{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"name":"2022-01-12-seq-mab-mushroom.ipynb","provenance":[{"file_id":"https://github.com/recohut/nbs/blob/main/raw/P296669%20%7C%20Sequential%20Batch%20Learning%20in%20Stochastic%20MAB%20and%20Contextual%20MAB%20on%20Mushroom%20and%20Synthetic%20data.ipynb","timestamp":1644607243833}],"collapsed_sections":["aJdhotjJuyCg","F_7QyQIsvraK","6SzDe38oviFg","Y635PDE6vgdI","UiTTsyhFvdwc","-IL38-vIzIG1","4L_OJQQEvP4k","oAuUEthsu9TG"],"authorship_tag":"ABX9TyNovRrJ5FnlW7E7l5a6PFNk"},"kernelspec":{"name":"python3","display_name":"Python 3"},"language_info":{"name":"python"},"widgets":{"application/vnd.jupyter.widget-state+json":{"2b72cbdd08374942859625d2047795f2":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_396e4164d325484e9956f597c2e133e0","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_46ce32b9cbc74045b96fc2fbc93f0575","IPY_MODEL_596798eaeac14aaea8c3cc476b7f329f","IPY_MODEL_c60ea6ae22a2457786ca6b8e3446e3f2"]}},"396e4164d325484e9956f597c2e133e0":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"46ce32b9cbc74045b96fc2fbc93f0575":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_342e5fcfeeb04720834331a4274b5e28","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_9cbda8d73a0a41f8acb9c3bdacdc7a9b"}},"596798eaeac14aaea8c3cc476b7f329f":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_042addbfa66d4dd4a8543124cc312905","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":23,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":23,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_3e9ddf2ca50f4c24bcf276eeeee63d01"}},"c60ea6ae22a2457786ca6b8e3446e3f2":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_ce6fa32eee4044df843b45c5693e065e","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 23/23 [00:09&lt;00:00,  3.28it/s]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_5ac6a207b502473fa8bd11d2adcc1201"}},"342e5fcfeeb04720834331a4274b5e28":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"9cbda8d73a0a41f8acb9c3bdacdc7a9b":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"042addbfa66d4dd4a8543124cc312905":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"3e9ddf2ca50f4c24bcf276eeeee63d01":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"ce6fa32eee4044df843b45c5693e065e":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"5ac6a207b502473fa8bd11d2adcc1201":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"3d6f8baea53a484faab78b4c50fbf620":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_9689c9d0d5354f7782671e9b301c5840","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_4550c6895fa244efb6dd5cbc9dd98324","IPY_MODEL_b70c31ffb8784655ad0d4e59e12343d6","IPY_MODEL_c14978517f8047718a1c084947d702b0"]}},"9689c9d0d5354f7782671e9b301c5840":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"4550c6895fa244efb6dd5cbc9dd98324":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_60d54986769b4e859de45bbce659b7f1","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"  6%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_0181512ceea0401db61675636ae223b5"}},"b70c31ffb8784655ad0d4e59e12343d6":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_4b34f4a343b242abaef58ebe57e07a4d","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"","max":16,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":16,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_476d47eec56d4ff0bec2900ed3505ba2"}},"c14978517f8047718a1c084947d702b0":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_27609bb3e0594b5084e01fae4b04cc9c","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 1/16 [00:08&lt;02:13,  8.89s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_606916b9b63b43969a01e84ec142f341"}},"60d54986769b4e859de45bbce659b7f1":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"0181512ceea0401db61675636ae223b5":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"4b34f4a343b242abaef58ebe57e07a4d":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"476d47eec56d4ff0bec2900ed3505ba2":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"27609bb3e0594b5084e01fae4b04cc9c":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"606916b9b63b43969a01e84ec142f341":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"b8cc488f50764d7a80c565e120acf3bd":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_73139f73ef7d4769807b6dc081a48fb1","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_8c0a7a56a8d946bf953390fd61675c4d","IPY_MODEL_63ead4d72b27495cac3bb6a385895123","IPY_MODEL_a1d8d9e611534997828fa51cdab2bd0a"]}},"73139f73ef7d4769807b6dc081a48fb1":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"8c0a7a56a8d946bf953390fd61675c4d":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_afe8ad0a833b4f399a1f6d8367786fd3","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"  7%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_c55e9f4461c24c9f88ba2098ff957634"}},"63ead4d72b27495cac3bb6a385895123":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_4472bedf87ec41aeb3306d6364386540","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"","max":15,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":15,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_edc2a60460ef48bdb78972f403c14175"}},"a1d8d9e611534997828fa51cdab2bd0a":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_d971324520f34ce5b3b6994ddd0d0102","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 1/15 [00:08&lt;02:03,  8.81s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_6d4282c72b80490cbdb027d3592849ea"}},"afe8ad0a833b4f399a1f6d8367786fd3":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"c55e9f4461c24c9f88ba2098ff957634":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"4472bedf87ec41aeb3306d6364386540":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"edc2a60460ef48bdb78972f403c14175":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"d971324520f34ce5b3b6994ddd0d0102":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"6d4282c72b80490cbdb027d3592849ea":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"77454d4570db40ad9d90ed0f9d80f019":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_efe993ec7ad34cfaac8ca10da8b06b25","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_a845d33bdbc048af807e19eb89699a1b","IPY_MODEL_e16dcd3650f64c61b1c53445f96d13d5","IPY_MODEL_124e06ade51442c3af9d364455ec278f"]}},"efe993ec7ad34cfaac8ca10da8b06b25":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"a845d33bdbc048af807e19eb89699a1b":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_a87b20c55b6547d7bad171519b9c2029","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 29%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_ce8d826cba9f41f8bb33997ceceab3c0"}},"e16dcd3650f64c61b1c53445f96d13d5":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_3ad0f66ba72c47ce8537870c6af56c0c","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"","max":14,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":14,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_3f68a6b94c614e87aea69a1fecf11d2e"}},"124e06ade51442c3af9d364455ec278f":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_3b3af273314945ef951076ac38deb74c","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 4/14 [00:08&lt;00:16,  1.64s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_ffd1831574034bc59136368ad3e60108"}},"a87b20c55b6547d7bad171519b9c2029":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"ce8d826cba9f41f8bb33997ceceab3c0":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"3ad0f66ba72c47ce8537870c6af56c0c":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"3f68a6b94c614e87aea69a1fecf11d2e":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"3b3af273314945ef951076ac38deb74c":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"ffd1831574034bc59136368ad3e60108":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"d8e841629bac43fbbea7f0a924aa9b94":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_dcde233fe4804bd8be4a35f7a781a985","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_a31bcdbac3f54376814d4b9cd40d545b","IPY_MODEL_806b8ca3d7fe4e05b22b038a82eb4545","IPY_MODEL_a26e8dfe75d540208dae3d35a0444551"]}},"dcde233fe4804bd8be4a35f7a781a985":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"a31bcdbac3f54376814d4b9cd40d545b":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_9a8c2b7f24ad47dea0cc84b99639f151","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 23%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_721da9071ffc493d97a523d2f2ac4f59"}},"806b8ca3d7fe4e05b22b038a82eb4545":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_0f5ac0bc528b4fa1b8d5a7aafeb36c79","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"","max":13,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":13,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_a37b2e9543074546bf43c7741f3a5650"}},"a26e8dfe75d540208dae3d35a0444551":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_f95e707810f3426eb20134bad69e6a7f","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 3/13 [00:08&lt;00:21,  2.19s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_c065af81128d4149b26e71b7a6cb4c14"}},"9a8c2b7f24ad47dea0cc84b99639f151":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"721da9071ffc493d97a523d2f2ac4f59":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"0f5ac0bc528b4fa1b8d5a7aafeb36c79":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"a37b2e9543074546bf43c7741f3a5650":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"f95e707810f3426eb20134bad69e6a7f":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"c065af81128d4149b26e71b7a6cb4c14":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"153549b4beb84e54b5fe4bec2eecf075":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_d7e2ff7ce49d44588f16e7abfbb9885c","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_6a821f0913784d64b2e98282b6ae15a1","IPY_MODEL_f736e0b03dea46dea371a5abcbcde9d7","IPY_MODEL_f0f773f30026425fa0dfdcefd1f7d2dc"]}},"d7e2ff7ce49d44588f16e7abfbb9885c":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"6a821f0913784d64b2e98282b6ae15a1":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_914c417613a446c28921a6cb430218dc","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"  8%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_76e6aa6b457347af8ef3146d3d1e7437"}},"f736e0b03dea46dea371a5abcbcde9d7":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_3d87aae69d6149fc98c0b0a642fd9194","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"","max":12,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":12,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_4a588c2572094039983c106ffd558d86"}},"f0f773f30026425fa0dfdcefd1f7d2dc":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_346efff4305041a5bd4913a00ebae73a","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 1/12 [00:07&lt;01:26,  7.87s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_3a36af3ccfd54942b5e4643c373f60a2"}},"914c417613a446c28921a6cb430218dc":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"76e6aa6b457347af8ef3146d3d1e7437":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"3d87aae69d6149fc98c0b0a642fd9194":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"4a588c2572094039983c106ffd558d86":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"346efff4305041a5bd4913a00ebae73a":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"3a36af3ccfd54942b5e4643c373f60a2":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"d1538f6f3b7641ccb48e2bbdef414389":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_3810f04feb094355bdc304447a3356b5","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_2528ae76b0a7423a8910adf401a5514a","IPY_MODEL_d8d6807eee1e44f78e10dcd5251c3e58","IPY_MODEL_21eac8d36aaf4fa2a0565329dedd965b"]}},"3810f04feb094355bdc304447a3356b5":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"2528ae76b0a7423a8910adf401a5514a":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_411e0eeb49834c04bbca486de6d10b5d","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"  9%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_d3282d958c424f84a1b0e2f65dc6602b"}},"d8d6807eee1e44f78e10dcd5251c3e58":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_28e68ecc22cf450d8ac9b8067ea0f801","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"","max":11,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":11,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_310d810b45114471bbc5a6658de789df"}},"21eac8d36aaf4fa2a0565329dedd965b":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_9c7bed0810f34fe5948110f9c4ad1894","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 1/11 [00:07&lt;01:14,  7.48s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_c8dc39d329de455ea39ec34406e6bd59"}},"411e0eeb49834c04bbca486de6d10b5d":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"d3282d958c424f84a1b0e2f65dc6602b":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"28e68ecc22cf450d8ac9b8067ea0f801":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"310d810b45114471bbc5a6658de789df":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"9c7bed0810f34fe5948110f9c4ad1894":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"c8dc39d329de455ea39ec34406e6bd59":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"87c240ca0fe8431096f5092d00b6d63f":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_6933623da13c468ea1caba8cb1e126f5","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_3beaa54b58654d2199153c4229b50f84","IPY_MODEL_317c46945fd64488ad7b2cc0bf30b7d2","IPY_MODEL_fa2e58ea24f84364b95fcc3f1587e9ad"]}},"6933623da13c468ea1caba8cb1e126f5":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"3beaa54b58654d2199153c4229b50f84":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_afa78843d54d404ab99cd87d55a36bb9","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_0c9761efc1a04534871cb6c05f232f4a"}},"317c46945fd64488ad7b2cc0bf30b7d2":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_21b40574fdb44788af67ece3c7b770c6","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":10,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":10,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_0a8e3b93b06740ad8cb9075dbf5d041a"}},"fa2e58ea24f84364b95fcc3f1587e9ad":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_bcaa037ccf1b47bf8d40956a370f93dc","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 10/10 [00:07&lt;00:00,  1.31it/s]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_2258f153219f48cc882eda53f0db33a2"}},"afa78843d54d404ab99cd87d55a36bb9":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"0c9761efc1a04534871cb6c05f232f4a":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"21b40574fdb44788af67ece3c7b770c6":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"0a8e3b93b06740ad8cb9075dbf5d041a":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"bcaa037ccf1b47bf8d40956a370f93dc":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"2258f153219f48cc882eda53f0db33a2":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"6574536e49dd498bae5d87ab60144c59":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_a787648e15174568a65f7b5e17ee43d2","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_9fccdb73039f4b649f5389aa2b4099ad","IPY_MODEL_701a04ff58fc492f88774aa7139dcbc2","IPY_MODEL_37296861808045328c2c4eb74625987a"]}},"a787648e15174568a65f7b5e17ee43d2":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"9fccdb73039f4b649f5389aa2b4099ad":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_8a9c221fcb4a46f9af525b5b0373a598","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_f155f2ec83cb416d8d375ae77330165b"}},"701a04ff58fc492f88774aa7139dcbc2":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_af831b75ff7146fc99abb8677c84bdf3","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":1000,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":1000,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_8fe71971bf8049bca4a5c6fde2bf7471"}},"37296861808045328c2c4eb74625987a":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_9a87fac6283543008f922a57f98b8514","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 1000/1000 [10:38&lt;00:00,  1.62it/s]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_ea8e8fab825a499c8fe628a021b731ca"}},"8a9c221fcb4a46f9af525b5b0373a598":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"f155f2ec83cb416d8d375ae77330165b":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"af831b75ff7146fc99abb8677c84bdf3":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"8fe71971bf8049bca4a5c6fde2bf7471":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"9a87fac6283543008f922a57f98b8514":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"ea8e8fab825a499c8fe628a021b731ca":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"d2a7f4cc89a44501b6423cff8f76f77d":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_9231d7a857624fbfa4c6dc1ba3307814","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_7dbee5d5ef884477be34e5a9889d13b1","IPY_MODEL_ed0108c515b94090bc7ee284284b535b","IPY_MODEL_ef228c5730d943149399ec81f4de4d46"]}},"9231d7a857624fbfa4c6dc1ba3307814":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"7dbee5d5ef884477be34e5a9889d13b1":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_0dd4ce26eb994bf89f9d429982672f1d","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_3c78bcaf65d745928c7cec9b3618f2f4"}},"ed0108c515b94090bc7ee284284b535b":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_3f9147ea50504b5e941cf600259b64c6","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":1000,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":1000,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_3cb22a9a850b42ce8d72132b34cede77"}},"ef228c5730d943149399ec81f4de4d46":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_3929bb10e2044d26869495a947ecd340","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 1000/1000 [13:01&lt;00:00,  1.28it/s]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_cf4c98d2628e418b84ef3c798588675a"}},"0dd4ce26eb994bf89f9d429982672f1d":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"3c78bcaf65d745928c7cec9b3618f2f4":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"3f9147ea50504b5e941cf600259b64c6":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"3cb22a9a850b42ce8d72132b34cede77":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"3929bb10e2044d26869495a947ecd340":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"cf4c98d2628e418b84ef3c798588675a":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"959b3e666a064b6d8917c8625ecd8fee":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_ba3a1e89d28342f683d0013065137cb2","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_53079098231c4906af72b59e6773052f","IPY_MODEL_3415712fbd16477b8eb7bf6844c56d51","IPY_MODEL_7269023608aa4b9c833466b4634a02cd"]}},"ba3a1e89d28342f683d0013065137cb2":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"53079098231c4906af72b59e6773052f":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_23ca025a590c4caba147cfee82b8fb82","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_6892149d580c4521bfbc8ce341445d61"}},"3415712fbd16477b8eb7bf6844c56d51":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_0767ce00fdc24b709c3ca2169e13e545","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":10,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":10,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_c11a3dcd03f24240826793c233950bea"}},"7269023608aa4b9c833466b4634a02cd":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_ff5a4612a01b47a8a805ca8a643956c3","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 10/10 [00:27&lt;00:00,  2.75s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_f8e2be434a1846c797556912bad038f1"}},"23ca025a590c4caba147cfee82b8fb82":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"6892149d580c4521bfbc8ce341445d61":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"0767ce00fdc24b709c3ca2169e13e545":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"c11a3dcd03f24240826793c233950bea":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"ff5a4612a01b47a8a805ca8a643956c3":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"f8e2be434a1846c797556912bad038f1":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"85a8b0c33c104b8eb22fe9e21dc3d543":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_f8a6e8d21951434b84789ff596bf188b","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_3825eaf811174f30b38725c44d1536a6","IPY_MODEL_f0fad1d578be4cb790bcbc91e2088d85","IPY_MODEL_367d61cced8a4b7a9eb3763f11856cce"]}},"f8a6e8d21951434b84789ff596bf188b":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"3825eaf811174f30b38725c44d1536a6":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_0655f854114a4705ba29eabd0cc99e31","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_f06c08cde5e14462a35c9b411e3cb167"}},"f0fad1d578be4cb790bcbc91e2088d85":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_d769f331f73d42a1ae05b69c07023890","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":10,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":10,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_52ec6649d8e94615bf9397cbd9d85b48"}},"367d61cced8a4b7a9eb3763f11856cce":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_fa0bd788f3a8421d949951e60f790c07","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 