPFWEJ6Omm9esUkuQPUrFmTzMxM3nzzTcCfuNesWXPUdrfeeiv//Oc/Qwn5ggsuYO7cuaER9vv37+fHH38E/Enc6/WWK45zzz2X119/HfCXgS2rKlavXr147bXXAPjoo4/45ZdfyjxWuNKv4O+i37p1K7Zt88Ybb9CjRw+6du3KJ598wt69e7Esizlz5nD++ecD0LBhQzZu3Iht27zzzjuh9zjRpWKVqo40wYfhs31aC16VW0FBAU2bNg09nnzySV599VX+9a9/0b59e84880zmzZt31H7169dn0KBBoe77tm3bMmXKFPr160dWVhYXXXRR6Dr32LFjycrKYtiwYVHHNW3aNGbNmkVWVhavvPLKUdf8wV8yc9myZZx55pm8/fbbRw3CCxo0aBBZWVm0b9+evn37lir9evbZZzNhwgTatGlDZmYmgwYNolGjRjzyyCP06dOH9u3b07lzZy677DLAX4xjwIABnHvuuaVqhw8dOpTHH3+cjh07huqwK6XKR8vFhvHCuhdIdiUzrE30/0RVfGm52PhaunQpTzzxxHHVQ1dKHVt5y8VqCz4Mn+3DKc54h6GUUkodF+2DDsNn6zV4pcqjd+/e9O7dO95hKKUCtAUfhtf26ih6pZRSVZYm+DAs28Lp0C56pZRSVZMm+DC8tle76JVSSlVZmuDDOFB0gO9++Y7th+NThEQppZT6LTTBl2HboW18nfs1n+74lEeXP6pJXkUt3uVie/fuTatWrejQoQMdOnQIHbu4uJghQ4bQokULunXrRk5ODgCff/45WVlZdOnShe+//x6AAwcO0K9fv6iqxi1dujQ0NW48xLOynlKVnSb4MuQcyuHU1FNpUbsFPttHzsGceIekYuRgbiE/rt/HwdzCuLx/LMrFvvrqq2RnZ5OdnR2av/5f//oXderUYfPmzdxyyy3ceeedAPz973/nww8/5KmnnmLGjBkATJkyhbvvvhuH4/j/PVTUFLnHogleqfB0mHgZMmtlUjOhJvuL9uNyuMiolRHvkFQ5fb9yD3m/FEfcpijPy9Y1uRgbxAGZ7RuQlBZ+3EVanURadmlY7lhOZLnYcObNm8fkyZMBGDx4MBMmTMAYg9vtpqCggIKCAtxuN1u2bGH79u0Rb3ebP38+kyZNIiUlhR49eoTWT548mS1btvDDDz/QrFkzHn744TJL1I4aNYqkpCRWrlzJoUOHePLJJxkwYABFRUWMGzeOlStX4nK5ePLJJ+nTpw+zZ89m5cqVPPPMMwAMGDCA2267jfnz5x/XuVDqZKEJvgzpNdK5s+ud5BzMIaNWBuk1YjNPuYqvwjwPxobEVBfF+T4K8zwRE/zxGjt2LDNmzKBly5Z89dVXjB8/nsWLFwOly8Xef//9oX1Klot1u92MHz8+VC72mWeeiVi8ZvTo0TidTq688kr+8pe/ICLs3LmT9MB8+y6Xi1q1arFv3z7+/Oc/c80115CcnMwrr7zCbbfdxpQpU8Ieu6ioiOuvv57FixfTokULhgwZUur1DRs28Nlnn5GcnMyll17KyJEjGTlyJC+++CI33XRTqExsTk4Oy5cvZ8uWLfTp04fNmzczffp0RIR169axadMm+vXrx3fffRc2lmjOhVInM03wYaTXSNfEXoVF09I+mFvIZ4e+w7YMyWluOvfPoFaD5AqN40SXi3311Vdp0qQJhw8f5sorr+SVV17hmmuuCbt9hw4d+PLLLwF/udhGjRphjGHIkCG43W7+/ve/07Dhr+dy06ZNZGZm0rJlSwCGDx/OzJkzQ68PHDiQ5GT/OYxUovYPf/gDDoeDli1bctppp7Fp0yY+++wzJk6cCEDr1q1p3rx5xASvlIpME7w6adVqkEyPP5zBgT0F1G6YUuHJHU58udgmTZoA/mpsf/zjH1m+fDnXXHMNTZo0Yfv27TRt2hSfz8fBgwepV69eqfebMmUKr7/+OhMnTuSxxx4jJyeHqVOn8uCDD0b9/rEoK+tyuUoN+NOyskpFRwfZqZNarQbJND+rXkySO5zYcrE+n4+9e/cC4PV6+eCDDzjrrLMAf8v6pZdeAmDu3Ln07du3VFJ9+eWXufjii6lbty4FBQU4HA4cDgcFBQWl3qN169bk5OSEKrzNmTMn7PceqUTtm2++iW3boWv2rVq1omfPnqHr6N999x3btm2jVatWZGRkkJ2djW3bbN++neXLl4eOczylc5U6WWgLXqkKFCwXG3Trrbfy6quvMm7cOKZMmYLX62Xo0KG0b9++1H7BcrH/+Mc/gNLlYm3bxu12M336dJo3bx4qF9upU6dSA8uKi4v53e9+h9frxbIsLrzwQq6//noArr32WkaMGEGLFi2oW7duKPEGY549ezYff/xxKOaLL76YhISEUH34oKSkJGbOnMkll1xCSkoKPXv2DFu3fdq0aYwePZrHH388NMguqFmzZnTt2pVDhw4xY8YMkpKSGD9+POPGjaNdu3a4XC5mz55NYmIi5513HpmZmbRt25Y2bdrQqVOn0HHCnQullJaLVdWIloutGkaNGsWAAQMYPHhwvENRqkrRcrFKKaWU0i56pdSJNXv27HiHoNRJQVvwqlqpLpeclFKqpOP536YJXlUbSUlJ7Nu3T5O8UqpaMcawb98+kpKSyrWfdtGraqNp06bs2LGD3NzceIeilFIVKikpqdQdOtHQBK+qDbfbTWZmZrzDUEqpSkG76JVSSqlqSBO8UkopVQ1pgldKKaWqoWozk52I5AI//sbD1Af2VkA4Jzs9jxVDz2PF0PNYMfQ8VoyKPo/NjTENynqh2iT4iiAiK8NN+aeip+exYuh5rBh6HiuGnseKcSLPo3bRK6WUUtWQJnillFKqGtIEX9rMeAdQTeh5rBh6HiuGnseKoeexYpyw86jX4JVSSqlqSFvwSimlVDWkCV4ppZSqhjTBAyLSX0S+FZHNInJXvOOpzETkRRH5WUTWl1hXV0QWiMj3ga91AutFRKYGzutaEekUv8grFxFJF5ElIrJBRL4RkZsD6/VcloOIJInIchFZEziP9wfWZ4rIV4Hz9YaIJATWJwaWNwdez4hn/JWNiDhF5GsR+SCwrOexnEQkR0TWiUi2iKwMrIvL3/VJn+BFxAlMB34PtAWuFpG28Y2qUpsN9D9i3V3AImNMS2BRYBn857Rl4DEWeO4ExVgV+ID/Z4xpC3QHbgz83um5LJ9ioK8xpj3QAegvIt2BR4F/GGNaAL8A1wa2vxb4JbD+H4Ht1K9uBjaWWNbzeHz6GGM6lLjfPS5/1yd9gge6ApuNMT8YYzzA68BlcY6p0jLGLAP2H7H6MuClwPOXgMtLrH/Z+H0J1BaRRicm0srNGLPbGLM68Pww/n+qTdBzWS6B85EXWHQHHgboC8wNrD/yPAbP71zgAhGRExRupSYiTYFLgBcCy4Kex4oSl79rTfD+f6rbSyzvCKxT0WtojNkdeP4T0DDwXM9tFALdmx2Br9BzWW6BbuVs4GdgAbAFOGCM8QU2KXmuQucx8PpBoN6JjbjSegq4A7ADy/XQ83g8DPCxiKwSkbGBdXH5u9Z68KpCGWOMiOi9l1ESkTTgLWCSMeZQyUaQnsvoGGMsoIOI1AbeAVrHOaQqR0QGAD8bY1aJSO94x1PF9TDG7BSRU4AFIrKp5Isn8u9aW/CwE0gvsdw0sE5Fb0+wWynw9efAej23EYiIG39yf9UY83ZgtZ7L42SMOQAsAc7B39UZbMCUPFeh8xh4vRaw7wSHWhmdBwwUkRz8lyn7Ak+j57HcjDE7A19/xv+Bsytx+rvWBA8rgJaB0aIJwFDgvTjHVNW8B4wMPB8JzCux/prASNHuwMES3VQntcD1yn8BG40xT5Z4Sc9lOYhIg0DLHRFJBi7CP55hCTA4sNmR5zF4fgcDi43O9oUx5s/GmKbGmAz8/wMXG2OGoeexXEQkVURqBJ8D/YD1xOvv2hhz0j+Ai4Hv8F+7uyfe8VTmBzAH2A148V8vuhb/tbdFwPfAQqBuYFvBf4fCFmAd0CXe8VeWB9AD/7W6tUB24HGxnstyn8cs4OvAeVwP3BtYfxqwHNgMvAkkBtYnBZY3B14/Ld7fQ2V7AL2BD/Q8Hte5Ow1YE3h8E8wn8fq71qlqlVJKqWpIu+iVUkqpakgTvFJKKVUNaYJXSimlqiFN8EoppVQ1pAleKaWUqoY0wSt1khMRK1D5KviosIqKIpIhJSoPKqVOHJ2qVilVaIzpEO8glFIVS1vwSqkyBepaPxaobb1cRFoE1meIyOJA/epFItIssL6hiLwj/trsa0Tk3MChnCLyvPjrtX8cmHEOEblJRDYEjvN6nL5NpaotTfBKqeQjuuiHlHjtoDGmHfAM/mpjANOAl4wxWcCrwNTA+qnAJ8Zfm70T/pm8wF/rerox5kzgAHBlYP1dQMfAcW6I1Ten1MlKZ7JT6iQnInnGmLQy1ucAfY0xPwQK4/xkjKknInuBRsYYb2D9bmNMfRHJBZoaY4pLHCMDWGCMaRlYvhNwG2OmiMh8IA94F3jX/FrXXSlVAbQFr5SKxIR5Xh7FJZ5b/Dr25xL883B3AlaUqFqmlKoAmuCVUpEMKfH1i8Dz/+GvOAYwDPg08HwRMA5ARJwiUivcQUXEAaQbY5YAd+IvN3pUL4JS6vjpJ2alVLKIZJdYnm+MCd4qV0dE1uJvhV8dWDcRmCUitwO5wOjA+puBmSJyLf6W+jj8lQfL4gT+HfgQIMBU46/nrpSqIHoNXilVpsA1+C7GmL3xjkUpVX7aRa+UUkpVQ9qCV0oppaohbcErpZRS1ZAmeKWUUqoa0gSvlFJKVUOa4JVSSqlqSBO8UkopVQ39fzzEznqraOA3AAAAAElFTkSuQmCC\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}