10/10 [00:26&lt;00:00,  2.71s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_2464f2977ba04e8a855461a081005305"}},"0655f854114a4705ba29eabd0cc99e31":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"f06c08cde5e14462a35c9b411e3cb167":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"d769f331f73d42a1ae05b69c07023890":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"52ec6649d8e94615bf9397cbd9d85b48":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"fa0bd788f3a8421d949951e60f790c07":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"2464f2977ba04e8a855461a081005305":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"07d74cc8c1034e8e88e0ba68681e7d83":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_20755197a75045769c76c25cf9997309","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_3aaecc6ab5334dbaaff91f3fd3d2dc92","IPY_MODEL_ccdfd084e55f4d48aa8d62ff900c54dc","IPY_MODEL_602801fccd6d4538bd22a5494c8a538e"]}},"20755197a75045769c76c25cf9997309":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"3aaecc6ab5334dbaaff91f3fd3d2dc92":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_81b219b6e4414ff888646a18f038c6d2","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_4622f6b8c7914be18ec086e119b14510"}},"ccdfd084e55f4d48aa8d62ff900c54dc":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_c603ecbfec974667805abf5f8366dde5","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":20,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":20,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_508403ada600443f97467dde6fb0dd3d"}},"602801fccd6d4538bd22a5494c8a538e":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_9840772f43134cee8eabacd2509f9fd7","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 20/20 [00:15&lt;00:00,  1.29it/s]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_4e96a7f4873f4360b2118e7a3ddbfc48"}},"81b219b6e4414ff888646a18f038c6d2":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"4622f6b8c7914be18ec086e119b14510":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"c603ecbfec974667805abf5f8366dde5":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"508403ada600443f97467dde6fb0dd3d":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"9840772f43134cee8eabacd2509f9fd7":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"4e96a7f4873f4360b2118e7a3ddbfc48":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"c19256a397ea4e79b3a0cbf5a424a3ea":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_4d1f5676bbaa4b3d8b81cc7be3cf6da3","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_eae2220f0119401a8759637bcf0e9f4c","IPY_MODEL_4aa44dd37d5243dfa24fea8f3dc6839b","IPY_MODEL_2deab0a58cd04226b834a99ee4ff4c90"]}},"4d1f5676bbaa4b3d8b81cc7be3cf6da3":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"eae2220f0119401a8759637bcf0e9f4c":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_16acd859901c4b8aa5f6b4473b74e249","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_e4edb80a747a4e9684a556ff96563e7f"}},"4aa44dd37d5243dfa24fea8f3dc6839b":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_3d74bbdd72dc4e0aadb4bd517358bbe7","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":20,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":20,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_809de5d470244a2b90daffd0477c28a4"}},"2deab0a58cd04226b834a99ee4ff4c90":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_4e13cae2ba7a44daa51794431d7af0c0","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 20/20 [00:15&lt;00:00,  1.31it/s]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_2780d1d40a0644c498e860cf4839ac25"}},"16acd859901c4b8aa5f6b4473b74e249":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"e4edb80a747a4e9684a556ff96563e7f":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"3d74bbdd72dc4e0aadb4bd517358bbe7":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"809de5d470244a2b90daffd0477c28a4":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"4e13cae2ba7a44daa51794431d7af0c0":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"2780d1d40a0644c498e860cf4839ac25":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"a2c68a56898548f8a7e190dcf2d034f4":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_e009631910e042a2a07afe8cf0e6d118","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_a850623290b14d07838e4203a29f9ba2","IPY_MODEL_9f140b090cb44361b39a6e09205c087c","IPY_MODEL_36fc4560cb99439c945c620138005a1b"]}},"e009631910e042a2a07afe8cf0e6d118":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"a850623290b14d07838e4203a29f9ba2":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_638ee1bca5d64cc7966556e3fb64c06e","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_a454e1e1460046e3a98a4b28c48ecda0"}},"9f140b090cb44361b39a6e09205c087c":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_d0a15e2054af430bad52b6fc28fd0994","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":10,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":10,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_f9295dd431fe4587a816d6c59f464399"}},"36fc4560cb99439c945c620138005a1b":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_41ca699547f14bda8a83777d78e6bc86","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 10/10 [15:36&lt;00:00, 95.65s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_c29bc574f0cd46309d8eb6ec40378025"}},"638ee1bca5d64cc7966556e3fb64c06e":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"a454e1e1460046e3a98a4b28c48ecda0":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"d0a15e2054af430bad52b6fc28fd0994":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"f9295dd431fe4587a816d6c59f464399":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"41ca699547f14bda8a83777d78e6bc86":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"c29bc574f0cd46309d8eb6ec40378025":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"858e460d943744248a4109a5c328d77c":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_700c0953693b48bea968c6c67964684d","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_c70158af000c40e59b588e745be4a2dd","IPY_MODEL_21b1b038a25f469295ea8e2b75c470ae","IPY_MODEL_c20bc14a583e452ba263c4e4f4f3147e"]}},"700c0953693b48bea968c6c67964684d":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"c70158af000c40e59b588e745be4a2dd":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_e1199401ceee4d0ca1aaa72020618241","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_57b794458691470ba34efd35f0cd8a9c"}},"21b1b038a25f469295ea8e2b75c470ae":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_4bd9a0cc11664bf4870c7fc46d151bda","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":10,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":10,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_301a50e882864065b9108e0a0cf0edde"}},"c20bc14a583e452ba263c4e4f4f3147e":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_93f8c74e75504aadbf1cdec1c5b6a43f","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 10/10 [00:14&lt;00:00,  1.43s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_a4d3c862e74242e4beeaaa880f4e3fc8"}},"e1199401ceee4d0ca1aaa72020618241":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"57b794458691470ba34efd35f0cd8a9c":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"4bd9a0cc11664bf4870c7fc46d151bda":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"301a50e882864065b9108e0a0cf0edde":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"93f8c74e75504aadbf1cdec1c5b6a43f":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"a4d3c862e74242e4beeaaa880f4e3fc8":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"79e9b193d9284e4b898cb4344db92ef3":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_8ea51857d7cf48c1beddb0a9bd3b764b","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_7b0e4960186b4307ae594b269043c16d","IPY_MODEL_aa39361015a84c158e922f8e0b8dca9a","IPY_MODEL_63acbf6a68e6455c91cbea40169d1ee9"]}},"8ea51857d7cf48c1beddb0a9bd3b764b":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"7b0e4960186b4307ae594b269043c16d":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_e2c2139f96b345768e075ef3593fac06","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_3dcbffa6176e4824b9a48cccfaf024cc"}},"aa39361015a84c158e922f8e0b8dca9a":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_740c9d9905b542d4a0edef78f1d8ca30","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":10,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":10,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_cd2dc9d9c7504cd4b118a33b7c24f683"}},"63acbf6a68e6455c91cbea40169d1ee9":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_33707c49e4fd42ad932dd74e59f4a060","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 10/10 [02:13&lt;00:00, 13.01s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_60745768aedc4942bad8883d708b0e25"}},"e2c2139f96b345768e075ef3593fac06":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"3dcbffa6176e4824b9a48cccfaf024cc":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"740c9d9905b542d4a0edef78f1d8ca30":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"cd2dc9d9c7504cd4b118a33b7c24f683":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"33707c49e4fd42ad932dd74e59f4a060":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"60745768aedc4942bad8883d708b0e25":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"48077097754d4b96b1345fb1c06560b8":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_view_name":"HBoxView","_dom_classes":[],"_model_name":"HBoxModel","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.5.0","box_style":"","layout":"IPY_MODEL_760943cea30f4236a0715b61a047fd99","_model_module":"@jupyter-widgets/controls","children":["IPY_MODEL_3bf9f4be47424b13b1db937f67894360","IPY_MODEL_28c1b766f56c41179840f2140da35b44","IPY_MODEL_487beca1cfc847c0a41a414968097b6b"]}},"760943cea30f4236a0715b61a047fd99":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"3bf9f4be47424b13b1db937f67894360":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_c28b5aecd2eb4353b99534e8eef2380c","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":"100%","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_f167c63d8ac54b07b31b261c77a98391"}},"28c1b766f56c41179840f2140da35b44":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_view_name":"ProgressView","style":"IPY_MODEL_4a8e550188ac481fa2838df0996d3b0b","_dom_classes":[],"description":"","_model_name":"FloatProgressModel","bar_style":"success","max":10,"_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":10,"_view_count":null,"_view_module_version":"1.5.0","orientation":"horizontal","min":0,"description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_50dfc1413ebd4298af9fc979de4ebb63"}},"487beca1cfc847c0a41a414968097b6b":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_view_name":"HTMLView","style":"IPY_MODEL_f36eb7ce107741159cfc86169724580f","_dom_classes":[],"description":"","_model_name":"HTMLModel","placeholder":"​","_view_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","value":" 10/10 [00:49&lt;00:00,  5.04s/it]","_view_count":null,"_view_module_version":"1.5.0","description_tooltip":null,"_model_module":"@jupyter-widgets/controls","layout":"IPY_MODEL_5b4edf526246433d8cca04eba8be3229"}},"c28b5aecd2eb4353b99534e8eef2380c":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"f167c63d8ac54b07b31b261c77a98391":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"4a8e550188ac481fa2838df0996d3b0b":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"ProgressStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","bar_color":null,"_model_module":"@jupyter-widgets/controls"}},"50dfc1413ebd4298af9fc979de4ebb63":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}},"f36eb7ce107741159cfc86169724580f":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_view_name":"StyleView","_model_name":"DescriptionStyleModel","description_width":"","_view_module":"@jupyter-widgets/base","_model_module_version":"1.5.0","_view_count":null,"_view_module_version":"1.2.0","_model_module":"@jupyter-widgets/controls"}},"5b4edf526246433d8cca04eba8be3229":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_view_name":"LayoutView","grid_template_rows":null,"right":null,"justify_content":null,"_view_module":"@jupyter-widgets/base","overflow":null,"_model_module_version":"1.2.0","_view_count":null,"flex_flow":null,"width":null,"min_width":null,"border":null,"align_items":null,"bottom":null,"_model_module":"@jupyter-widgets/base","top":null,"grid_column":null,"overflow_y":null,"overflow_x":null,"grid_auto_flow":null,"grid_area":null,"grid_template_columns":null,"flex":null,"_model_name":"LayoutModel","justify_items":null,"grid_row":null,"max_height":null,"align_content":null,"visibility":null,"align_self":null,"height":null,"min_height":null,"padding":null,"grid_auto_rows":null,"grid_gap":null,"max_width":null,"order":null,"_view_module_version":"1.2.0","grid_template_areas":null,"object_position":null,"object_fit":null,"grid_auto_columns":null,"margin":null,"display":null,"left":null}}}}},"cells":[{"cell_type":"markdown","source":["# Sequential Batch Learning in Stochastic MAB and Contextual MAB on Mushroom and Synthetic data"],"metadata":{"id":"i_1jvK-oGLjU"}},{"cell_type":"markdown","source":["## Executive summary\n","\n","| | |\n","| --- | --- |\n","| Problem | Learning user preferences online might have an impact of delay and training recommender system sequentially for every example is computationally heavy. |\n","| Hypothesis | A learning agent observes responses batched in groups over a certain time period. The impact of batch learning can be measured in terms of online behavior. |\n","| Prblm Stmt. | Given a finite set of arms ⁍, an environment ⁍ (⁍ is the distribution of rewards for action ⁍), and a time horizon ⁍, at each time step ⁍, the agent chooses an action ⁍ and receives a reward ⁍. The goal of the agent is to maximize the total reward ⁍. |\n","| Solution | Sequential batch learning is a more generalized way of learning which covers both offline and online settings as special cases bringing together their advantages. Unlike offline learning, sequential batch learning retains the sequential nature of the problem. Unlike online learning, it is often appealing to implement batch learning in large scale bandit problems. In this setting, responses are grouped in batches and observed by the agent only at the end of each batch. |\n","| Dataset | Mushroom, Synthetic |\n","| Preprocessing | Train/test split, label encoding |\n","| Metrics | Conversion rate, regret |\n","| Credits | [Danil Provodin](https://github.com/danilprov) |"],"metadata":{"id":"bLatuDpMGLge"}},{"cell_type":"markdown","source":["### Environments\n","\n","| Name | Type | Rewards |\n","| --- | --- | --- |\n","| env1 | 2-arm environment | [0.7, 0.5] |\n","| env2 | 2-arm environment | [0.7, 0.4] |\n","| env3 | 2-arm environment | [0.7, 0.1] |\n","| env4 | 4-arm environment | [0.35, 0.18, 0.47, 0.61] |\n","| env5 | 4-arm environment | [0.40, 0.75, 0.57, 0.49] |\n","| env6 | 4-arm environment | [0.70, 0.50, 0.30, 0.10] |"],"metadata":{"id":"y-pKQhr9GaAv"}},{"cell_type":"markdown","source":["### Simulation\n","\n","| Application | Policy |\n","| --- | --- |\n","| Multi-armed bandit (MAB) | Thompson Sampling (TS) |\n","| Multi-armed bandit (MAB) | Upper Confidence Bound (UCB) |\n","| Contextual MAB (CMAB) | Linear Thompson Sampling (LinTS) |\n","| Contextual MAB (CMAB) | Linear UCB (LinUCB) |"],"metadata":{"id":"QPyNX42KGiaR"}},{"cell_type":"markdown","source":["## Process flow"],"metadata":{"id":"4veQdUjFGCwk"}},{"cell_type":"markdown","source":["![](https://github.com/RecoHut-Stanzas/S873634/raw/main/images/process_flow.svg)"],"metadata":{"id":"Pm8qSJsaGEUN"}},{"cell_type":"markdown","source":["## Setup"],"metadata":{"id":"VAwEfoPurXGo"}},{"cell_type":"markdown","source":["### Imports"],"metadata":{"id":"ngwffbD4rXE3"}},{"cell_type":"code","source":["import pandas as pd\n","import numpy as np\n","from numpy.linalg import inv\n","from sklearn.metrics import roc_auc_score\n","from sklearn.model_selection import GridSearchCV, train_test_split, cross_validate\n","from sklearn.tree import DecisionTreeClassifier\n","from scipy.optimize import minimize\n","from lightgbm import LGBMClassifier\n","from scipy.stats import beta\n","import pickle\n","import os\n","import shutil\n","\n","import tqdm\n","from tqdm.notebook import tqdm\n","from multiprocessing.dummy import Pool\n","from IPython.display import clear_output\n","import matplotlib.pyplot as plt\n","\n","from torch.utils.data import Dataset, DataLoader\n","\n","from __future__ import print_function\n","from abc import ABCMeta, abstractmethod"],"metadata":{"id":"vSHj786VrXC8"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["## Data"],"metadata":{"id":"XXMnN2vioutk"}},{"cell_type":"markdown","source":["### Download"],"metadata":{"id":"W_L8uOd0rfK6"}},{"cell_type":"markdown","source":["This data set includes descriptions of hypothetical samples corresponding to 23 species of gilled mushrooms in the Agaricus and Lepiota Family (pp. 500-525). Each species is identified as definitely edible, definitely poisonous, or of unknown edibility and not recommended. This latter class was combined with the poisonous one. The Guide clearly states that there is no simple rule for determining the edibility of a mushroom; no rule like ``leaflets three, let it be'' for Poisonous Oak and Ivy. More details [here](https://archive.ics.uci.edu/ml/datasets/mushroom)."],"metadata":{"id":"-f9daEHdrG5E"}},{"cell_type":"code","source":["!mkdir -p data\n","!cd data && wget -q --show-progress https://archive.ics.uci.edu/ml/machine-learning-databases/mushroom/agaricus-lepiota.data\n","!cd data && wget -q --show-progress https://archive.ics.uci.edu/ml/machine-learning-databases/mushroom/agaricus-lepiota.names"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"adGWyfxFrHij","executionInfo":{"status":"ok","timestamp":1639145418029,"user_tz":-330,"elapsed":1715,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"eb9c650f-63bc-4fcf-ef24-b1f698d71d4a"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["agaricus-lepiota.da 100%[===================>] 364.95K  1.31MB/s    in 0.3s    \n","agaricus-lepiota.na 100%[===================>]   6.66K  --.-KB/s    in 0s      \n"]}]},{"cell_type":"markdown","source":["### Preprocessing"],"metadata":{"id":"Aal7wGRfrhfe"}},{"cell_type":"code","source":["mushroom_data = pd.read_csv(\"data/agaricus-lepiota.data\", header=None)\n","\n","column_names = [\"classes\", \"cap-shape\", \"cap-surface\", \"cap-color\", \"bruises?\", \"odor\", \"gill-attachment\",\n","                \"gill-spacing\", \"gill-size\", \"gill-color\", \"stalk-shape\", \"stalk-root\", \"stalk-surface-above-ring\",\n","                \"stalk-surface-below-ring\", \"stalk-color-above-ring\", \"stalk-color-below-ring\", \"veil-type\", \n","                \"veil-color\", \"ring-number\", \"ring-type\", \"spore-print-color\", \"population\", \"habitat\"]\n","                \n","mushroom_data.columns = column_names\n","mushroom_data.head()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":278},"id":"TshNYFbQrhdi","executionInfo":{"status":"ok","timestamp":1639145489020,"user_tz":-330,"elapsed":516,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"88c714b1-d1c5-4666-8f70-9db8b6fe6584"},"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>classes</th>\n","      <th>cap-shape</th>\n","      <th>cap-surface</th>\n","      <th>cap-color</th>\n","      <th>bruises?</th>\n","      <th>odor</th>\n","      <th>gill-attachment</th>\n","      <th>gill-spacing</th>\n","      <th>gill-size</th>\n","      <th>gill-color</th>\n","      <th>stalk-shape</th>\n","      <th>stalk-root</th>\n","      <th>stalk-surface-above-ring</th>\n","      <th>stalk-surface-below-ring</th>\n","      <th>stalk-color-above-ring</th>\n","      <th>stalk-color-below-ring</th>\n","      <th>veil-type</th>\n","      <th>veil-color</th>\n","      <th>ring-number</th>\n","      <th>ring-type</th>\n","      <th>spore-print-color</th>\n","      <th>population</th>\n","      <th>habitat</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0</th>\n","      <td>p</td>\n","      <td>x</td>\n","      <td>s</td>\n","      <td>n</td>\n","      <td>t</td>\n","      <td>p</td>\n","      <td>f</td>\n","      <td>c</td>\n","      <td>n</td>\n","      <td>k</td>\n","      <td>e</td>\n","      <td>e</td>\n","      <td>s</td>\n","      <td>s</td>\n","      <td>w</td>\n","      <td>w</td>\n","      <td>p</td>\n","      <td>w</td>\n","      <td>o</td>\n","      <td>p</td>\n","      <td>k</td>\n","      <td>s</td>\n","      <td>u</td>\n","    </tr>\n","    <tr>\n","      <th>1</th>\n","      <td>e</td>\n","      <td>x</td>\n","      <td>s</td>\n","      <td>y</td>\n","      <td>t</td>\n","      <td>a</td>\n","      <td>f</td>\n","      <td>c</td>\n","      <td>b</td>\n","      <td>k</td>\n","      <td>e</td>\n","      <td>c</td>\n","      <td>s</td>\n","      <td>s</td>\n","      <td>w</td>\n","      <td>w</td>\n","      <td>p</td>\n","      <td>w</td>\n","      <td>o</td>\n","      <td>p</td>\n","      <td>n</td>\n","      <td>n</td>\n","      <td>g</td>\n","    </tr>\n","    <tr>\n","      <th>2</th>\n","      <td>e</td>\n","      <td>b</td>\n","      <td>s</td>\n","      <td>w</td>\n","      <td>t</td>\n","      <td>l</td>\n","      <td>f</td>\n","      <td>c</td>\n","      <td>b</td>\n","      <td>n</td>\n","      <td>e</td>\n","      <td>c</td>\n","      <td>s</td>\n","      <td>s</td>\n","      <td>w</td>\n","      <td>w</td>\n","      <td>p</td>\n","      <td>w</td>\n","      <td>o</td>\n","      <td>p</td>\n","      <td>n</td>\n","      <td>n</td>\n","      <td>m</td>\n","    </tr>\n","    <tr>\n","      <th>3</th>\n","      <td>p</td>\n","      <td>x</td>\n","      <td>y</td>\n","      <td>w</td>\n","      <td>t</td>\n","      <td>p</td>\n","      <td>f</td>\n","      <td>c</td>\n","      <td>n</td>\n","      <td>n</td>\n","      <td>e</td>\n","      <td>e</td>\n","      <td>s</td>\n","      <td>s</td>\n","      <td>w</td>\n","      <td>w</td>\n","      <td>p</td>\n","      <td>w</td>\n","      <td>o</td>\n","      <td>p</td>\n","      <td>k</td>\n","      <td>s</td>\n","      <td>u</td>\n","    </tr>\n","    <tr>\n","      <th>4</th>\n","      <td>e</td>\n","      <td>x</td>\n","      <td>s</td>\n","      <td>g</td>\n","      <td>f</td>\n","      <td>n</td>\n","      <td>f</td>\n","      <td>w</td>\n","      <td>b</td>\n","      <td>k</td>\n","      <td>t</td>\n","      <td>e</td>\n","      <td>s</td>\n","      <td>s</td>\n","      <td>w</td>\n","      <td>w</td>\n","      <td>p</td>\n","      <td>w</td>\n","      <td>o</td>\n","      <td>e</td>\n","      <td>n</td>\n","      <td>a</td>\n","      <td>g</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["  classes cap-shape cap-surface  ... spore-print-color population habitat\n","0       p         x           s  ...                 k          s       u\n","1       e         x           s  ...                 n          n       g\n","2       e         b           s  ...                 n          n       m\n","3       p         x           y  ...                 k          s       u\n","4       e         x           s  ...                 n          a       g\n","\n","[5 rows x 23 columns]"]},"metadata":{},"execution_count":3}]},{"cell_type":"code","source":["mushroom_data.dtypes"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"sIbAiLjtrhbZ","executionInfo":{"status":"ok","timestamp":1639145504479,"user_tz":-330,"elapsed":518,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"f0602325-8225-4047-b8a9-503031bba8c4"},"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["classes                     object\n","cap-shape                   object\n","cap-surface                 object\n","cap-color                   object\n","bruises?                    object\n","odor                        object\n","gill-attachment             object\n","gill-spacing                object\n","gill-size                   object\n","gill-color                  object\n","stalk-shape                 object\n","stalk-root                  object\n","stalk-surface-above-ring    object\n","stalk-surface-below-ring    object\n","stalk-color-above-ring      object\n","stalk-color-below-ring      object\n","veil-type                   object\n","veil-color                  object\n","ring-number                 object\n","ring-type                   object\n","spore-print-color           object\n","population                  object\n","habitat                     object\n","dtype: object"]},"metadata":{},"execution_count":4}]},{"cell_type":"code","source":["# label encoding\n","for column in column_names:\n","    mushroom_data[column] = mushroom_data[column].astype('category')\n","    mushroom_data[column] = mushroom_data[column].cat.codes\n","\n","# split\n","idx_trn, idx_tst = train_test_split(mushroom_data.index, test_size=0.2, random_state=42, \n","                                    stratify=mushroom_data[['classes']])"],"metadata":{"id":"I420evWNrhZQ"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["gini by factors"],"metadata":{"id":"c6AARTV0sSEw"}},{"cell_type":"code","source":["def gini(var):\n","    df = mushroom_data.copy()\n","    x_trn = df.loc[idx_trn, var]\n","    y_trn = df.loc[idx_trn, 'classes']\n","    x_tst = df.loc[idx_tst, var]\n","    y_tst = df.loc[idx_tst, 'classes']\n","    \n","    if x_trn.dtype in ['O','object']:\n","        cats = pd.DataFrame({'x': x_trn, 'y': y_trn}).fillna('#NAN#').groupby('x').agg('mean').sort_values('y').index.values\n","        X_trn = pd.Categorical(x_trn.fillna('#NAN#'), categories=cats, ordered=True).codes.reshape(-1, 1)\n","        X_tst = pd.Categorical(x_tst.fillna('#NAN#'), categories=cats, ordered=True).codes.reshape(-1, 1)\n","    else:\n","        repl = min(x_trn.min(), x_tst.min())-1 if np.isfinite(min(x_trn.min(), x_tst.min())-1) else -999999\n","        #repl = x_trn.min()-1 if np.isfinite(x_trn.min())-1 else -999999\n","        X_trn = x_trn.fillna(repl).replace(np.inf, repl).replace(-np.inf, repl).values.reshape(-1, 1)\n","        X_tst = x_tst.fillna(repl).replace(np.inf, repl).replace(-np.inf, repl).values.reshape(-1, 1)\n","    \n","    obvious_gini_trn = 2*roc_auc_score(y_trn, X_trn)-1\n","    obvious_gini_tst = 2*roc_auc_score(y_tst, X_tst)-1\n","\n","    if obvious_gini_trn < 0:\n","        obvious_gini_trn = -obvious_gini_trn\n","        obvious_gini_tst = -obvious_gini_tst\n","\n","    parameters = {'min_samples_leaf':[0.01, 0.025, 0.05, 0.1]}\n","    dt = DecisionTreeClassifier(random_state=1)\n","    clf = GridSearchCV(dt, parameters, cv=4, scoring='roc_auc', n_jobs=10)\n","    clf.fit(X_trn, y_trn)\n","\n","    true_gini_trn = 2*clf.best_score_-1\n","    true_gini_tst = 2*roc_auc_score(y_tst, clf.predict_proba(X_tst)[:, 1])-1\n","\n","    if true_gini_trn < 0:\n","        true_gini_trn = -true_gini_trn\n","        true_gini_tst = -true_gini_tst\n","\n","    if obvious_gini_trn > true_gini_trn:\n","        return [var, obvious_gini_trn, obvious_gini_tst]\n","    else:\n","        return [var, true_gini_trn, true_gini_tst]"],"metadata":{"id":"hjgbgx7NrhWv"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["with Pool(20) as p:\n","    vars_gini = list(tqdm(p.imap(gini, column_names), total=len(column_names)))"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":49,"referenced_widgets":["2b72cbdd08374942859625d2047795f2","396e4164d325484e9956f597c2e133e0","46ce32b9cbc74045b96fc2fbc93f0575","596798eaeac14aaea8c3cc476b7f329f","c60ea6ae22a2457786ca6b8e3446e3f2","342e5fcfeeb04720834331a4274b5e28","9cbda8d73a0a41f8acb9c3bdacdc7a9b","042addbfa66d4dd4a8543124cc312905","3e9ddf2ca50f4c24bcf276eeeee63d01","ce6fa32eee4044df843b45c5693e065e","5ac6a207b502473fa8bd11d2adcc1201"]},"id":"pwtreallsHVN","executionInfo":{"status":"ok","timestamp":1639145595544,"user_tz":-330,"elapsed":9639,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"d6e113bd-e273-4922-f04d-a4d6448d26b2"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"2b72cbdd08374942859625d2047795f2","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/23 [00:00<?, ?it/s]"]},"metadata":{}}]},{"cell_type":"code","source":["vars_gini = pd.DataFrame(vars_gini)\n","vars_gini.set_index(0, inplace=True)\n","vars_gini.columns = ['gini_train', 'gini_test']\n","\n","vars_gini.T"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":184},"id":"wqBNXxJVsLoh","executionInfo":{"status":"ok","timestamp":1639145626752,"user_tz":-330,"elapsed":527,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"98514987-c253-42d6-8c63-a7fb38cf9c5a"},"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>classes</th>\n","      <th>cap-shape</th>\n","      <th>cap-surface</th>\n","      <th>cap-color</th>\n","      <th>bruises?</th>\n","      <th>odor</th>\n","      <th>gill-attachment</th>\n","      <th>gill-spacing</th>\n","      <th>gill-size</th>\n","      <th>gill-color</th>\n","      <th>stalk-shape</th>\n","      <th>stalk-root</th>\n","      <th>stalk-surface-above-ring</th>\n","      <th>stalk-surface-below-ring</th>\n","      <th>stalk-color-above-ring</th>\n","      <th>stalk-color-below-ring</th>\n","      <th>veil-type</th>\n","      <th>veil-color</th>\n","      <th>ring-number</th>\n","      <th>ring-type</th>\n","      <th>spore-print-color</th>\n","      <th>population</th>\n","      <th>habitat</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>gini_train</th>\n","      <td>1.0</td>\n","      <td>0.185225</td>\n","      <td>0.180956</td>\n","      <td>0.207793</td>\n","      <td>0.495815</td>\n","      <td>0.967487</td>\n","      <td>0.040117</td>\n","      <td>0.256251</td>\n","      <td>0.490563</td>\n","      <td>0.758349</td>\n","      <td>0.10296</td>\n","      <td>0.429898</td>\n","      <td>0.543743</td>\n","      <td>0.534091</td>\n","      <td>0.544061</td>\n","      <td>0.534553</td>\n","      <td>0.0</td>\n","      <td>0.045791</td>\n","      <td>0.114617</td>\n","      <td>0.627000</td>\n","      <td>0.761604</td>\n","      <td>0.519078</td>\n","      <td>0.444780</td>\n","    </tr>\n","    <tr>\n","      <th>gini_test</th>\n","      <td>1.0</td>\n","      <td>0.212779</td>\n","      <td>0.201118</td>\n","      <td>0.237278</td>\n","      <td>0.490001</td>\n","      <td>0.969555</td>\n","      <td>0.044683</td>\n","      <td>0.257858</td>\n","      <td>0.535203</td>\n","      <td>0.760057</td>\n","      <td>0.09395</td>\n","      <td>0.426011</td>\n","      <td>0.530422</td>\n","      <td>0.532951</td>\n","      <td>0.528023</td>\n","      <td>0.516607</td>\n","      <td>0.0</td>\n","      <td>0.054705</td>\n","      <td>0.117193</td>\n","      <td>0.622052</td>\n","      <td>0.746050</td>\n","      <td>0.510436</td>\n","      <td>0.476353</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["0           classes  cap-shape  ...  population   habitat\n","gini_train      1.0   0.185225  ...    0.519078  0.444780\n","gini_test       1.0   0.212779  ...    0.510436  0.476353\n","\n","[2 rows x 23 columns]"]},"metadata":{},"execution_count":8}]},{"cell_type":"markdown","source":["Correlation analysis"],"metadata":{"id":"2CWLsNxLsVYI"}},{"cell_type":"code","source":["vars_corrs = mushroom_data.loc[:, column_names].corr().abs().stack().reset_index().drop_duplicates()\n","vars_corrs = vars_corrs[vars_corrs.level_0!=vars_corrs.level_1]\n","vars_corrs.columns = ['var_1', 'var_2', 'correlation']\n","vars_corrs = vars_corrs.set_index(['var_1', 'var_2'], drop=True).sort_values(by='correlation', ascending=False)\n","\n","vars_drop = []\n","\n","for v in vars_corrs[vars_corrs.correlation > 0.7].index.values:\n","    if v[0] not in vars_drop and v[1] not in vars_drop:\n","        vars_drop.append(v[1] if vars_gini.loc[v[0], 'gini_train'] > vars_gini.loc[v[1], 'gini_train'] else v[0])\n","        \n","del v"],"metadata":{"id":"U94p4ETPscxi"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["Feature selection"],"metadata":{"id":"q1MuoCp4sk8r"}},{"cell_type":"code","source":["# all variables\n","vars0 = column_names[1:]\n","\n","# drop values with gini less than 3%\n","vars1 = [v for v in vars0 if vars_gini.loc[v, 'gini_train'] >= 0.03]\n","\n","# drop correlated variables\n","vars2 = [v for v in vars1 if v not in vars_drop]\n","\n","i = 0\n","\n","for var_lst in [vars0, vars1, vars2]:\n","    i += 1\n","    lgb = LGBMClassifier(max_depth=1, n_estimators=250, random_state=42, n_jobs=30)\n","    \n","    cv = cross_validate(lgb, mushroom_data.loc[:, var_lst], mushroom_data.loc[:, 'classes'], \n","                        cv=5, scoring='roc_auc', n_jobs=20, return_train_score=True)\n","    \n","    lgb.fit(mushroom_data[var_lst], mushroom_data['classes'])\n","    \n","    print({'Variables': len(var_lst), \n","           'Train CV': round(cv['train_score'].mean()*2-1, 4), \n","           'Test CV': round(cv['test_score'].mean()*2-1, 4)})\n","    \n","var_lst_imp = pd.Series(dict(zip(var_lst, lgb.feature_importances_)))\n","var_lst = [i for i in var_lst_imp.index if var_lst_imp.loc[i]>0]\n","print({'exclude': [i for i in var_lst_imp.index if var_lst_imp.loc[i]<=0]})\n","print(len(var_lst))\n","\n","forw_cols = []\n","current_ginis = pd.Series({'Train CV':0, 'Test CV':0})\n","\n","def forw(x):\n","    lgb = LGBMClassifier(max_depth=1, n_estimators=250, random_state=42, n_jobs=1)\n","    cv = cross_validate(lgb, mushroom_data.loc[:, forw_cols+[x]], mushroom_data.loc[:, 'classes'],\n","                        cv=5, scoring='roc_auc', n_jobs=1, return_train_score=True)\n","    lgb.fit(mushroom_data.loc[:, forw_cols+[x]], mushroom_data.loc[:, 'classes'])\n","    return x, pd.Series({\n","        'Train CV': cv['train_score'].mean()*2-1,\n","        'Test CV': cv['test_score'].mean()*2-1\n","    })\n","\n","forwards_log = []\n","while len(forw_cols)<30:\n","    with Pool(20) as p:\n","        res = list(tqdm(p.imap(forw, [i for i in var_lst if i not in forw_cols]), total=len(var_lst)-len(forw_cols), leave=False))\n","    res = pd.DataFrame({i[0]:i[1] for i in res}).T\n","    delta = res - current_ginis\n","    if delta['Test CV'].max()<0:\n","        break\n","    best_var = delta['Test CV'].idxmax()\n","    forw_cols = forw_cols + [best_var]\n","    current_ginis = res.loc[best_var]\n","    forwards_log.append(current_ginis)\n","    clear_output()\n","    print(pd.DataFrame(forwards_log))\n","\n","clear_output()\n","forwards_log = pd.DataFrame(forwards_log)\n","forwards_log['Uplift Train CV'] = forwards_log['Train CV']-forwards_log['Train CV'].shift(1).fillna(0)\n","forwards_log['Uplift Test CV'] = forwards_log['Test CV']-forwards_log['Test CV'].shift(1).fillna(0)\n","print(forwards_log)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":153,"referenced_widgets":["3d6f8baea53a484faab78b4c50fbf620","9689c9d0d5354f7782671e9b301c5840","4550c6895fa244efb6dd5cbc9dd98324","b70c31ffb8784655ad0d4e59e12343d6","c14978517f8047718a1c084947d702b0","60d54986769b4e859de45bbce659b7f1","0181512ceea0401db61675636ae223b5","4b34f4a343b242abaef58ebe57e07a4d","476d47eec56d4ff0bec2900ed3505ba2","27609bb3e0594b5084e01fae4b04cc9c","606916b9b63b43969a01e84ec142f341","b8cc488f50764d7a80c565e120acf3bd","73139f73ef7d4769807b6dc081a48fb1","8c0a7a56a8d946bf953390fd61675c4d","63ead4d72b27495cac3bb6a385895123","a1d8d9e611534997828fa51cdab2bd0a","afe8ad0a833b4f399a1f6d8367786fd3","c55e9f4461c24c9f88ba2098ff957634","4472bedf87ec41aeb3306d6364386540","edc2a60460ef48bdb78972f403c14175","d971324520f34ce5b3b6994ddd0d0102","6d4282c72b80490cbdb027d3592849ea","77454d4570db40ad9d90ed0f9d80f019","efe993ec7ad34cfaac8ca10da8b06b25","a845d33bdbc048af807e19eb89699a1b","e16dcd3650f64c61b1c53445f96d13d5","124e06ade51442c3af9d364455ec278f","a87b20c55b6547d7bad171519b9c2029","ce8d826cba9f41f8bb33997ceceab3c0","3ad0f66ba72c47ce8537870c6af56c0c","3f68a6b94c614e87aea69a1fecf11d2e","3b3af273314945ef951076ac38deb74c","ffd1831574034bc59136368ad3e60108","d8e841629bac43fbbea7f0a924aa9b94","dcde233fe4804bd8be4a35f7a781a985","a31bcdbac3f54376814d4b9cd40d545b","806b8ca3d7fe4e05b22b038a82eb4545","a26e8dfe75d540208dae3d35a0444551","9a8c2b7f24ad47dea0cc84b99639f151","721da9071ffc493d97a523d2f2ac4f59","0f5ac0bc528b4fa1b8d5a7aafeb36c79","a37b2e9543074546bf43c7741f3a5650","f95e707810f3426eb20134bad69e6a7f","c065af81128d4149b26e71b7a6cb4c14","153549b4beb84e54b5fe4bec2eecf075","d7e2ff7ce49d44588f16e7abfbb9885c","6a821f0913784d64b2e98282b6ae15a1","f736e0b03dea46dea371a5abcbcde9d7","f0f773f30026425fa0dfdcefd1f7d2dc","914c417613a446c28921a6cb430218dc","76e6aa6b457347af8ef3146d3d1e7437","3d87aae69d6149fc98c0b0a642fd9194","4a588c2572094039983c106ffd558d86","346efff4305041a5bd4913a00ebae73a","3a36af3ccfd54942b5e4643c373f60a2","d1538f6f3b7641ccb48e2bbdef414389","3810f04feb094355bdc304447a3356b5","2528ae76b0a7423a8910adf401a5514a","d8d6807eee1e44f78e10dcd5251c3e58","21eac8d36aaf4fa2a0565329dedd965b","411e0eeb49834c04bbca486de6d10b5d","d3282d958c424f84a1b0e2f65dc6602b","28e68ecc22cf450d8ac9b8067ea0f801","310d810b45114471bbc5a6658de789df","9c7bed0810f34fe5948110f9c4ad1894","c8dc39d329de455ea39ec34406e6bd59"]},"id":"WzyBNZ9psmGi","executionInfo":{"status":"ok","timestamp":1639145844301,"user_tz":-330,"elapsed":68523,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"c12cae13-09ca-4583-fa7d-1cb2fe20519f"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["              Train CV   Test CV  Uplift Train CV  Uplift Test CV\n","odor          0.970307  0.843798         0.970307        0.843798\n","gill-size     0.979971  0.958396         0.009664        0.114598\n","gill-spacing  0.984611  0.964531         0.004640        0.006135\n","stalk-shape   0.993921  0.982878         0.009310        0.018347\n","habitat       0.995208  0.984953         0.001287        0.002075\n"]}]},{"cell_type":"code","source":["ids_vars = forwards_log[forwards_log['Uplift Test CV']>0.001].index.values.tolist()\n","vars_gini.loc[ids_vars,:]"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":238},"id":"jjPfCIG-sw2k","executionInfo":{"status":"ok","timestamp":1639145844304,"user_tz":-330,"elapsed":47,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"9fa96409-314f-4dab-a148-c38601d0e10b"},"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>gini_train</th>\n","      <th>gini_test</th>\n","    </tr>\n","    <tr>\n","      <th>0</th>\n","      <th></th>\n","      <th></th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>odor</th>\n","      <td>0.967487</td>\n","      <td>0.969555</td>\n","    </tr>\n","    <tr>\n","      <th>gill-size</th>\n","      <td>0.490563</td>\n","      <td>0.535203</td>\n","    </tr>\n","    <tr>\n","      <th>gill-spacing</th>\n","      <td>0.256251</td>\n","      <td>0.257858</td>\n","    </tr>\n","    <tr>\n","      <th>stalk-shape</th>\n","      <td>0.102960</td>\n","      <td>0.093950</td>\n","    </tr>\n","    <tr>\n","      <th>habitat</th>\n","      <td>0.444780</td>\n","      <td>0.476353</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["              gini_train  gini_test\n","0                                  \n","odor            0.967487   0.969555\n","gill-size       0.490563   0.535203\n","gill-spacing    0.256251   0.257858\n","stalk-shape     0.102960   0.093950\n","habitat         0.444780   0.476353"]},"metadata":{},"execution_count":12}]},{"cell_type":"code","source":["mushroom_data_features = mushroom_data[ids_vars + [\"classes\"]]\n","mushroom_data = mushroom_data_features.loc[mushroom_data_features.index.repeat(4)].reset_index(drop=True)\n","\n","mushroom_data[\"a\"] = np.random.choice([0, 1], mushroom_data.shape[0])\n","mushroom_data[\"probs\"] = 1\n","mushroom_data[\"y\"] = 0\n","\n","eat_edible = (1-mushroom_data[\"classes\"]) * mushroom_data[\"a\"] * 1\n","eat_poisonous = mushroom_data[\"classes\"] * mushroom_data[\"a\"] * np.random.choice([1, -1], mushroom_data.shape[0])\n","mushroom_data[\"y\"] = eat_edible + eat_poisonous\n","new_names = ['X_' + str(i+1) for i in range(len(ids_vars))]    \n","mushroom_data = mushroom_data.rename(columns=dict(zip(ids_vars, new_names)))\n","\n","mushroom_data_final = mushroom_data[new_names + ['a', 'y', 'probs']]\n","mushroom_data_final.head()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":206},"id":"vEXcrpMis7Pi","executionInfo":{"status":"ok","timestamp":1639145844306,"user_tz":-330,"elapsed":39,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"bcb3014b-6da3-4627-e4a8-9e224a253f2a"},"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>X_1</th>\n","      <th>X_2</th>\n","      <th>X_3</th>\n","      <th>X_4</th>\n","      <th>X_5</th>\n","      <th>a</th>\n","      <th>y</th>\n","      <th>probs</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0</th>\n","      <td>6</td>\n","      <td>1</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>5</td>\n","      <td>1</td>\n","      <td>1</td>\n","      <td>1</td>\n","    </tr>\n","    <tr>\n","      <th>1</th>\n","      <td>6</td>\n","      <td>1</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>5</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>1</td>\n","    </tr>\n","    <tr>\n","      <th>2</th>\n","      <td>6</td>\n","      <td>1</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>5</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>1</td>\n","    </tr>\n","    <tr>\n","      <th>3</th>\n","      <td>6</td>\n","      <td>1</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>5</td>\n","      <td>1</td>\n","      <td>-1</td>\n","      <td>1</td>\n","    </tr>\n","    <tr>\n","      <th>4</th>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>1</td>\n","      <td>1</td>\n","      <td>1</td>\n","      <td>1</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["   X_1  X_2  X_3  X_4  X_5  a  y  probs\n","0    6    1    0    0    5  1  1      1\n","1    6    1    0    0    5  0  0      1\n","2    6    1    0    0    5  0  0      1\n","3    6    1    0    0    5  1 -1      1\n","4    0    0    0    0    1  1  1      1"]},"metadata":{},"execution_count":13}]},{"cell_type":"code","source":["with open('data/mushroom_data_final.pickle', 'wb') as handle:\n","    pickle.dump(mushroom_data_final, handle, protocol=pickle.HIGHEST_PROTOCOL)"],"metadata":{"id":"ZTavtFcks9iR"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["## Utilities"],"metadata":{"id":"p4B8_mndtBlr"}},{"cell_type":"markdown","source":["### Softmax"],"metadata":{"id":"dHg46NSstX9e"}},{"cell_type":"code","source":["def softmax(action_values, tau=1.0):\n","    \"\"\"\n","    Args:\n","        action_values (Numpy array): A 2D array of shape (batch_size, num_actions).\n","                       The action-values computed by an action-value network.\n","        tau (float): The temperature parameter scalar.\n","    Returns:\n","        A 2D array of shape (batch_size, num_actions). Where each column is a probability distribution over\n","        the actions representing the policy.\n","    \"\"\"\n","\n","    # Compute the preferences by dividing the action-values by the temperature parameter tau\n","    preferences = action_values / tau\n","    # Compute the maximum preference across the actions\n","    max_preference = np.max(preferences, axis=1)\n","\n","    # your code here\n","\n","    # Reshape max_preference array which has shape [Batch,] to [Batch, 1]. This allows NumPy broadcasting\n","    # when subtracting the maximum preference from the preference of each action.\n","    reshaped_max_preference = max_preference.reshape((-1, 1))\n","    # print(reshaped_max_preference)\n","\n","    # Compute the numerator, i.e., the exponential of the preference - the max preference.\n","    exp_preferences = np.exp(preferences - reshaped_max_preference)\n","    # print(exp_preferences)\n","    # Compute the denominator, i.e., the sum over the numerator along the actions axis.\n","    sum_of_exp_preferences = np.sum(exp_preferences, axis=1)\n","    # print(sum_of_exp_preferences)\n","\n","    # your code here\n","\n","    # Reshape sum_of_exp_preferences array which has shape [Batch,] to [Batch, 1] to  allow for NumPy broadcasting\n","    # when dividing the numerator by the denominator.\n","    reshaped_sum_of_exp_preferences = sum_of_exp_preferences.reshape((-1, 1))\n","    # print(reshaped_sum_of_exp_preferences)\n","\n","    # Compute the action probabilities according to the equation in the previous cell.\n","    action_probs = exp_preferences / reshaped_sum_of_exp_preferences\n","    # print(action_probs)\n","\n","    # your code here\n","\n","    # squeeze() removes any singleton dimensions. It is used here because this function is used in the\n","    # agent policy when selecting an action (for which the batch dimension is 1.) As np.random.choice is used in\n","    # the agent policy and it expects 1D arrays, we need to remove this singleton batch dimension.\n","    action_probs = action_probs.squeeze()\n","    return action_probs"],"metadata":{"id":"utuZpkQktdjN"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# if __name__ == '__main__':\n","#     rand_generator = np.random.RandomState(0)\n","#     action_values = rand_generator.normal(0, 1, (2, 4))\n","#     tau = 0.5\n","\n","#     action_probs = softmax(action_values, tau)\n","#     print(\"action_probs\", action_probs)\n","\n","#     assert (np.allclose(action_probs, np.array([\n","#         [0.25849645, 0.01689625, 0.05374514, 0.67086216],\n","#         [0.84699852, 0.00286345, 0.13520063, 0.01493741]\n","#     ])))\n","\n","#     action_values = np.array([[0.0327, 0.0127, 0.0688]])\n","#     tau = 1.\n","#     action_probs = softmax(action_values, tau)\n","#     print(\"action_probs\", action_probs)\n","\n","#     assert np.allclose(action_probs, np.array([0.3315, 0.3249, 0.3436]), atol=1e-04)\n","\n","#     print(\"Passed the asserts! (Note: These are however limited in scope, additional testing is encouraged.)\")"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"Mbk-YyyHtfqc","executionInfo":{"status":"ok","timestamp":1639145939697,"user_tz":-330,"elapsed":477,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"c0665afb-a0c1-460c-e4d7-7d9e20103146"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["action_probs [[0.25849645 0.01689625 0.05374514 0.67086216]\n"," [0.84699852 0.00286345 0.13520063 0.01493741]]\n","action_probs [0.33145968 0.32489634 0.34364398]\n","Passed the asserts! (Note: These are however limited in scope, additional testing is encouraged.)\n"]}]},{"cell_type":"markdown","source":["### Replay buffer"],"metadata":{"id":"CsJuwC4nth4A"}},{"cell_type":"code","source":["class ReplayBuffer:\n","    def __init__(self, size, seed):\n","        \"\"\"\n","        Args:\n","            size (integer): The size of the replay buffer.\n","            minibatch_size (integer): The sample size.\n","            seed (integer): The seed for the random number generator.\n","        \"\"\"\n","        self.buffer = []\n","        self.rand_generator = np.random.RandomState(seed)\n","        self.max_size = size\n","\n","    def append(self, state, action, reward):\n","        \"\"\"\n","        Args:\n","            state (Numpy array): The state.\n","            action (integer): The action.\n","            reward (float): The reward.\n","            terminal (integer): 1 if the next state is a terminal state and 0 otherwise.\n","            next_state (Numpy array): The next state.\n","        \"\"\"\n","        if len(self.buffer) == self.max_size:\n","            del self.buffer[0]\n","        self.buffer.append([state, action, reward])\n","\n","    def sample(self, last_action):\n","        \"\"\"\n","        Returns:\n","            A list of transition tuples including state, action, reward, terinal, and next_state\n","        \"\"\"\n","        state, action, reward = map(list, zip(*self.buffer))\n","        idxs = [elem == last_action for elem in action]\n","        X = [b for a, b in zip(idxs, state) if a]\n","        y = [b for a, b in zip(idxs, reward) if a]\n","\n","        return X, y\n","\n","    def size(self):\n","        return len(self.buffer)"],"metadata":{"id":"mPZIB5GOtm7D"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# if __name__ == \"__main__\":\n","\n","#     buffer = ReplayBuffer(size=100000, seed=1)\n","#     buffer.append([1, 2, 3], 0, 1)\n","#     buffer.append([4, 21, 3], 1, 1)\n","#     buffer.append([0, 1, 1], 0, 0)\n","\n","#     print(buffer.sample(0))"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"XCsLBOzDtozO","executionInfo":{"status":"ok","timestamp":1639145971783,"user_tz":-330,"elapsed":9,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"fbd79fdd-1c19-4e74-ce5d-77d8b66a834c"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["([[1, 2, 3], [0, 1, 1]], [1, 0])\n"]}]},{"cell_type":"markdown","source":["### Data Generator"],"metadata":{"id":"MGlV70x3tpu7"}},{"cell_type":"code","source":["def generate_samples(num_samples, num_features, num_arms, return_dataframe=False):\n","    np.random.seed(1)\n","    # generate pseudo features X and \"true\" arms' weights\n","    X = np.random.randint(0, 4, size=(num_samples, num_features))\n","    actions_weights = np.random.normal(loc=-1., scale=1, size=(num_arms, num_features))\n","\n","    # apply data generating policy\n","    policy_weights = np.random.normal(size=(num_arms, num_features))\n","    action_scores = np.dot(X, policy_weights.T)\n","    action_probs = softmax(action_scores, tau=10)\n","    A = np.zeros((num_samples, 1))\n","    for i in range(num_samples):\n","        A[i, 0] = np.random.choice(range(num_arms), 1, p=action_probs[i, :])\n","\n","    # store probabilities of choosing a particular action\n","    _rows = np.zeros_like(A, dtype=np.intp)\n","    _columns = A.astype(int)\n","    probs = action_probs[_rows, _columns]\n","\n","    # calculate \"true\" outcomes Y\n","    ## broadcasting chosen actions to action weights\n","    matrix_multiplicator = actions_weights[_columns].squeeze()  # (num_samples x num_features) matrix\n","    rewards = np.sum(X * matrix_multiplicator, axis=1).reshape(-1, 1)\n","    Y = (np.sign(rewards) + 1) / 2\n","\n","    if return_dataframe:\n","        column_names = ['X_' + str(i+1) for i in range(num_features)]\n","        X = pd.DataFrame(X, columns=column_names)\n","        A = pd.DataFrame(A, columns=['a'])\n","        Y = pd.DataFrame(Y, columns=['y'])\n","        probs = pd.DataFrame(probs, columns=['probs'])\n","\n","        return pd.concat([X, A, Y, probs], axis=1)\n","    else:\n","        return X, A, Y, probs"],"metadata":{"id":"2bxVP6RGtxDa"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# dataset = generate_samples(100000, 4, 3, True)\n","# dataset.head()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":206},"id":"UH5mKo3htyWH","executionInfo":{"status":"ok","timestamp":1639146016409,"user_tz":-330,"elapsed":4932,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"fca14966-0540-4ed6-e390-3ad54fb80465"},"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>X_1</th>\n","      <th>X_2</th>\n","      <th>X_3</th>\n","      <th>X_4</th>\n","      <th>a</th>\n","      <th>y</th>\n","      <th>probs</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0</th>\n","      <td>1</td>\n","      <td>3</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.266661</td>\n","    </tr>\n","    <tr>\n","      <th>1</th>\n","      <td>3</td>\n","      <td>1</td>\n","      <td>3</td>\n","      <td>1</td>\n","      <td>1.0</td>\n","      <td>0.0</td>\n","      <td>0.236514</td>\n","    </tr>\n","    <tr>\n","      <th>2</th>\n","      <td>3</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>1</td>\n","      <td>1.0</td>\n","      <td>0.0</td>\n","      <td>0.236514</td>\n","    </tr>\n","    <tr>\n","      <th>3</th>\n","      <td>0</td>\n","      <td>3</td>\n","      <td>1</td>\n","      <td>0</td>\n","      <td>1.0</td>\n","      <td>0.0</td>\n","      <td>0.236514</td>\n","    </tr>\n","    <tr>\n","      <th>4</th>\n","      <td>2</td>\n","      <td>1</td>\n","      <td>2</td>\n","      <td>0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.266661</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["   X_1  X_2  X_3  X_4    a    y     probs\n","0    1    3    0    0  0.0  0.0  0.266661\n","1    3    1    3    1  1.0  0.0  0.236514\n","2    3    0    0    1  1.0  0.0  0.236514\n","3    0    3    1    0  1.0  0.0  0.236514\n","4    2    1    2    0  0.0  0.0  0.266661"]},"metadata":{},"execution_count":20}]},{"cell_type":"markdown","source":["### Data loader"],"metadata":{"id":"PYr5NIxatzVv"}},{"cell_type":"code","source":["def data_randomizer(pickle_file, seed=None):\n","    if isinstance(pickle_file, str):\n","        with open(pickle_file, 'rb') as f:\n","            dataset = pickle.load(f)\n","    else:\n","        dataset = pickle_file\n","\n","    actions = sorted(dataset.iloc[:, -3].unique().tolist())\n","    tst_smpl = pd.DataFrame().reindex_like(dataset).dropna()\n","    ratio = 0.1\n","\n","    for action in actions:\n","        action_subsample = dataset[dataset.iloc[:, -3] == action]\n","        action_drop, action_use = train_test_split(action_subsample.index, test_size=ratio,\n","                                                   random_state=seed,\n","                                                   stratify=action_subsample.iloc[:, -2])\n","        tst_smpl = pd.concat([tst_smpl,\n","                              action_subsample.loc[action_use]]).sample(frac=1, random_state=seed)\n","\n","    tst_smpl = tst_smpl.reset_index(drop=True)\n","\n","    del action_drop, action_use\n","\n","    X = tst_smpl.iloc[:, :-3].to_numpy()\n","    A = tst_smpl.iloc[:, -3].to_numpy()\n","    Y = tst_smpl.iloc[:, -2].to_numpy()\n","    probs = tst_smpl.iloc[:, -1].to_numpy()\n","\n","    return X, A, Y/probs"],"metadata":{"id":"3sW2YJahuCbl"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["class BanditDataset(Dataset):\n","    def __init__(self, pickle_file, seed=None):\n","        # load dataset\n","        X, A, Y = data_randomizer(pickle_file, seed)\n","        self.features = X\n","        self.actions = A\n","        self.rewards = Y\n","\n","    def __len__(self):\n","        return len(self.rewards)\n","\n","    def __getitem__(self, idx):\n","        feature_vec = self.features[idx]\n","        action = self.actions[idx]\n","        reward = self.rewards[idx]\n","\n","        return feature_vec, action, reward"],"metadata":{"id":"NcdcUSeruEUD"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# if __name__ == '__main__':\n","#     dir = 'data/mushroom_data_final.pickle'\n","#     data = data_randomizer(dir)\n","\n","#     dataset = BanditDataset(pickle_file=dir, seed=1)\n","#     print(len(dataset))\n","#     print(dataset.__len__())\n","#     print(dataset[420])\n","#     print(dataset[421])\n","#     print(dataset[0])\n","#     print(dataset[1])\n","\n","#     dl = DataLoader(dataset, batch_size=2, shuffle=True)\n","\n","#     print(next(iter(dl)))\n","\n","#     dataset = generate_samples(100000, 4, 3, True)\n","#     dataset = BanditDataset(pickle_file=dataset, seed=1)\n","#     print(len(dataset))\n","#     print(dataset.__len__())\n","#     print(dataset[420])\n","#     print(dataset[421])\n","#     print(dataset[0])\n","#     print(dataset[1])"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"8v7s0-QIuGPR","executionInfo":{"status":"ok","timestamp":1639146132864,"user_tz":-330,"elapsed":5534,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"dada2526-7ef3-476c-e6f7-141d7f62e7c0"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["3251\n","3251\n","(array([5., 0., 0., 1., 0.]), 1.0, 1.0)\n","(array([7., 1., 0., 1., 0.]), 0.0, 0.0)\n","(array([3., 0., 0., 0., 1.]), 0.0, 0.0)\n","(array([7., 1., 0., 1., 0.]), 1.0, 1.0)\n","[tensor([[5., 0., 0., 1., 0.],\n","        [5., 0., 1., 1., 1.]], dtype=torch.float64), tensor([0., 0.], dtype=torch.float64), tensor([0., 0.], dtype=torch.float64)]\n","10001\n","10001\n","(array([1., 1., 1., 1.]), 2.0, 0.0)\n","(array([1., 3., 3., 3.]), 2.0, 0.0)\n","(array([3., 3., 0., 3.]), 2.0, 0.0)\n","(array([2., 1., 0., 1.]), 1.0, 0.0)\n"]}]},{"cell_type":"markdown","source":["### Plot script"],"metadata":{"id":"IsZ9fTsiuPwc"}},{"cell_type":"code","source":["def get_leveled_data(arr):\n","    \"\"\"\n","    Args:\n","        arr: list of lists os different length\n","    Returns:\n","        average result over arr, axis=0\n","    \"\"\"\n","    b = np.zeros([len(arr), len(max(arr, key=lambda x: len(x)))])\n","    b[:, :] = np.nan\n","    for i, j in enumerate(arr):\n","        b[i][0:len(j)] = j\n","\n","    return b"],"metadata":{"id":"Whuf3fm3uUdS"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["def smooth(data, k):\n","    num_episodes = data.shape[1]\n","    num_runs = data.shape[0]\n","\n","    smoothed_data = np.zeros((num_runs, num_episodes))\n","\n","    for i in range(num_episodes):\n","        if i < k:\n","            smoothed_data[:, i] = np.mean(data[:, :i + 1], axis=1)\n","        else:\n","            smoothed_data[:, i] = np.mean(data[:, i - k:i + 1], axis=1)\n","\n","    return smoothed_data"],"metadata":{"id":"eDcvasCOubnj"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["def plot_result(result_batch, result_online, batch_size):\n","    plt_agent_sweeps = []\n","    num_steps = np.inf\n","\n","    fig, ax = plt.subplots(figsize=(8, 6))\n","\n","    for data, label in zip([result_batch, result_online], ['batch', 'online']):\n","        sum_reward_data = get_leveled_data(data)\n","\n","        # smooth data\n","        smoothed_sum_reward = smooth(data=sum_reward_data, k=100)\n","\n","        mean_smoothed_sum_reward = np.mean(smoothed_sum_reward, axis=0)\n","\n","        if mean_smoothed_sum_reward.shape[0] < num_steps:\n","            num_steps = mean_smoothed_sum_reward.shape[0]\n","\n","        plot_x_range = np.arange(0, mean_smoothed_sum_reward.shape[0])\n","        graph_current_agent_sum_reward, = ax.plot(plot_x_range, mean_smoothed_sum_reward[:],\n","                                                  label=label)\n","        plt_agent_sweeps.append(graph_current_agent_sum_reward)\n","\n","\n","    update_points = np.ceil(np.arange(num_steps) / batch_size).astype(int)\n","    ax.plot(plot_x_range, mean_smoothed_sum_reward[update_points], label='upper bound')\n","\n","    ax.legend(handles=plt_agent_sweeps, fontsize=13)\n","    ax.set_title(\"Learning Curve\", fontsize=15)\n","    ax.set_xlabel('Episodes', fontsize=14)\n","    ax.set_ylabel('reward', rotation=0, labelpad=40, fontsize=14)\n","    # ax.set_ylim([-300, 300])\n","\n","    plt.tight_layout()\n","    plt.show()"],"metadata":{"id":"6DI-dp6fuawY"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["## Agents"],"metadata":{"id":"ndaq3Ha6ucgA"}},{"cell_type":"markdown","source":["### Base Agent"],"metadata":{"id":"nK6ccVZfukKl"}},{"cell_type":"markdown","source":["An abstract class that specifies the Agent API for RL-Glue-py."],"metadata":{"id":"5vxRMmXluqMk"}},{"cell_type":"code","source":["class BaseAgent:\n","    \"\"\"Implements the agent for an RL-Glue environment.\n","    Note:\n","        agent_init, agent_start, agent_step, agent_end, agent_cleanup, and\n","        agent_message are required methods.\n","    \"\"\"\n","\n","    __metaclass__ = ABCMeta\n","\n","    def __init__(self):\n","        pass\n","\n","    @abstractmethod\n","    def agent_init(self, agent_info={}):\n","        \"\"\"Setup for the agent called when the experiment first starts.\"\"\"\n","\n","    @abstractmethod\n","    def agent_start(self, observation):\n","        \"\"\"The first method called when the experiment starts, called after\n","        the environment starts.\n","        Args:\n","            observation (Numpy array): the state observation from the environment's evn_start function.\n","        Returns:\n","            The first action the agent takes.\n","        \"\"\"\n","\n","    @abstractmethod\n","    def agent_step(self, reward, observation):\n","        \"\"\"A step taken by the agent.\n","        Args:\n","            reward (float): the reward received for taking the last action taken\n","            observation (Numpy array): the state observation from the\n","                environment's step based, where the agent ended up after the\n","                last step\n","        Returns:\n","            The action the agent is taking.\n","        \"\"\"\n","\n","    @abstractmethod\n","    def agent_end(self, reward):\n","        \"\"\"Run when the agent terminates.\n","        Args:\n","            reward (float): the reward the agent received for entering the terminal state.\n","        \"\"\"\n","\n","    @abstractmethod\n","    def agent_cleanup(self):\n","        \"\"\"Cleanup done after the agent ends.\"\"\"\n","\n","    @abstractmethod\n","    def agent_message(self, message):\n","        \"\"\"A function used to pass information from the agent to the experiment.\n","        Args:\n","            message: The message passed to the agent.\n","        Returns:\n","            The response (or answer) to the message.\n","        \"\"\""],"metadata":{"id":"tIP4FINouoP1"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### Random Agent"],"metadata":{"id":"aJdhotjJuyCg"}},{"cell_type":"code","source":["class RandomAgent(BaseAgent):\n","    def __init__(self):\n","        super().__init__()\n","        self.num_actions = None\n","\n","    def agent_init(self, agent_info=None):\n","        if agent_info is None:\n","            agent_info = {}\n","        self.num_actions = agent_info.get('num_actions', 2)\n","\n","    def agent_start(self, observation):\n","        pass\n","\n","    def agent_step(self, reward, observation):\n","        pass\n","\n","    def agent_end(self, reward):\n","        pass\n","\n","    def agent_cleanup(self):\n","        pass\n","\n","    def agent_message(self, message):\n","        pass\n","\n","    def agent_policy(self, observation):\n","        return np.random.choice(self.num_actions)"],"metadata":{"id":"_nofgAnQu3em"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# if __name__ == '__main__':\n","#     ag = RandomAgent()\n","#     print(ag.num_actions)\n","\n","#     ag.agent_init()\n","#     print(ag.num_actions)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"HwKhrIoVu5M3","executionInfo":{"status":"ok","timestamp":1639146301441,"user_tz":-330,"elapsed":6,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"57147a1f-d407-49ad-8e62-4d6c9d90922e"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["None\n","2\n"]}]},{"cell_type":"code","source":["class Agent(BaseAgent):\n","    \"\"\"agent does *no* learning, selects random action always\"\"\"\n","\n","    def __init__(self):\n","        super().__init__()\n","        self.arm_count = None\n","        self.last_action = None\n","        self.num_actions = None\n","        self.q_values = None\n","        self.step_size = None\n","        self.initial_value = 0.0\n","        self.batch_size = None\n","        self.q_values_oracle = None  # used for batch updates\n","\n","    def agent_init(self, agent_info=None):\n","        \"\"\"Setup for the agent called when the experiment first starts.\"\"\"\n","\n","        if agent_info is None:\n","            agent_info = {}\n","\n","        self.num_actions = agent_info.get(\"num_actions\", 2)\n","        self.initial_value = agent_info.get(\"initial_value\", 0.0)\n","        self.q_values = np.ones(agent_info.get(\"num_actions\", 2)) * self.initial_value\n","        self.step_size = agent_info.get(\"step_size\", 0.1)\n","        self.batch_size = agent_info.get('batch_size', 1)\n","        self.q_values_oracle = self.q_values.copy()\n","        self.arm_count = np.zeros(self.num_actions)  # [0.0 for _ in range(self.num_actions)]\n","        # self.last_action = np.random.choice(self.num_actions)  # set first action to random\n","\n","    def agent_start(self, observation):\n","        \"\"\"The first method called when the experiment starts, called after\n","        the environment starts.\n","        Args:\n","            observation (Numpy array): the state observation from the\n","                environment's evn_start function.\n","        Returns:\n","            The first action the agent takes.\n","        \"\"\"\n","        self.last_action = np.random.choice(self.num_actions)\n","\n","        return self.last_action\n","\n","    def agent_step(self, reward, observation):\n","        \"\"\"A step taken by the agent.\n","        Args:\n","            reward (float): the reward received for taking the last action taken\n","            observation (Numpy array): the state observation from the\n","                environment's step based, where the agent ended up after the\n","                last step\n","        Returns:\n","            The action the agent is taking.\n","        \"\"\"\n","        # local_action = 0  # choose the action here\n","        self.last_action = np.random.choice(self.num_actions)\n","\n","        return self.last_action\n","\n","    def agent_end(self, reward):\n","        pass\n","\n","    def agent_cleanup(self):\n","        pass\n","\n","    def agent_message(self, message):\n","        pass"],"metadata":{"id":"egiYidMHvP7D"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["def argmax(q_values):\n","    \"\"\"\n","    Takes in a list of q_values and returns the index of the item\n","    with the highest value. Breaks ties randomly.\n","    returns: int - the index of the highest value in q_values\n","    \"\"\"\n","    top_value = float(\"-inf\")\n","    ties = []\n","\n","    for i in range(len(q_values)):\n","        if q_values[i] > top_value:\n","            ties = [i]\n","            top_value = q_values[i]\n","        elif q_values[i] == top_value:\n","            ties.append(i)\n","\n","    return np.random.choice(ties)"],"metadata":{"id":"5w620FFCvTu8"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### Greedy Agent"],"metadata":{"id":"F_7QyQIsvraK"}},{"cell_type":"code","source":["class GreedyAgent(Agent):\n","    def __init__(self):\n","        super().__init__()\n","\n","    def agent_init(self, agent_info=None):\n","        if agent_info is None:\n","            agent_info = {}\n","\n","        super().agent_init(agent_info)\n","\n","    def agent_step(self, reward, observation):\n","        \"\"\"\n","        Takes one step for the agent. It takes in a reward and observation and\n","        returns the action the agent chooses at that time step.\n","        Arguments:\n","        reward -- float, the reward the agent received from the environment after taking the last action.\n","        observation -- float, the observed state the agent is in. Do not worry about this as you will not use it\n","                              until future lessons\n","        Returns:\n","        current_action -- int, the action chosen by the agent at the current time step.\n","        \"\"\"\n","\n","        a = self.last_action\n","        self.arm_count[a] += 1\n","        self.q_values_oracle[a] = self.q_values_oracle[a] + 1 / self.arm_count[a] * (reward - self.q_values_oracle[a])\n","\n","        if sum(self.arm_count) % self.batch_size == 0:\n","            self.q_values = self.q_values_oracle.copy()\n","\n","        current_action = argmax(self.q_values)\n","        self.last_action = current_action\n","\n","        return current_action"],"metadata":{"id":"my_xM4bpvVDf"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### ϵ-Greedy Agent"],"metadata":{"id":"6SzDe38oviFg"}},{"cell_type":"code","source":["class EpsilonGreedyAgent(Agent):\n","    def __init__(self):\n","        super().__init__()\n","        self.epsilon = None\n","\n","    def agent_init(self, agent_info=None):\n","        if agent_info is None:\n","            agent_info = {}\n","\n","        super().agent_init(agent_info)\n","        self.epsilon = agent_info.get(\"epsilon\", 0.1)\n","\n","    def agent_step(self, reward, observation):\n","        \"\"\"\n","        Takes one step for the agent. It takes in a reward and observation and\n","        returns the action the agent chooses at that time step.\n","        Arguments:\n","        reward -- float, the reward the agent received from the environment after taking the last action.\n","        observation -- float, the observed state the agent is in. Do not worry about this as you will not use it\n","                              until future lessons\n","        Returns:\n","        current_action -- int, the action chosen by the agent at the current time step.\n","        \"\"\"\n","\n","        a = self.last_action\n","\n","        self.arm_count[a] += 1\n","        self.q_values_oracle[a] = self.q_values_oracle[a] + 1 / self.arm_count[a] * (reward - self.q_values_oracle[a])\n","\n","        if np.sum(self.arm_count) % self.batch_size == 0:\n","            self.q_values = self.q_values_oracle.copy()\n","\n","        if np.random.random() < self.epsilon:\n","            current_action = np.random.choice(range(len(self.arm_count)))\n","        else:\n","            current_action = argmax(self.q_values)\n","\n","        self.last_action = current_action\n","\n","        return current_action"],"metadata":{"id":"bPTYzIzTvWYZ"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### UCB Agent"],"metadata":{"id":"Y635PDE6vgdI"}},{"cell_type":"code","source":["class UCBAgent(Agent):\n","    def __init__(self):\n","        super().__init__()\n","        self.upper_bounds = None\n","        self.alpha = None  # exploration parameter\n","\n","    def agent_init(self, agent_info=None):\n","        if agent_info is None:\n","            agent_info = {}\n","\n","        super().agent_init(agent_info)\n","        self.alpha = agent_info.get(\"alpha\", 1.0)\n","        self.arm_count = np.ones(self.num_actions)\n","        self.upper_bounds = np.sqrt(np.log(np.sum(self.arm_count)) / self.arm_count)\n","\n","    def agent_step(self, reward, observation):\n","        a = self.last_action\n","\n","        self.arm_count[a] += 1\n","        self.q_values_oracle[a] = self.q_values_oracle[a] + 1 / self.arm_count[a] * (reward - self.q_values_oracle[a])\n","\n","        # since we start with arms_count = np.ones(num_actions),\n","        # we should subtract num_actions to get number of the current round\n","        if (np.sum(self.arm_count) - self.num_actions) % self.batch_size == 0:\n","            self.q_values = self.q_values_oracle.copy()\n","            self.upper_bounds = np.sqrt(np.log(np.sum(self.arm_count)) / self.arm_count)\n","\n","        # if min(self.q_values + self.alpha * self.upper_bounds) < max(self.q_values):\n","        #     print(f'Distinguish suboptimal arm at step {sum(self.arm_count)}')\n","        current_action = argmax(self.q_values + self.alpha * self.upper_bounds)\n","        # current_action = np.argmax(self.q_values + self.alpha * self.upper_bounds)\n","\n","        self.last_action = current_action\n","\n","        return current_action"],"metadata":{"id":"PV_TaTk1vX0T"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### TS Agent"],"metadata":{"id":"UiTTsyhFvdwc"}},{"cell_type":"code","source":["class TSAgent(Agent):\n","    def agent_step(self, reward, observation):\n","        a = self.last_action\n","        self.arm_count[a] += 1\n","        self.q_values_oracle[a] = self.q_values_oracle[a] + 1 / self.arm_count[a] * (reward - self.q_values_oracle[a])\n","\n","        if (np.sum(self.arm_count) - self.num_actions) % self.batch_size == 0:\n","            self.q_values = self.q_values_oracle.copy()\n","\n","        # sample from posteriors\n","        theta = [beta.rvs(a + 1, b + 1, size=1) for a, b in\n","                 zip(self.q_values * self.arm_count, self.arm_count - self.q_values * self.arm_count)]\n","        # choose the max realization\n","        current_action = argmax(theta)\n","        self.last_action = current_action\n","\n","        return current_action"],"metadata":{"id":"TT3oZeNavY4e"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### LinUCB Agent"],"metadata":{"id":"-IL38-vIzIG1"}},{"cell_type":"code","source":["class LinUCBAgent(BaseAgent):\n","\n","    def __init__(self):\n","        super().__init__()\n","        self.name = \"LinUCB\"\n","\n","    def agent_init(self, agent_info=None):\n","\n","        if agent_info is None:\n","            agent_info = {}\n","\n","        self.num_actions = agent_info.get('num_actions', 3)\n","        self.alpha = agent_info.get('alpha', 1)\n","        self.batch_size = agent_info.get('batch_size', 1)\n","        # Set random seed for policy for each run\n","        self.policy_rand_generator = np.random.RandomState(agent_info.get(\"seed\", None))\n","\n","        self.last_action = None\n","        self.last_state = None\n","        self.num_round = None\n","\n","    def agent_policy(self, observation):\n","        p_t = np.zeros(self.num_actions)\n","\n","        for i in range(self.num_actions):\n","            # initialize theta hat\n","            self.theta = inv(self.A[i]).dot(self.b[i])\n","            # get context of each arm from flattened vector of length 100\n","            cntx = observation\n","            # get gain reward of each arm\n","            p_t[i] = self.theta.T.dot(cntx) + self.alpha * np.sqrt(np.maximum(cntx.dot(inv(self.A[i]).dot(cntx)), 0))\n","        # action = np.random.choice(np.where(p_t == max(p_t))[0])\n","        action = self.policy_rand_generator.choice(np.where(p_t == max(p_t))[0])\n","\n","        return action\n","\n","    def agent_start(self, observation):\n","        # Specify feature dimension\n","        self.ndims = len(observation)\n","\n","        self.A = np.zeros((self.num_actions, self.ndims, self.ndims))\n","        # Instantiate b as a 0 vector of length ndims.\n","        self.b = np.zeros((self.num_actions, self.ndims, 1))\n","        # set each A per arm as identity matrix of size ndims\n","        for arm in range(self.num_actions):\n","            self.A[arm] = np.eye(self.ndims)\n","\n","        self.A_oracle = self.A.copy()\n","        self.b_oracle = self.b.copy()\n","\n","        self.last_state = observation\n","        self.last_action = self.agent_policy(self.last_state)\n","        self.num_round = 0\n","\n","        return self.last_action\n","\n","    def agent_update(self, reward):\n","        self.A_oracle[self.last_action] = self.A_oracle[self.last_action] + np.outer(self.last_state, self.last_state)\n","        self.b_oracle[self.last_action] = np.add(self.b_oracle[self.last_action].T, self.last_state * reward).reshape(self.ndims, 1)\n","\n","    def agent_step(self, reward, observation):\n","        if reward is not None:\n","            self.agent_update(reward)\n","            # it is a good question whether I should increment num_round outside\n","            # condition or not (since theoretical result doesn't clarify this\n","            self.num_round += 1\n","\n","        if self.num_round % self.batch_size == 0:\n","            self.A = self.A_oracle.copy()\n","            self.b = self.b_oracle.copy()\n","\n","        self.last_state = observation\n","        self.last_action = self.agent_policy(self.last_state)\n","\n","        return self.last_action\n","\n","    def agent_end(self, reward):\n","        if reward is not None:\n","            self.agent_update(reward)\n","            self.num_round += 1\n","\n","        if self.num_round % self.batch_size == 0:\n","            self.A = self.A_oracle.copy()\n","            self.b = self.b_oracle.copy()\n","\n","    def agent_message(self, message):\n","        pass\n","\n","    def agent_cleanup(self):\n","        pass"],"metadata":{"id":"gyXQwM2-zKR_"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# if __name__ == '__main__':\n","#     agent_info = {'alpha': 2,\n","#                   'num_actions': 4,\n","#                   'seed': 1}\n","\n","#     # check initialization\n","#     linucb = LinUCBAgent()\n","#     linucb.agent_init(agent_info)\n","#     print(linucb.num_actions, linucb.alpha)\n","\n","#     assert linucb.num_actions == 4\n","#     assert linucb.alpha == 2\n","\n","#     # check policy\n","#     observation = np.array([1, 2, 5, 0])\n","#     linucb.A = np.zeros((linucb.num_actions, len(observation), len(observation)))\n","#     # Instantiate b as a 0 vector of length ndims.\n","#     linucb.b = np.zeros((linucb.num_actions, len(observation), 1))\n","#     # set each A per arm as identity matrix of size ndims\n","#     for arm in range(linucb.num_actions):\n","#         linucb.A[arm] = np.eye(len(observation))\n","\n","#     action = linucb.agent_policy(observation)\n","#     print(action)\n","\n","#     assert action == 1\n","\n","#     # check start\n","#     observation = np.array([1, 2, 5, 0])\n","#     linucb.agent_start(observation)\n","#     print(linucb.ndims)\n","#     print(linucb.last_state, linucb.last_action)\n","\n","#     assert linucb.ndims == len(observation)\n","#     assert np.allclose(linucb.last_state, observation)\n","#     assert np.allclose(linucb.b, np.zeros((linucb.num_actions, len(observation), 1)))\n","#     assert np.allclose(linucb.A, np.array([np.eye(len(observation)), np.eye(len(observation)),\n","#                                            np.eye(len(observation)), np.eye(len(observation))]))\n","#     assert linucb.last_action == 3\n","\n","#     # check step\n","#     observation = np.array([5, 3, 1, 2])\n","#     reward = 1\n","\n","#     action = linucb.agent_step(reward, observation)\n","#     print(linucb.A)\n","#     print(linucb.b)\n","#     print(action)\n","\n","#     true_A = np.array([[2., 2., 5., 0.],\n","#                        [2., 5., 10., 0.],\n","#                        [5., 10., 26., 0.],\n","#                        [0., 0., 0., 1.]])\n","\n","#     true_b = np.array([[1.],\n","#                        [2.],\n","#                        [5.],\n","#                        [0.]])\n","\n","#     for i in range(3):\n","#         assert np.allclose(linucb.A[i], np.eye(4))\n","#         assert np.allclose(linucb.b[i], np.zeros((linucb.num_actions, 4, 1)))\n","#     assert np.allclose(linucb.A[3], true_A)\n","#     assert np.allclose(linucb.b[3], true_b)\n","#     assert linucb.last_action == 0\n","\n","#     observation = np.array([3, 1, 3, 5])\n","#     reward = None\n","\n","#     action = linucb.agent_step(reward, observation)\n","#     print(linucb.A)\n","#     print(linucb.b)\n","#     print(action)\n","\n","#     assert np.allclose(linucb.A[3], true_A)\n","#     assert np.allclose(linucb.b[3], true_b)\n","#     assert action == 0\n","\n","#     # check batch size\n","#     agent_info = {'alpha': 2,\n","#                   'num_actions': 4,\n","#                   'seed': 1,\n","#                   'batch_size': 2}\n","#     linucb = LinUCBAgent()\n","#     linucb.agent_init(agent_info)\n","#     observation = np.array([1, 2, 5, 0])\n","#     linucb.agent_start(observation)\n","#     assert linucb.num_round == 0\n","#     assert linucb.last_action == 1\n","\n","#     observation = np.array([5, 3, 1, 2])\n","#     reward = 1\n","\n","#     action = linucb.agent_step(reward, observation)\n","#     assert linucb.num_round == 1\n","#     assert np.allclose(linucb.b, np.zeros((linucb.num_actions, len(observation), 1)))\n","#     assert np.allclose(linucb.A, np.array([np.eye(len(observation)), np.eye(len(observation)),\n","#                                            np.eye(len(observation)), np.eye(len(observation))]))\n","\n","#     for i in [0, 2, 3]:\n","#         assert np.allclose(linucb.A_oracle[i], np.eye(4))\n","#         assert np.allclose(linucb.b_oracle[i], np.zeros((linucb.num_actions, 4, 1)))\n","#     assert np.allclose(linucb.A_oracle[1], true_A)\n","#     assert np.allclose(linucb.b_oracle[1], true_b)\n","\n","#     observation = np.array([3, 1, 3, 5])\n","#     reward = None\n","#     action = linucb.agent_step(reward, observation)\n","#     # sinse reward is None, nothing should happen\n","#     assert linucb.num_round == 1\n","#     assert np.allclose(linucb.b, np.zeros((linucb.num_actions, len(observation), 1)))\n","#     assert np.allclose(linucb.A, np.array([np.eye(len(observation)), np.eye(len(observation)),\n","#                                            np.eye(len(observation)), np.eye(len(observation))]))\n","\n","#     for i in [0, 2, 3]:\n","#         assert np.allclose(linucb.A_oracle[i], np.eye(4))\n","#         assert np.allclose(linucb.b_oracle[i], np.zeros((linucb.num_actions, 4, 1)))\n","#     assert np.allclose(linucb.A_oracle[1], true_A)\n","#     assert np.allclose(linucb.b_oracle[1], true_b)\n","\n","#     observation = np.array([3, 0, 2, 5])\n","#     reward = 0\n","#     action = linucb.agent_step(reward, observation)\n","\n","#     assert linucb.num_round == 2\n","#     assert np.allclose(linucb.b, linucb.b_oracle)\n","#     assert np.allclose(linucb.A, linucb.A_oracle)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"Nkdq0ya9zNlE","executionInfo":{"status":"ok","timestamp":1639148410291,"user_tz":-330,"elapsed":5,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"b7a9b9c5-e3d1-4fc8-bdc5-321172c6ebdb"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["4 2\n","1\n","4\n","[1 2 5 0] 3\n","[[[ 1.  0.  0.  0.]\n","  [ 0.  1.  0.  0.]\n","  [ 0.  0.  1.  0.]\n","  [ 0.  0.  0.  1.]]\n","\n"," [[ 1.  0.  0.  0.]\n","  [ 0.  1.  0.  0.]\n","  [ 0.  0.  1.  0.]\n","  [ 0.  0.  0.  1.]]\n","\n"," [[ 1.  0.  0.  0.]\n","  [ 0.  1.  0.  0.]\n","  [ 0.  0.  1.  0.]\n","  [ 0.  0.  0.  1.]]\n","\n"," [[ 2.  2.  5.  0.]\n","  [ 2.  5. 10.  0.]\n","  [ 5. 10. 26.  0.]\n","  [ 0.  0.  0.  1.]]]\n","[[[0.]\n","  [0.]\n","  [0.]\n","  [0.]]\n","\n"," [[0.]\n","  [0.]\n","  [0.]\n","  [0.]]\n","\n"," [[0.]\n","  [0.]\n","  [0.]\n","  [0.]]\n","\n"," [[1.]\n","  [2.]\n","  [5.]\n","  [0.]]]\n","0\n","[[[ 1.  0.  0.  0.]\n","  [ 0.  1.  0.  0.]\n","  [ 0.  0.  1.  0.]\n","  [ 0.  0.  0.  1.]]\n","\n"," [[ 1.  0.  0.  0.]\n","  [ 0.  1.  0.  0.]\n","  [ 0.  0.  1.  0.]\n","  [ 0.  0.  0.  1.]]\n","\n"," [[ 1.  0.  0.  0.]\n","  [ 0.  1.  0.  0.]\n","  [ 0.  0.  1.  0.]\n","  [ 0.  0.  0.  1.]]\n","\n"," [[ 2.  2.  5.  0.]\n","  [ 2.  5. 10.  0.]\n","  [ 5. 10. 26.  0.]\n","  [ 0.  0.  0.  1.]]]\n","[[[0.]\n","  [0.]\n","  [0.]\n","  [0.]]\n","\n"," [[0.]\n","  [0.]\n","  [0.]\n","  [0.]]\n","\n"," [[0.]\n","  [0.]\n","  [0.]\n","  [0.]]\n","\n"," [[1.]\n","  [2.]\n","  [5.]\n","  [0.]]]\n","0\n"]}]},{"cell_type":"markdown","source":["### LinTS Agent"],"metadata":{"id":"4L_OJQQEvP4k"}},{"cell_type":"code","source":["class LinTSAgent(BaseAgent):\n","\n","    def __init__(self):\n","        super().__init__()\n","        self.name = \"LinTS\"\n","\n","    def agent_init(self, agent_info=None):\n","\n","        if agent_info is None:\n","            agent_info = {}\n","\n","        self.num_actions = agent_info.get('num_actions', 3)\n","        self.alpha = agent_info.get('alpha', 1)\n","        self.lambda_ = agent_info.get('lambda', 1)\n","        self.batch_size = agent_info.get('batch_size', 1)\n","        # Set random seed for policy for each run\n","        self.policy_rand_generator = np.random.RandomState(agent_info.get(\"seed\", None))\n","\n","        self.replay_buffer = ReplayBuffer(agent_info['replay_buffer_size'],\n","                                          agent_info.get(\"seed\"))\n","\n","\n","        self.last_action = None\n","        self.last_state = None\n","        self.num_round = None\n","\n","    def agent_policy(self, observation, mode='sample'):\n","        p_t = np.zeros(self.num_actions)\n","        cntx = observation\n","\n","        for i in range(self.num_actions):\n","            # sampling weights after update\n","            self.w = self.get_weights(i)\n","\n","            # using weight depending on mode\n","            if mode == 'sample':\n","                w = self.w  # weights are samples of posteriors\n","            elif mode == 'expected':\n","                w = self.m[i]  # weights are expected values of posteriors\n","            else:\n","                raise Exception('mode not recognized!')\n","\n","            # calculating probabilities\n","            p_t[i] = 1 / (1 + np.exp(-1 * cntx.dot(w)))\n","            action = self.policy_rand_generator.choice(np.where(p_t == max(p_t))[0])\n","            # probs = softmax(p_t.reshape(1, -1))\n","            # action = self.policy_rand_generator.choice(a=range(self.num_actions), p=probs)\n","\n","        return action\n","\n","    def get_weights(self, arm):\n","        return np.random.normal(self.m[arm], self.alpha * self.q[arm] ** (-1.0), size=len(self.w))\n","\n","        # the loss function\n","    def loss(self, w, *args):\n","        X, y = args\n","        return 0.5 * (self.q[self.last_action] * (w - self.m[self.last_action])).dot(w - self.m[self.last_action]) + np.sum(\n","            [np.log(1 + np.exp(-y[j] * w.dot(X[j]))) for j in range(y.shape[0])])\n","\n","    # the gradient\n","    def grad(self, w, *args):\n","        X, y = args\n","        return self.q[self.last_action] * (w - self.m[self.last_action]) + (-1) * np.array(\n","            [y[j] * X[j] / (1. + np.exp(y[j] * w.dot(X[j]))) for j in range(y.shape[0])]).sum(axis=0)\n","\n","    # fitting method\n","    def agent_update(self, X, y):\n","        # step 1, find w\n","        self.w = minimize(self.loss, self.w, args=(X, y), jac=self.grad, method=\"L-BFGS-B\",\n","                          options={'maxiter': 20, 'disp': False}).x\n","        # self.m_oracle[self.last_action] = self.w\n","        self.m[self.last_action] = self.w\n","\n","        # step 2, update q\n","        P = (1 + np.exp(1 - X.dot(self.m[self.last_action]))) ** (-1)\n","        #self.q_oracle[self.last_action] = self.q[self.last_action] + (P * (1 - P)).dot(X ** 2)\n","        self.q[self.last_action] = self.q[self.last_action] + (P * (1 - P)).dot(X ** 2)\n","\n","    def agent_start(self, observation):\n","        # Specify feature dimension\n","        self.ndims = len(observation)\n","\n","        # initializing parameters of the model\n","        self.m = np.zeros((self.num_actions, self.ndims))\n","        self.q = np.ones((self.num_actions, self.ndims)) * self.lambda_\n","        # initializing weights using any arm (e.g. 0) because they all equal\n","        self.w = np.array([0.]*self.ndims, dtype=np.float64)\n","\n","        # self.m_oracle = self.m.copy()\n","        # self.q_oracle = self.q.copy()\n","\n","        self.last_state = observation\n","        self.last_action = self.agent_policy(self.last_state)\n","        self.num_round = 0\n","\n","        return self.last_action\n","\n","\n","    def agent_step(self, reward, observation):\n","        # Append new experience to replay buffer\n","        if reward is not None:\n","            self.replay_buffer.append(self.last_state, self.last_action, reward)\n","            # it is a good question whether I should increment num_round outside\n","            # condition or not (since theoretical result doesn't clarify this\n","            self.num_round += 1\n","\n","            if self.num_round % self.batch_size == 0:\n","                X, y = self.replay_buffer.sample(self.last_action)\n","                X = np.array(X)\n","                y = np.array(y)\n","                self.agent_update(X, y)\n","                # self.m = self.m_oracle.copy()\n","                # self.q = self.q_oracle.copy()\n","\n","        self.last_state = observation\n","        self.last_action = self.agent_policy(self.last_state)\n","\n","        return self.last_action\n","\n","    def agent_end(self, reward):\n","        # Append new experience to replay buffer\n","        if reward is not None:\n","            self.replay_buffer.append(self.last_state, self.last_action, reward)\n","            # it is a good question whether I should increment num_round outside\n","            # condition or not (since theoretical result doesn't clarify this\n","            self.num_round += 1\n","\n","            if self.num_round % self.batch_size == 0:\n","                X, y = self.replay_buffer.sample(self.last_action)\n","                X = np.array(X)\n","                y = np.array(y)\n","                self.agent_update(X, y)\n","                # self.m = self.m_oracle.copy()\n","                # self.q = self.q_oracle.copy()\n","\n","    def agent_message(self, message):\n","        pass\n","\n","    def agent_cleanup(self):\n","        pass"],"metadata":{"id":"X7mSGLS4vP20"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# if __name__ == '__main__':\n","#     agent_info = {'alpha': 2,\n","#                   'num_actions': 3,\n","#                   'seed': 1,\n","#                   'lambda': 2,\n","#                   'replay_buffer_size': 100000}\n","\n","#     np.random.seed(1)\n","#     # check initialization\n","#     lints = LinTSAgent()\n","#     lints.agent_init(agent_info)\n","#     print(lints.num_actions, lints.alpha, lints.lambda_)\n","\n","#     assert lints.num_actions == 3\n","#     assert lints.alpha == 2\n","#     assert lints.lambda_ == 2\n","\n","#     # check agent policy\n","#     observation = np.array([1, 2, 5, 0])\n","#     lints.m = np.zeros((lints.num_actions, len(observation)))\n","#     lints.q = np.ones((lints.num_actions, len(observation))) * lints.lambda_\n","#     lints.w = np.random.normal(lints.m[0], lints.alpha * lints.q[0] ** (-1.0), size=len(observation))\n","#     print(lints.w)\n","#     action = lints.agent_policy(observation)\n","#     print(action)\n","\n","#     # check agent start\n","#     observation = np.array([1, 2, 5, 0])\n","#     lints.agent_start(observation)\n","#     # manually reassign w to np.random.normal, because I np.seed doesn't work inside the class\n","#     np.random.seed(1)\n","#     lints.w = np.random.normal(lints.m[0], lints.alpha * lints.q[0] ** (-1.0), size=len(observation))\n","#     print(lints.ndims)\n","#     print(lints.last_state, lints.last_action)\n","#     print(lints.last_action)\n","#     assert lints.ndims == len(observation)\n","#     assert np.allclose(lints.last_state, observation)\n","#     assert np.allclose(lints.m, np.zeros((lints.num_actions, lints.ndims)))\n","#     assert np.allclose(lints.q, np.ones((lints.num_actions, lints.ndims)) * lints.lambda_)\n","#     assert np.allclose(lints.w, np.array([ 1.62434536, -0.61175641, -0.52817175, -1.07296862]))\n","#     # assert lints.last_action == 1\n","\n","#     # check step\n","#     observation = np.array([5, 3, 1, 2])\n","#     reward = 1\n","#     action = lints.agent_step(reward, observation)\n","#     print(action)\n","\n","#     observation = np.array([1, 3, 2, 1])\n","#     reward = 0\n","#     action = lints.agent_step(reward, observation)\n","#     print(action)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"yr9HfGeJy-vR","executionInfo":{"status":"ok","timestamp":1639148425016,"user_tz":-330,"elapsed":3,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"4037b3ed-73b8-4057-8ea7-c1cb4d7fd177"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["3 2 2\n","[ 1.62434536 -0.61175641 -0.52817175 -1.07296862]\n","1\n","4\n","[1 2 5 0] 1\n","1\n","1\n","1\n"]}]},{"cell_type":"markdown","source":["## Environments"],"metadata":{"id":"IQr77JFsu6Jq"}},{"cell_type":"markdown","source":["### Base Environment"],"metadata":{"id":"oAuUEthsu9TG"}},{"cell_type":"markdown","source":["Abstract environment base class for RL-Glue-py."],"metadata":{"id":"pEz4-17VvAZQ"}},{"cell_type":"code","source":["class BaseEnvironment:\n","    \"\"\"Implements the environment for an RLGlue environment\n","    Note:\n","        env_init, env_start, env_step, env_cleanup, and env_message are required\n","        methods.\n","    \"\"\"\n","\n","    __metaclass__ = ABCMeta\n","\n","    def __init__(self):\n","        reward = None\n","        observation = None\n","        termination = None\n","        self.reward_state_term = (reward, observation, termination)\n","\n","    @abstractmethod\n","    def env_init(self, env_info={}):\n","        \"\"\"Setup for the environment called when the experiment first starts.\n","        Note:\n","            Initialize a tuple with the reward, first state observation, boolean\n","            indicating if it's terminal.\n","        \"\"\"\n","\n","    @abstractmethod\n","    def env_start(self):\n","        \"\"\"The first method called when the experiment starts, called before the\n","        agent starts.\n","        Returns:\n","            The first state observation from the environment.\n","        \"\"\"\n","\n","    @abstractmethod\n","    def env_step(self, action):\n","        \"\"\"A step taken by the environment.\n","        Args:\n","            action: The action taken by the agent\n","        Returns:\n","            (float, state, Boolean): a tuple of the reward, state observation,\n","                and boolean indicating if it's terminal.\n","        \"\"\"\n","\n","    @abstractmethod\n","    def env_cleanup(self):\n","        \"\"\"Cleanup done after the environment ends\"\"\"\n","\n","    @abstractmethod\n","    def env_message(self, message):\n","        \"\"\"A message asking the environment for information\n","        Args:\n","            message: the message passed to the environment\n","        Returns:\n","            the response (or answer) to the message\n","        \"\"\""],"metadata":{"id":"qqZDb6A9u-85"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### k-arm Environment"],"metadata":{"id":"_Mm2zTKyv3Fb"}},{"cell_type":"code","source":["class Environment(BaseEnvironment):\n","    \"\"\"Implements the environment for an RLGlue environment\n","    Note:\n","        env_init, env_start, env_step, env_cleanup, and env_message are required\n","        methods.\n","    \"\"\"\n","\n","    actions = [0]\n","\n","    def __init__(self):\n","        super().__init__()\n","        reward = None\n","        observation = None\n","        termination = None\n","        self.seed = None\n","        self.k = None\n","        self.reward_type = None\n","        self.custom_arms = None\n","        self.reward_state_term = (reward, observation, termination)\n","        self.count = 0\n","        self.arms = []\n","        self.subopt_gaps = None\n","\n","    def env_init(self, env_info=None):\n","        \"\"\"Setup for the environment called when the experiment first starts.\n","        Note:\n","            Initialize a tuple with the reward, first state observation, boolean\n","            indicating if it's terminal.\n","        \"\"\"\n","\n","        if env_info is None:\n","            env_info = {}\n","        self.k = env_info.get(\"num_actions\", 2)\n","        self.reward_type = env_info.get(\"reward_type\", \"subgaussian\")\n","        self.custom_arms = env_info.get(\"arms_values\", None)\n","\n","        if self.reward_type not in ['Bernoulli', 'subgaussian']:\n","            raise ValueError('Unknown reward_type: ' + str(self.reward_type))\n","\n","        if self.custom_arms is None:\n","            if self.reward_type == 'Bernoulli':\n","                self.arms = np.random.uniform(0, 1, self.k)\n","            else:\n","                self.arms = np.random.randn(self.k)\n","        else:\n","            self.arms = self.custom_arms\n","        self.subopt_gaps = np.max(self.arms) - self.arms\n","\n","        local_observation = 0  # An empty NumPy array\n","\n","        self.reward_state_term = (0.0, local_observation, False)\n","\n","    def env_start(self):\n","        \"\"\"The first method called when the experiment starts, called before the\n","        agent starts.\n","        Returns:\n","            The first state observation from the environment.\n","        \"\"\"\n","\n","        return self.reward_state_term[1]\n","\n","    def env_step(self, action):\n","        \"\"\"A step taken by the environment.\n","        Args:\n","            action: The action taken by the agent\n","        Returns:\n","            (float, state, Boolean): a tuple of the reward, state observation,\n","                and boolean indicating if it's terminal.\n","        \"\"\"\n","        if self.reward_type == 'Bernoulli':\n","            reward = np.random.binomial(1, self.arms[action], 1)\n","        else:\n","            reward = self.arms[action] + np.random.randn()\n","        obs = self.reward_state_term[1]\n","\n","        self.reward_state_term = (reward, obs, False)\n","\n","        return self.reward_state_term\n","\n","    def env_cleanup(self):\n","        \"\"\"Cleanup done after the environment ends\"\"\"\n","        pass\n","\n","    def env_message(self, message):\n","        \"\"\"A message asking the environment for information\n","        Args:\n","            message (string): the message passed to the environment\n","        Returns:\n","            string: the response (or answer) to the message\n","        \"\"\"\n","        if message == \"what is the current reward?\":\n","            return \"{}\".format(self.reward_state_term[0])\n","\n","        # else\n","        return \"I don't know how to respond to your message\""],"metadata":{"id":"YLTkxvLGvC4g"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### Offline Evaluator"],"metadata":{"id":"naYFRhirzYfu"}},{"cell_type":"code","source":["class OfflineEvaluator:\n","    def __init__(self, eval_info=None):\n","\n","        if eval_info is None:\n","            eval_info = {}\n","\n","        self.dataset = eval_info['dataset']\n","        self.agent = eval_info['agent']\n","\n","        if not isinstance(self.dataset, Dataset):\n","            raise TypeError('dataset ' + \"must be a \" + str(Dataset))\n","        if not isinstance(self.agent, BaseAgent):\n","            raise TypeError('agent ' + \"must be a \" + str(BaseAgent))\n","\n","        self.total_reward = None\n","        self.average_reward = None\n","        self.num_matches = None\n","        self.idxs = range(self.dataset.__len__())\n","        self.counter = None\n","\n","    def eval_start(self):\n","        self.total_reward = 0\n","        self.average_reward = [0]\n","        self.num_matches = 0\n","        self.idxs = range(self.dataset.__len__())\n","        self.counter = 0\n","\n","    def _get_observation(self):\n","        idx = self.idxs[self.counter]\n","        self.counter += 1\n","\n","        return self.dataset.__getitem__(idx)\n","\n","    def eval_step(self):\n","        observation = self._get_observation()\n","\n","        state = observation[0]\n","        true_action = observation[1]\n","        reward = observation[2]\n","\n","        pred_action = self.agent.agent_policy(state)\n","\n","        if true_action != pred_action:\n","            return\n","\n","        self.num_matches += 1\n","        aw_reward = self.average_reward[-1] + (reward - self.average_reward[-1]) / self.num_matches\n","        self.average_reward.append(aw_reward)\n","        self.total_reward += reward\n","\n","    def eval_run(self):\n","        self.eval_start()\n","\n","        while self.counter < self.dataset.__len__():\n","            self.eval_step()\n","\n","        return self.average_reward"],"metadata":{"id":"iJLVLYunzbPe"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# if __name__ == '__main__':\n","\n","#     dir1 = 'data/mushroom_data_final.pickle'\n","\n","#     ra = RandomAgent()\n","#     agent_info = {'num_actions': 2}\n","#     ra.agent_init(agent_info)\n","\n","#     result = []\n","#     result1 = []\n","\n","#     for seed_ in [1, 5, 10]:  # , 2, 3, 32, 123, 76, 987, 2134]:\n","#         dataset = BanditDataset(pickle_file=dir1, seed=seed_)\n","\n","#         eval_info = {'dataset': dataset, 'agent': ra}\n","#         evaluator = OfflineEvaluator(eval_info)\n","\n","#         reward = evaluator.eval_run()\n","\n","#         result.append(reward)\n","#         result1.append(evaluator.total_reward)\n","\n","#     for elem in result:\n","#         plt.plot(elem)\n","#     plt.legend()\n","#     plt.show()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":282},"id":"BY8sASekzc6S","executionInfo":{"status":"ok","timestamp":1639148441141,"user_tz":-330,"elapsed":14,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"1ddaab3a-5b2f-41a8-842f-407aaa3b1f62"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stderr","text":["No handles with labels found to put in legend.\n"]},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","source":["### Replay Environment"],"metadata":{"id":"tL1Ur1CtzlGx"}},{"cell_type":"code","source":["class ReplayEnvironment(BaseEnvironment):\n","    dataset: BanditDataset\n","\n","    def __init__(self):\n","        super().__init__()\n","        self.counter = None\n","        self.last_observation = None\n","\n","    def env_init(self, env_info=None):\n","        \"\"\"\n","        Set parameters needed to setup the replay SavePilot environment.\n","        Assume env_info dict contains:\n","        {\n","            pickle_file: data directory [str]\n","        }\n","        Args:\n","            env_info (dict):\n","        \"\"\"\n","        if env_info is None:\n","            env_info = {}\n","\n","        directory = env_info['pickle_file']\n","        seed = env_info.get('seed', None)\n","        self.dataset = BanditDataset(directory, seed)\n","        self.idxs = range(self.dataset.__len__())\n","        self.counter = 0\n","\n","    def _get_observation(self):\n","        idx = self.idxs[self.counter]\n","\n","        return self.dataset.__getitem__(idx)\n","\n","    def env_start(self):\n","        self.last_observation = self._get_observation()\n","\n","        state = self.last_observation[0]\n","        reward = None\n","        is_terminal = False\n","\n","        self.reward_state_term = (reward, state, is_terminal)\n","        self.counter += 1\n","\n","        # return first state from the environment\n","        return self.reward_state_term[1]\n","\n","    def env_step(self, action):\n","        true_action = self.last_observation[1]\n","        reward = self.last_observation[2]\n","\n","        if true_action != action:\n","            reward = None\n","\n","        observation = self._get_observation()\n","        state = observation[0]\n","\n","        if self.counter == self.dataset.__len__() - 1:\n","            is_terminal = True\n","        else:\n","            is_terminal = False\n","\n","        self.reward_state_term = (reward, state, is_terminal)\n","\n","        self.last_observation = observation\n","        self.counter += 1\n","\n","        return self.reward_state_term\n","\n","    def env_cleanup(self):\n","        pass\n","\n","    def env_message(self, message):\n","        pass"],"metadata":{"id":"YK8fE2VyzlEO"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["## Wrappers"],"metadata":{"id":"s42XsgHev0lv"}},{"cell_type":"markdown","source":["### RL Glue"],"metadata":{"id":"OlbboMWrwRV2"}},{"cell_type":"code","source":["class RLGlue:\n","    \"\"\"RLGlue class\n","    args:\n","        env_name (string): the name of the module where the Environment class can be found\n","        agent_name (string): the name of the module where the Agent class can be found\n","    \"\"\"\n","\n","    def __init__(self, env_class, agent_class):\n","        self.environment = env_class()\n","        self.agent = agent_class()\n","\n","        self.total_reward = None\n","        self.average_reward = None\n","        self.last_action = None\n","        self.num_steps = None\n","        self.num_episodes = None\n","        self.num_matches = None\n","\n","    def rl_init(self, agent_init_info={}, env_init_info={}):\n","        \"\"\"Initial method called when RLGlue experiment is created\"\"\"\n","        self.environment.env_init(env_init_info)\n","        self.agent.agent_init(agent_init_info)\n","\n","        self.total_reward = 0.0\n","        self.average_reward = [0]\n","        self.num_steps = 0\n","        self.num_episodes = 0\n","        self.num_matches = 0\n","\n","    def rl_start(self):\n","        \"\"\"Starts RLGlue experiment\n","        Returns:\n","            tuple: (state, action)\n","        \"\"\"\n","\n","        last_state = self.environment.env_start()\n","        self.last_action = self.agent.agent_start(last_state)\n","\n","        observation = (last_state, self.last_action)\n","\n","        return observation\n","\n","    def rl_agent_start(self, observation):\n","        \"\"\"Starts the agent.\n","        Args:\n","            observation: The first observation from the environment\n","        Returns:\n","            The action taken by the agent.\n","        \"\"\"\n","        return self.agent.agent_start(observation)\n","\n","    def rl_agent_step(self, reward, observation):\n","        \"\"\"Step taken by the agent\n","        Args:\n","            reward (float): the last reward the agent received for taking the\n","                last action.\n","            observation : the state observation the agent receives from the\n","                environment.\n","        Returns:\n","            The action taken by the agent.\n","        \"\"\"\n","        return self.agent.agent_step(reward, observation)\n","\n","    def rl_agent_end(self, reward):\n","        \"\"\"Run when the agent terminates\n","        Args:\n","            reward (float): the reward the agent received when terminating\n","        \"\"\"\n","        self.agent.agent_end(reward)\n","\n","    def rl_env_start(self):\n","        \"\"\"Starts RL-Glue environment.\n","        Returns:\n","            (float, state, Boolean): reward, state observation, boolean\n","                indicating termination\n","        \"\"\"\n","        self.total_reward = 0.0\n","        self.num_steps = 1\n","\n","        this_observation = self.environment.env_start()\n","\n","        return this_observation\n","\n","    def rl_env_step(self, action):\n","        \"\"\"Step taken by the environment based on action from agent\n","        Args:\n","            action: Action taken by agent.\n","        Returns:\n","            (float, state, Boolean): reward, state observation, boolean\n","                indicating termination.\n","        \"\"\"\n","        ro = self.environment.env_step(action)\n","        (this_reward, _, terminal) = ro\n","\n","        self.total_reward += this_reward\n","\n","        if terminal:\n","            self.num_episodes += 1\n","        else:\n","            self.num_steps += 1\n","\n","        return ro\n","\n","    def rl_step(self):\n","        \"\"\"Step taken by RLGlue, takes environment step and either step or\n","            end by agent.\n","        Returns:\n","            (float, state, action, Boolean): reward, last state observation,\n","                last action, boolean indicating termination\n","        \"\"\"\n","\n","        (reward, last_state, term) = self.environment.env_step(self.last_action)\n","\n","        if reward is not None:\n","            self.num_matches += 1\n","            aw_reward = self.average_reward[-1] + (reward - self.average_reward[-1]) / self.num_matches\n","            self.average_reward.append(aw_reward)\n","            self.total_reward += reward\n","\n","        if term:\n","            self.num_episodes += 1\n","            self.agent.agent_end(reward)\n","            roat = (reward, last_state, None, term)\n","        else:\n","            self.num_steps += 1\n","            self.last_action = self.agent.agent_step(reward, last_state)\n","            roat = (reward, last_state, self.last_action, term)\n","\n","        return roat\n","\n","    def rl_cleanup(self):\n","        \"\"\"Cleanup done at end of experiment.\"\"\"\n","        self.environment.env_cleanup()\n","        self.agent.agent_cleanup()\n","\n","    def rl_agent_message(self, message):\n","        \"\"\"Message passed to communicate with agent during experiment\n","        Args:\n","            message: the message (or question) to send to the agent\n","        Returns:\n","            The message back (or answer) from the agent\n","        \"\"\"\n","\n","        return self.agent.agent_message(message)\n","\n","    def rl_env_message(self, message):\n","        \"\"\"Message passed to communicate with environment during experiment\n","        Args:\n","            message: the message (or question) to send to the environment\n","        Returns:\n","            The message back (or answer) from the environment\n","        \"\"\"\n","        return self.environment.env_message(message)\n","\n","    def rl_episode(self, max_steps_this_episode):\n","        \"\"\"Runs an RLGlue episode\n","        Args:\n","            max_steps_this_episode (Int): the maximum steps for the experiment to run in an episode\n","        Returns:\n","            Boolean: if the episode should terminate\n","        \"\"\"\n","        is_terminal = False\n","\n","        self.rl_start()\n","\n","        while (not is_terminal) and ((max_steps_this_episode == 0) or\n","                                     (self.num_steps < max_steps_this_episode)):\n","            rl_step_result = self.rl_step()\n","            is_terminal = rl_step_result[3]\n","\n","        return is_terminal\n","\n","    def rl_return(self):\n","        \"\"\"The total reward\n","        Returns:\n","            float: the total reward\n","        \"\"\"\n","        return self.total_reward\n","\n","    def rl_num_steps(self):\n","        \"\"\"The total number of steps taken\n","        Returns:\n","            Int: the total number of steps taken\n","        \"\"\"\n","        return self.num_steps\n","\n","    def rl_num_episodes(self):\n","        \"\"\"The number of episodes\n","        Returns\n","            Int: the total number of episodes\n","        \"\"\"\n","        return self.num_episodes"],"metadata":{"id":"InGf3vAFwTIF"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### Policy"],"metadata":{"id":"MCXJTVo5xRpq"}},{"cell_type":"code","source":["class Policy:\n","    def __init__(self, env, agent):\n","        self.env = env\n","        self.agent = agent\n","        self.rl_glue = None\n","\n","    @abstractmethod\n","    def get_average_performance(self, agent_info=None, env_info=None, exper_info=None):\n","        raise NotImplementedError"],"metadata":{"id":"9RsPVJpnwmad"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### Bandit wrapper"],"metadata":{"id":"RbiWk0wWxTD-"}},{"cell_type":"code","source":["class BanditWrapper(Policy):\n","    def get_average_performance(self, agent_info=None, env_info=None, exper_info=None):\n","\n","        if exper_info is None:\n","            exper_info = {}\n","        if env_info is None:\n","            env_info = {}\n","        if agent_info is None:\n","            agent_info = {}\n","\n","        num_runs = exper_info.get(\"num_runs\", 100)\n","        num_steps = exper_info.get(\"num_steps\", 1000)\n","        return_type = exper_info.get(\"return_type\", None)\n","        seed = exper_info.get(\"seed\", None)\n","\n","        np.random.seed(seed)\n","        seeds = np.random.randint(0, num_runs * 100, num_runs)\n","\n","        all_averages = []\n","        subopt_arm_average = []\n","        best_arm = []\n","        worst_arm = []\n","        all_chosen_arm = []\n","        average_regret = []\n","\n","        for run in tqdm(range(num_runs)):\n","            np.random.seed(seeds[run])\n","\n","            self.rl_glue = RLGlue(self.env, self.agent)\n","            self.rl_glue.rl_init(agent_info, env_info)\n","            (first_state, first_action) = self.rl_glue.rl_start()\n","\n","            worst_position = np.argmin(self.rl_glue.environment.arms)\n","            best_value = np.max(self.rl_glue.environment.arms)\n","            worst_value = np.min(self.rl_glue.environment.arms)\n","            best_arm.append(best_value)\n","            worst_arm.append(worst_value)\n","\n","            scores = [0]\n","            averages = []\n","            subopt_arm = []\n","            chosen_arm_log = []\n","\n","            cum_regret = [0]\n","            delta = self.rl_glue.environment.subopt_gaps[first_action]\n","            cum_regret.append(cum_regret[-1] + delta)\n","\n","            # first action was made in rl_start, that's why run over num_steps-1\n","            for i in range(num_steps-1):\n","                reward, _, action, _ = self.rl_glue.rl_step()\n","                chosen_arm_log.append(action)\n","                scores.append(scores[-1] + reward)\n","                averages.append(scores[-1] / (i + 1))\n","                subopt_arm.append(self.rl_glue.agent.arm_count[worst_position])\n","\n","                delta = self.rl_glue.environment.subopt_gaps[action]\n","                cum_regret.append(cum_regret[-1] + delta)\n","\n","            all_averages.append(averages)\n","            subopt_arm_average.append(subopt_arm)\n","            all_chosen_arm.append(chosen_arm_log)\n","\n","            average_regret.append(cum_regret)\n","\n","        if return_type is None:\n","            returns = (np.mean(all_averages, axis=0),\n","                       np.mean(best_arm))\n","        elif return_type == 'regret':\n","            returns = np.mean(average_regret, axis=0)\n","        elif return_type == 'regret_reward':\n","            returns = (np.mean(average_regret, axis=0),\n","                       np.mean(all_averages, axis=0))\n","        elif return_type == 'arm_choice_analysis':\n","            returns = (np.mean(all_averages, axis=0),\n","                       np.mean(best_arm),\n","                       np.mean(all_chosen_arm, axis=0))\n","        elif return_type == 'complex':\n","            returns = (np.mean(all_averages, axis=0),\n","                       np.mean(subopt_arm_average, axis=0),\n","                       np.array(best_arm), np.array(worst_arm),\n","                       np.mean(average_regret, axis=0))\n","\n","        return returns"],"metadata":{"id":"zAzjgURdwsl5"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### Run experiments"],"metadata":{"id":"58Lm6nzi3c7a"}},{"cell_type":"code","source":["def run_experiment(environment, agent, environment_parameters, agent_parameters,\n","                   experiment_parameters, save_data=True, dir=''):\n","    rl_glue = RLGlue(environment, agent)\n","\n","    # save sum of reward at the end of each episode\n","    agent_sum_reward = []\n","\n","    env_info = environment_parameters\n","    agent_info = agent_parameters\n","\n","    # one agent setting\n","    for run in tqdm(range(1, experiment_parameters[\"num_runs\"] + 1)):\n","        env_info[\"seed\"] = run\n","\n","        rl_glue.rl_init(agent_info, env_info)\n","        rl_glue.rl_episode(0)\n","        agent_sum_reward.append(rl_glue.average_reward)\n","\n","    leveled_result = get_leveled_data(agent_sum_reward)\n","    if save_data:\n","        save_name = \"{}-{}\".format(rl_glue.agent.name, rl_glue.agent.batch_size)\n","        file_dir = \"results/{}\".format(dir)\n","        if not os.path.exists(file_dir):\n","            os.makedirs(file_dir)\n","        np.save(\"{}/sum_reward_{}\".format(file_dir, save_name), leveled_result)\n","\n","    return leveled_result"],"metadata":{"id":"GWc2CLhz3c2E"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# if __name__ == '__main__':\n","\n","#     num_experements = 10\n","#     batch_size = 100\n","#     data_dir = 'data/mushroom_data_final.pickle'\n","\n","#     experiment_parameters = {\"num_runs\": num_experements}\n","#     env_info = {'pickle_file': data_dir}\n","#     agent_info = {'alpha': 2,\n","#                   'num_actions': 3,\n","#                   'seed': 1,\n","#                   'batch_size': 1}\n","\n","#     agent = LinUCBAgent\n","#     environment = ReplayEnvironment\n","\n","#     result = run_experiment(environment, agent, env_info, agent_info, experiment_parameters, save_data=False)\n","\n","#     smoothed_leveled_result = smooth(result, 100)\n","#     mean_smoothed_leveled_result = np.mean(smoothed_leveled_result, axis=0)\n","\n","#     plt.plot(mean_smoothed_leveled_result, lw=3, ls='-.', label='online policy')\n","#     plt.show()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":297,"referenced_widgets":["87c240ca0fe8431096f5092d00b6d63f","6933623da13c468ea1caba8cb1e126f5","3beaa54b58654d2199153c4229b50f84","317c46945fd64488ad7b2cc0bf30b7d2","fa2e58ea24f84364b95fcc3f1587e9ad","afa78843d54d404ab99cd87d55a36bb9","0c9761efc1a04534871cb6c05f232f4a","21b40574fdb44788af67ece3c7b770c6","0a8e3b93b06740ad8cb9075dbf5d041a","bcaa037ccf1b47bf8d40956a370f93dc","2258f153219f48cc882eda53f0db33a2"]},"id":"oLbyRqGi3miH","executionInfo":{"status":"ok","timestamp":1639148611213,"user_tz":-330,"elapsed":8524,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"d2341afc-3a7d-4ca0-ff62-a1904cad3dc8"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"87c240ca0fe8431096f5092d00b6d63f","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/10 [00:00<?, ?it/s]"]},"metadata":{}},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","source":["## Experiments"],"metadata":{"id":"kD7w8VVTwmV-"}},{"cell_type":"markdown","source":["### UCB"],"metadata":{"id":"S9-ruIHEwmTS"}},{"cell_type":"markdown","source":["#### UCB dynamic by timesteps"],"metadata":{"id":"iSarqXSj5buA"}},{"cell_type":"code","source":["env = Environment\n","agent = UCBAgent\n","\n","alpha = 1\n","\n","num_runs = 1000\n","num_steps = 10000\n","seed = None\n","if_save = False\n","exper_info = {\"num_runs\": num_runs,\n","              \"num_steps\": num_steps,\n","              \"seed\": seed,\n","              \"return_type\": \"regret\"}\n","\n","k = 2\n","arms_values = [0.7, 0.65]\n","reward_type = 'Bernoulli'\n","env_info = {\"num_actions\": k,\n","            \"reward_type\": reward_type,\n","            \"arms_values\": arms_values}\n","\n","# batch-online experiment\n","batch_res = []\n","online_res = []\n","batch = 10\n","agent_info_batch = {\"num_actions\": k, \"batch_size\": batch, \"alpha\": alpha}\n","agent_info_online = {\"num_actions\": k, \"batch_size\": 1, \"alpha\": alpha}\n","\n","exp1 = BanditWrapper(env, agent)\n","batch_res.append(exp1.get_average_performance(agent_info_batch, env_info, exper_info))\n","online_res.append(exp1.get_average_performance(agent_info_online, env_info, exper_info))\n","\n","av_online_res = np.mean(online_res, axis=0)\n","av_batch_res = np.mean(batch_res, axis=0)\n","\n","plt.plot(av_batch_res, label='batch')\n","plt.plot(av_online_res, label='online')\n","\n","M = int(num_steps / batch)\n","update_points = np.ceil(np.arange(num_steps) / batch).astype(int)\n","plt.plot(av_online_res[update_points] * batch, ls='--',\n","         label='upper bound, batch size = 10')\n","plt.title('Cumulative Regret averaged over ' + str(num_runs) + ' runs')\n","plt.xlabel('time steps')\n","plt.ylabel('regret')\n","plt.grid(b=True, which='major', linestyle='--', alpha=0.5)\n","plt.minorticks_on()\n","plt.grid(b=True, which='minor', linestyle=':', alpha=0.2)\n","plt.legend()\n","if if_save:\n","    plt.savefig('results/UCB transform example.png', bbox_inches='tight')\n","plt.show()\n","\n","if if_save:\n","    name = 'batch_result, runs=' + str(num_runs) + ', steps=' + str(num_steps)\n","    with open('results/' + '/' + name + '.pickle', 'wb') as handle:\n","        pickle.dump(batch_res, handle, protocol=pickle.HIGHEST_PROTOCOL)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":359,"referenced_widgets":["6574536e49dd498bae5d87ab60144c59","a787648e15174568a65f7b5e17ee43d2","9fccdb73039f4b649f5389aa2b4099ad","701a04ff58fc492f88774aa7139dcbc2","37296861808045328c2c4eb74625987a","8a9c221fcb4a46f9af525b5b0373a598","f155f2ec83cb416d8d375ae77330165b","af831b75ff7146fc99abb8677c84bdf3","8fe71971bf8049bca4a5c6fde2bf7471","9a87fac6283543008f922a57f98b8514","ea8e8fab825a499c8fe628a021b731ca","d2a7f4cc89a44501b6423cff8f76f77d","9231d7a857624fbfa4c6dc1ba3307814","7dbee5d5ef884477be34e5a9889d13b1","ed0108c515b94090bc7ee284284b535b","ef228c5730d943149399ec81f4de4d46","0dd4ce26eb994bf89f9d429982672f1d","3c78bcaf65d745928c7cec9b3618f2f4","3f9147ea50504b5e941cf600259b64c6","3cb22a9a850b42ce8d72132b34cede77","3929bb10e2044d26869495a947ecd340","cf4c98d2628e418b84ef3c798588675a"]},"id":"Vc7uTIz8xARE","executionInfo":{"status":"ok","timestamp":1639148297436,"user_tz":-330,"elapsed":1433361,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"38fb8767-6cf3-41cc-8cd3-1b9abd760ed6"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"6574536e49dd498bae5d87ab60144c59","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/1000 [00:00<?, ?it/s]"]},"metadata":{}},{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"d2a7f4cc89a44501b6423cff8f76f77d","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/1000 [00:00<?, ?it/s]"]},"metadata":{}},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","source":["#### UCB dynamic by batches"],"metadata":{"id":"82qKEl1F5dMB"}},{"cell_type":"code","source":["model_dir = 'results/UCB/dynamic_by_batches'\n","if not os.path.exists(model_dir):\n","    print(f'Creating a new model directory: {model_dir}')\n","    os.makedirs(model_dir)\n","\n","num_runs = 10 # 500\n","num_steps = 10001\n","seed = None\n","exper_info = {\"num_runs\": num_runs,\n","              \"num_steps\": num_steps,\n","              \"seed\": seed,\n","              \"return_type\": \"regret\"}\n","\n","environments = [[0.7, 0.5], [0.7, 0.4], [0.7, 0.1],\n","                [0.35, 0.18, 0.47, 0.61],\n","                [0.4, 0.75, 0.57, 0.49],\n","                [0.70, 0.50, 0.30, 0.10]]\n","\n","for arms_values in environments:\n","    k = len(arms_values)\n","    reward_type = 'Bernoulli'\n","    env_info = {\"num_actions\": k,\n","                \"reward_type\": reward_type,\n","                \"arms_values\": arms_values}\n","    env = Environment\n","    agent = UCBAgent\n","    alpha = 1\n","\n","    # run online agent\n","    agent_info_online = {\"num_actions\": k, \"batch_size\": 1, \"alpha\": alpha}\n","    experiment = BanditWrapper(env, agent)\n","    online_regret = experiment.get_average_performance(agent_info_online, env_info, exper_info)\n","\n","    # run batch agent\n","    batches = np.logspace(1.0, 3.0, num=20).astype(int)\n","    actual_regret = []\n","    upper_bound = []\n","\n","    for batch in batches:\n","        agent_info_batch = {\"num_actions\": k, \"batch_size\": batch, \"alpha\": alpha}\n","        experiment = BanditWrapper(env, agent)\n","        batch_regret = experiment.get_average_performance(agent_info_batch, env_info, exper_info)\n","        actual_regret.append(batch_regret[-1])\n","        M = int(num_steps / batch)\n","        upper_bound.append(online_regret[M] * batch)\n","\n","    # save data\n","    name = 'dyn_by_batch_' + str(arms_values)\n","    name1 = name + ' batch_regret'\n","    with open(model_dir + '/' + name1 + '.pickle', 'wb') as handle:\n","        pickle.dump(actual_regret, handle, protocol=pickle.HIGHEST_PROTOCOL)\n","\n","    name2 = name + ' online_regret'\n","    with open(model_dir + '/' + name2 + '.pickle', 'wb') as handle:\n","        pickle.dump(online_regret, handle, protocol=pickle.HIGHEST_PROTOCOL)\n","\n","print(\"End!\")"],"metadata":{"id":"L3KoEguL5dIz"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### TS"],"metadata":{"id":"t618mNgBxKKM"}},{"cell_type":"markdown","source":["#### TS dynamic by timesteps"],"metadata":{"id":"oNTgxitH5Dy_"}},{"cell_type":"code","source":["env = Environment\n","agent = TSAgent\n","\n","num_runs = 10 # 1000\n","num_steps = 10000\n","seed = None\n","if_save = False\n","exper_info = {\"num_runs\": num_runs,\n","              \"num_steps\": num_steps,\n","              \"seed\": seed,\n","              \"return_type\": \"regret\"}\n","\n","k = 2\n","arms_values = [0.7, 0.65]\n","reward_type = 'Bernoulli'\n","env_info = {\"num_actions\": k,\n","            \"reward_type\": reward_type,\n","            \"arms_values\": arms_values}\n","\n","# batch-online experiment\n","batch_res = []\n","online_res = []\n","batch = 10\n","agent_info_batch = {\"num_actions\": k, \"batch_size\": batch}\n","agent_info_online = {\"num_actions\": k, \"batch_size\": 1}\n","\n","exp1 = BanditWrapper(env, agent)\n","batch_res.append(exp1.get_average_performance(agent_info_batch, env_info, exper_info))\n","online_res.append(exp1.get_average_performance(agent_info_online, env_info, exper_info))\n","\n","av_online_res = np.mean(online_res, axis=0)\n","av_batch_res = np.mean(batch_res, axis=0)\n","\n","plt.plot(av_batch_res, label='batch')\n","plt.plot(av_online_res, label='online')\n","\n","M = int(num_steps / batch)\n","update_points = np.ceil(np.arange(num_steps) / batch).astype(int)\n","plt.plot(av_online_res[update_points] * batch, ls='--',\n","         label='upper bound, batch size = 10')\n","plt.title('Cumulative Regret averaged over ' + str(num_runs) + ' runs')\n","plt.xlabel('time steps')\n","plt.ylabel('regret')\n","plt.grid(b=True, which='major', linestyle='--', alpha=0.5)\n","plt.minorticks_on()\n","plt.grid(b=True, which='minor', linestyle=':', alpha=0.2)\n","plt.legend()\n","if if_save:\n","    plt.savefig('results/TS  example.png', bbox_inches='tight')\n","plt.show()\n","\n","if if_save:\n","    name = 'batch_result, runs=' + str(num_runs) + ', steps=' + str(num_steps)\n","    with open('results/' + '/' + name + '.pickle', 'wb') as handle:\n","        pickle.dump(batch_res, handle, protocol=pickle.HIGHEST_PROTOCOL)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":359,"referenced_widgets":["959b3e666a064b6d8917c8625ecd8fee","ba3a1e89d28342f683d0013065137cb2","53079098231c4906af72b59e6773052f","3415712fbd16477b8eb7bf6844c56d51","7269023608aa4b9c833466b4634a02cd","23ca025a590c4caba147cfee82b8fb82","6892149d580c4521bfbc8ce341445d61","0767ce00fdc24b709c3ca2169e13e545","c11a3dcd03f24240826793c233950bea","ff5a4612a01b47a8a805ca8a643956c3","f8e2be434a1846c797556912bad038f1","85a8b0c33c104b8eb22fe9e21dc3d543","f8a6e8d21951434b84789ff596bf188b","3825eaf811174f30b38725c44d1536a6","f0fad1d578be4cb790bcbc91e2088d85","367d61cced8a4b7a9eb3763f11856cce","0655f854114a4705ba29eabd0cc99e31","f06c08cde5e14462a35c9b411e3cb167","d769f331f73d42a1ae05b69c07023890","52ec6649d8e94615bf9397cbd9d85b48","fa0bd788f3a8421d949951e60f790c07","2464f2977ba04e8a855461a081005305"]},"id":"9POfXn2TxLWQ","executionInfo":{"status":"ok","timestamp":1639148363702,"user_tz":-330,"elapsed":55499,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"15423772-d107-4547-f6fa-59c4a54546e3"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"959b3e666a064b6d8917c8625ecd8fee","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/10 [00:00<?, ?it/s]"]},"metadata":{}},{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"85a8b0c33c104b8eb22fe9e21dc3d543","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/10 [00:00<?, ?it/s]"]},"metadata":{}},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","source":["#### TS dynamic by batches"],"metadata":{"id":"koXASf5O5GR2"}},{"cell_type":"code","source":["model_dir = 'results/TS/dynamic_by_batches'\n","if not os.path.exists(model_dir):\n","    print(f'Creating a new model directory: {model_dir}')\n","    os.makedirs(model_dir)\n","\n","num_runs = 10 # 500\n","num_steps = 10001\n","seed = None\n","exper_info = {\"num_runs\": num_runs,\n","              \"num_steps\": num_steps,\n","              \"seed\": seed,\n","              \"return_type\": \"regret\"}\n","\n","environments = [[0.7, 0.5], [0.7, 0.4], [0.7, 0.1],\n","                [0.35, 0.18, 0.47, 0.61],\n","                [0.4, 0.75, 0.57, 0.49],\n","                [0.70, 0.50, 0.30, 0.10]]\n","\n","for arms_values in environments:\n","    k = len(arms_values)\n","    reward_type = 'Bernoulli'\n","    env_info = {\"num_actions\": k,\n","                \"reward_type\": reward_type,\n","                \"arms_values\": arms_values}\n","    env = Environment\n","    agent = TSAgent\n","\n","    # run online agent\n","    agent_info_online = {\"num_actions\": k, \"batch_size\": 1}\n","    experiment = BanditWrapper(env, agent)\n","    online_regret = experiment.get_average_performance(agent_info_online, env_info, exper_info)\n","\n","    # run batch agent\n","    batches = np.logspace(1.0, 3.0, num=20).astype(int)\n","    actual_regret = []\n","    upper_bound = []\n","\n","    for batch in batches:\n","        agent_info_batch = {\"num_actions\": k, \"batch_size\": batch}\n","        experiment = BanditWrapper(env, agent)\n","        batch_regret = experiment.get_average_performance(agent_info_batch, env_info, exper_info)\n","        actual_regret.append(batch_regret[-1])\n","        M = int(num_steps / batch)\n","        upper_bound.append(online_regret[M] * batch)\n","\n","    # save data\n","    name = 'dyn_by_batch_' + str(k) + str(arms_values)\n","    name1 = name + ' batch_regret'\n","    with open(model_dir + '/' + name1 + '.pickle', 'wb') as handle:\n","        pickle.dump(actual_regret, handle, protocol=pickle.HIGHEST_PROTOCOL)\n","\n","    name2 = name + ' online_regret'\n","    with open(model_dir + '/' + name2 + '.pickle', 'wb') as handle:\n","        pickle.dump(online_regret, handle, protocol=pickle.HIGHEST_PROTOCOL)\n","\n","print(\"End!\")"],"metadata":{"id":"T-CjxemW5GNq"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### LinUCB"],"metadata":{"id":"L-4KgfMU0iE9"}},{"cell_type":"markdown","source":["#### LinUCB by timesteps"],"metadata":{"id":"elLqw6xk4083"}},{"cell_type":"code","source":["num_experiments = 20\n","batch_size = 100\n","data_dir = 'data/mushroom_data_final.pickle'\n","env_info = {'pickle_file': data_dir}\n","output_dir = 'LinUCB/dynamic_by_timesteps'\n","\n","agent_info = {'alpha': 2,\n","              'num_actions': 3,\n","              'seed': 1,\n","              'batch_size': 1}\n","agent_info_batch = {'alpha': 2,\n","                    'num_actions': 3,\n","                    'seed': 1,\n","                    'batch_size': batch_size}\n","experiment_parameters = {\"num_runs\": num_experiments}\n","\n","agent = LinUCBAgent\n","environment = ReplayEnvironment\n","\n","online_result = run_experiment(environment, agent, env_info, agent_info,\n","                               experiment_parameters, True, output_dir)\n","batch_result = run_experiment(environment, agent, env_info, agent_info_batch,\n","                              experiment_parameters, True, output_dir)\n","\n","smoothed_leveled_result = smooth(online_result, 100)\n","smoothed_leveled_result1 = smooth(batch_result, 100)\n","\n","mean_smoothed_leveled_result = np.mean(smoothed_leveled_result, axis=0)\n","mean_smoothed_leveled_result1 = np.mean(smoothed_leveled_result1, axis=0)\n","\n","num_steps = np.minimum(len(mean_smoothed_leveled_result), len(mean_smoothed_leveled_result1))\n","update_points = np.ceil(np.arange(num_steps) / batch_size).astype(int)\n","\n","pic_filename = \"results/{}/UCB_transform_timesteps.png\".format(output_dir)\n","plt.plot(mean_smoothed_leveled_result1, lw=3, label='batch, batch size = ' + str(batch_size))\n","plt.plot(mean_smoothed_leveled_result, lw=3, ls='-.', label='online policy')\n","plt.plot(mean_smoothed_leveled_result[update_points], lw=3, ls='-.', label='dumb policy')\n","plt.legend()\n","plt.xlabel('time steps')\n","plt.title(\"Smooth Cumulative Reward averaged over {} runs\".format(num_experiments))\n","plt.ylabel('smoothed reward')\n","plt.grid(b=True, which='major', linestyle='--', alpha=0.5)\n","plt.minorticks_on()\n","plt.grid(b=True, which='minor', linestyle=':', alpha=0.2)\n","plt.savefig(pic_filename, bbox_inches='tight')\n","plt.show()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":359,"referenced_widgets":["07d74cc8c1034e8e88e0ba68681e7d83","20755197a75045769c76c25cf9997309","3aaecc6ab5334dbaaff91f3fd3d2dc92","ccdfd084e55f4d48aa8d62ff900c54dc","602801fccd6d4538bd22a5494c8a538e","81b219b6e4414ff888646a18f038c6d2","4622f6b8c7914be18ec086e119b14510","c603ecbfec974667805abf5f8366dde5","508403ada600443f97467dde6fb0dd3d","9840772f43134cee8eabacd2509f9fd7","4e96a7f4873f4360b2118e7a3ddbfc48","c19256a397ea4e79b3a0cbf5a424a3ea","4d1f5676bbaa4b3d8b81cc7be3cf6da3","eae2220f0119401a8759637bcf0e9f4c","4aa44dd37d5243dfa24fea8f3dc6839b","2deab0a58cd04226b834a99ee4ff4c90","16acd859901c4b8aa5f6b4473b74e249","e4edb80a747a4e9684a556ff96563e7f","3d74bbdd72dc4e0aadb4bd517358bbe7","809de5d470244a2b90daffd0477c28a4","4e13cae2ba7a44daa51794431d7af0c0","2780d1d40a0644c498e860cf4839ac25"]},"id":"tXey4Vyi06pm","executionInfo":{"status":"ok","timestamp":1639148649109,"user_tz":-330,"elapsed":32385,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"77e181e9-b27e-474c-834d-4db5a099aeb1"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"07d74cc8c1034e8e88e0ba68681e7d83","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/20 [00:00<?, ?it/s]"]},"metadata":{}},{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"c19256a397ea4e79b3a0cbf5a424a3ea","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/20 [00:00<?, ?it/s]"]},"metadata":{}},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","source":["#### LinUCB by batches"],"metadata":{"id":"q7xl1u_F45d5"}},{"cell_type":"code","source":["num_experiments = 20\n","data_dir = 'data/mushroom_data_final.pickle'\n","env_info = {'pickle_file': data_dir}\n","output_dir = 'LinUCB/dynamic_by_batches'\n","\n","agent_info = {'alpha': 2,\n","              'num_actions': 3,\n","              'seed': 1,\n","              'batch_size': 1}\n","experiment_parameters = {\"num_runs\": num_experiments}\n","\n","agent = LinUCBAgent\n","environment = ReplayEnvironment\n","\n","# run online agent\n","online_result = run_experiment(environment, agent, env_info, agent_info,\n","                               experiment_parameters, True, output_dir)\n","# smooth and average the result\n","smoothed_leveled_result = smooth(online_result, 100)\n","mean_smoothed_leveled_result = np.mean(smoothed_leveled_result, axis=0)\n","mean_smoothed_leveled_result = mean_smoothed_leveled_result[~np.isnan(mean_smoothed_leveled_result)]\n","\n","# run batch agent\n","batch_sizes = np.logspace(1.0, 2.7, num=20).astype(int)\n","actual_regret = []\n","upper_bound = []\n","for batch in batch_sizes:\n","    agent_info_batch = {'alpha': 2,\n","                        'num_actions': 3,\n","                        'seed': 1,\n","                        'batch_size': batch}\n","    batch_result = run_experiment(environment, agent, env_info, agent_info_batch,\n","                                  experiment_parameters, True, output_dir)\n","    # smooth and average the result\n","    smoothed_leveled_result1 = smooth(batch_result, 100)\n","    mean_smoothed_leveled_result1 = np.mean(smoothed_leveled_result1, axis=0)\n","    mean_smoothed_leveled_result1 = mean_smoothed_leveled_result1[~np.isnan(mean_smoothed_leveled_result1)]\n","\n","    actual_regret.append(mean_smoothed_leveled_result1[-1])\n","\n","    # fetch dumb result\n","    M = int(len(mean_smoothed_leveled_result1) / batch)\n","    upper_bound.append(mean_smoothed_leveled_result[M])\n","\n","pic_filename = \"results/{}/UCB_transform_batchsize.png\".format(output_dir)\n","plt.plot(batch_sizes, actual_regret, label='actual regret')\n","plt.plot(batch_sizes, [mean_smoothed_leveled_result[-1]]*len(batch_sizes), label='online policy')\n","plt.plot(batch_sizes, upper_bound, label='dumb policy')\n","plt.legend()\n","plt.title(\"Reward as a f-n of batch size (each point is averaged over {} runs)\".format(num_experiments))\n","plt.xlabel('batch size (log scale)')\n","plt.ylabel('reward')\n","plt.grid(b=True, which='major', linestyle='--', alpha=0.5)\n","plt.minorticks_on()\n","plt.grid(b=True, which='minor', linestyle=':', alpha=0.2)\n","plt.savefig(pic_filename, bbox_inches='tight')\n","plt.show()"],"metadata":{"id":"XxBGmxib45aV"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### LinTS"],"metadata":{"id":"zaZs-8O31BHz"}},{"cell_type":"markdown","source":["#### LinTS by timesteps"],"metadata":{"id":"apXIaCEY4nih"}},{"cell_type":"code","source":["num_experiments = 10\n","batch_size = 100\n","data_dir = 'data/mushroom_data_final.pickle'\n","env_info = {'pickle_file': data_dir}\n","output_dir = 'LinTS/dynamic_by_timesteps'\n","\n","agent_info = {'alpha': 1,\n","              'num_actions': 3,\n","              'seed': 1,\n","              'batch_size': 1,\n","              'replay_buffer_size': 100000}\n","agent_info_batch = {'alpha': 1,\n","                    'num_actions': 3,\n","                    'seed': 1,\n","                    'batch_size': batch_size,\n","                    'replay_buffer_size': 100000}\n","experiment_parameters = {\"num_runs\": num_experiments}\n","\n","agent = LinTSAgent\n","environment = ReplayEnvironment\n","\n","online_result = run_experiment(environment, agent, env_info, agent_info,\n","                               experiment_parameters, True, output_dir)\n","batch_result = run_experiment(environment, agent, env_info, agent_info_batch,\n","                              experiment_parameters, True, output_dir)\n","\n","smoothed_leveled_result = smooth(online_result, 100)\n","smoothed_leveled_result1 = smooth(batch_result, 100)\n","\n","mean_smoothed_leveled_result = np.mean(smoothed_leveled_result, axis=0)\n","mean_smoothed_leveled_result1 = np.mean(smoothed_leveled_result1, axis=0)\n","\n","num_steps = np.minimum(len(mean_smoothed_leveled_result), len(mean_smoothed_leveled_result1))\n","update_points = np.ceil(np.arange(num_steps) / batch_size).astype(int)\n","\n","pic_filename = \"results/{}/TS_transform_timesteps.png\".format(output_dir)\n","plt.plot(mean_smoothed_leveled_result1, lw=3, label='batch, batch size = ' + str(batch_size))\n","plt.plot(mean_smoothed_leveled_result, lw=3, ls='-.', label='online policy')\n","plt.plot(mean_smoothed_leveled_result[update_points], lw=3, ls='-.', label='dumb policy')\n","plt.legend()\n","plt.xlabel('time steps')\n","plt.title(\"Smooth Cumulative Reward averaged over {} runs\".format(num_experiments))\n","plt.ylabel('smoothed reward')\n","plt.grid(b=True, which='major', linestyle='--', alpha=0.5)\n","plt.minorticks_on()\n","plt.grid(b=True, which='minor', linestyle=':', alpha=0.2)\n","plt.savefig(pic_filename, bbox_inches='tight')\n","plt.show()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":359,"referenced_widgets":["a2c68a56898548f8a7e190dcf2d034f4","e009631910e042a2a07afe8cf0e6d118","a850623290b14d07838e4203a29f9ba2","9f140b090cb44361b39a6e09205c087c","36fc4560cb99439c945c620138005a1b","638ee1bca5d64cc7966556e3fb64c06e","a454e1e1460046e3a98a4b28c48ecda0","d0a15e2054af430bad52b6fc28fd0994","f9295dd431fe4587a816d6c59f464399","41ca699547f14bda8a83777d78e6bc86","c29bc574f0cd46309d8eb6ec40378025","858e460d943744248a4109a5c328d77c","700c0953693b48bea968c6c67964684d","c70158af000c40e59b588e745be4a2dd","21b1b038a25f469295ea8e2b75c470ae","c20bc14a583e452ba263c4e4f4f3147e","e1199401ceee4d0ca1aaa72020618241","57b794458691470ba34efd35f0cd8a9c","4bd9a0cc11664bf4870c7fc46d151bda","301a50e882864065b9108e0a0cf0edde","93f8c74e75504aadbf1cdec1c5b6a43f","a4d3c862e74242e4beeaaa880f4e3fc8"]},"id":"6uyZk7PT1Cjy","executionInfo":{"status":"ok","timestamp":1639149620282,"user_tz":-330,"elapsed":952413,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"72dd4d89-25ac-408a-c033-d05d6bee7628"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"a2c68a56898548f8a7e190dcf2d034f4","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/10 [00:00<?, ?it/s]"]},"metadata":{}},{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"858e460d943744248a4109a5c328d77c","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/10 [00:00<?, ?it/s]"]},"metadata":{}},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","source":["#### LinTS by batches"],"metadata":{"id":"pVbE43Lm4p5H"}},{"cell_type":"code","source":["num_experiments = 20\n","data_dir = 'data/mushroom_data_final.pickle'\n","env_info = {'pickle_file': data_dir}\n","output_dir = 'LinTS/dynamic_by_batches'\n","\n","agent_info = {'alpha': 1,\n","              'num_actions': 3,\n","              'seed': 1,\n","              'batch_size': 1,\n","              'replay_buffer_size': 100000}\n","experiment_parameters = {\"num_runs\": num_experiments}\n","\n","agent = LinTSAgent\n","environment = ReplayEnvironment\n","\n","# run online agent\n","online_result = run_experiment(environment, agent, env_info, agent_info,\n","                               experiment_parameters, True, output_dir)\n","# smooth and average the result\n","smoothed_leveled_result = smooth(online_result, 100)\n","mean_smoothed_leveled_result = np.mean(smoothed_leveled_result, axis=0)\n","mean_smoothed_leveled_result = mean_smoothed_leveled_result[~np.isnan(mean_smoothed_leveled_result)]\n","\n","# run batch agent\n","batch_sizes = np.logspace(1.0, 2.7, num=20).astype(int)\n","actual_regret = []\n","upper_bound = []\n","for batch in batch_sizes:\n","    agent_info_batch = {'alpha': 1,\n","                        'num_actions': 3,\n","                        'seed': 1,\n","                        'batch_size': batch,\n","                        'replay_buffer_size': 100000}\n","    batch_result = run_experiment(environment, agent, env_info, agent_info_batch,\n","                                  experiment_parameters, True, output_dir)\n","    # smooth and average the result\n","    smoothed_leveled_result1 = smooth(batch_result, 100)\n","    mean_smoothed_leveled_result1 = np.mean(smoothed_leveled_result1, axis=0)\n","    mean_smoothed_leveled_result1 = mean_smoothed_leveled_result1[~np.isnan(mean_smoothed_leveled_result1)]\n","\n","    actual_regret.append(mean_smoothed_leveled_result1[-1])\n","\n","    # fetch dumb result\n","    M = int(len(mean_smoothed_leveled_result1) / batch)\n","    upper_bound.append(mean_smoothed_leveled_result[M])\n","\n","pic_filename = \"results/{}/TS_transform_batchsize.png\".format(output_dir)\n","plt.plot(batch_sizes, actual_regret, label='actual regret')\n","plt.plot(batch_sizes, [mean_smoothed_leveled_result[-1]]*len(batch_sizes), label='online policy')\n","plt.plot(batch_sizes, upper_bound, label='dumb policy')\n","plt.legend()\n","plt.title(\"Reward as a f-n of batch size (each point is averaged over {} runs)\".format(num_experiments))\n","plt.xlabel('batch size (log scale)')\n","plt.ylabel('reward')\n","plt.grid(b=True, which='major', linestyle='--', alpha=0.5)\n","plt.minorticks_on()\n","plt.grid(b=True, which='minor', linestyle=':', alpha=0.2)\n","plt.savefig(pic_filename, bbox_inches='tight')\n","plt.show()"],"metadata":{"id":"K63E-61k4p03"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["### CMAB demo on Mushroom dataset"],"metadata":{"id":"w5S8ABzm1JMM"}},{"cell_type":"code","source":["data_dir = 'data/mushroom_data_final.pickle'\n","env_info = {'pickle_file': data_dir,\n","            'seed': 1}\n","# init env\n","environment = ReplayEnvironment\n","\n","# init random agent\n","random_agent_info = {'num_actions': 2}\n","ra = RandomAgent()\n","ra.agent_init(random_agent_info)\n","\n","# learn LinUCB agent\n","agent_info = {'alpha': 2,\n","              'num_actions': 2,\n","              'seed': 1,\n","              'batch_size': 1}\n","\n","agent = LinUCBAgent\n","rl_glue = RLGlue(environment, agent)\n","\n","for i in range(4):    \n","    rl_glue.rl_init(agent_info, env_info)\n","    rl_glue.rl_episode(0)\n","UCB_agent = rl_glue.agent\n","\n","# learn LinTS agent\n","agent_info = {'num_actions': 2,\n","              'replay_buffer_size': 200,\n","              'seed': 1,\n","              'batch_size': 1}\n","agent = LinTSAgent\n","rl_glue = RLGlue(environment, agent)\n","\n","for i in range(4):    \n","    rl_glue.rl_init(agent_info, env_info)\n","    rl_glue.rl_episode(0)\n","\n","TS_agent = rl_glue.agent\n","result = []\n","result1 = []\n","result2 = []\n","\n","exper_seeds = [2, 5, 10, 12, 54, 32, 15, 76, 45, 56]\n","for seed_ in exper_seeds:\n","    dataset = BanditDataset(pickle_file=data_dir, seed=seed_)\n","\n","    eval_info = {'dataset': dataset, 'agent': UCB_agent}\n","    eval_info1 = {'dataset': dataset, 'agent': TS_agent}\n","    eval_info2 = {'dataset': dataset, 'agent': ra}\n","\n","    evaluator = OfflineEvaluator(eval_info)\n","    evaluator1 = OfflineEvaluator(eval_info1)\n","    evaluator2 = OfflineEvaluator(eval_info2)\n","\n","    reward = evaluator.eval_run()\n","    reward1 = evaluator1.eval_run()\n","    reward2 = evaluator2.eval_run()\n","\n","    result.append(reward)\n","    result1.append(reward1)\n","    result2.append(reward2)\n","\n","labels = ['UCB agent', 'TS agent', 'Random agent']\n","for i, res in enumerate([result, result1, result2]):\n","    for elem in res:\n","        plt.plot(elem, linewidth=0.1)\n","    avg = [float(sum(col))/len(col) for col in zip(*res)]\n","    plt.plot(avg, label=labels[i])"],"metadata":{"id":"pHKrIm_Q1R7S"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["labels = ['UCB agent', 'TS agent', 'Random agent']\n","for i, res in enumerate([result, result1, result2]):\n","    for elem in res:\n","        plt.plot(elem, linewidth=0.1)\n","    avg = [float(sum(col))/len(col) for col in zip(*res)]\n","    plt.plot(avg, label=labels[i])\n","plt.legend()\n","plt.ylim([0.1, 0.7])\n","plt.grid(b=True, which='major', linestyle='--', alpha=0.5)\n","plt.minorticks_on()\n","plt.grid(b=True, which='minor', linestyle=':', alpha=0.2)\n","plt.show()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":269},"id":"ZbxPyHo_1Zgo","executionInfo":{"status":"ok","timestamp":1639152031963,"user_tz":-330,"elapsed":1780,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"0b963035-788e-4b43-b73d-9ee1df8e7cc7"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","source":["### CMAB demo on Simulated dataset"],"metadata":{"id":"o_S529BK1qop"}},{"cell_type":"code","source":["# generate 100 000 samples with 4 features and 3 actions\n","dataset = generate_samples(100000, 4, 3, True)\n","dataset.head()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":206},"id":"8OOtls5G1qlT","executionInfo":{"status":"ok","timestamp":1639149837227,"user_tz":-330,"elapsed":4833,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"597535e2-9ebb-4bce-a443-6c90344656dc"},"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>X_1</th>\n","      <th>X_2</th>\n","      <th>X_3</th>\n","      <th>X_4</th>\n","      <th>a</th>\n","      <th>y</th>\n","      <th>probs</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0</th>\n","      <td>1</td>\n","      <td>3</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.266661</td>\n","    </tr>\n","    <tr>\n","      <th>1</th>\n","      <td>3</td>\n","      <td>1</td>\n","      <td>3</td>\n","      <td>1</td>\n","      <td>1.0</td>\n","      <td>0.0</td>\n","      <td>0.236514</td>\n","    </tr>\n","    <tr>\n","      <th>2</th>\n","      <td>3</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>1</td>\n","      <td>1.0</td>\n","      <td>0.0</td>\n","      <td>0.236514</td>\n","    </tr>\n","    <tr>\n","      <th>3</th>\n","      <td>0</td>\n","      <td>3</td>\n","      <td>1</td>\n","      <td>0</td>\n","      <td>1.0</td>\n","      <td>0.0</td>\n","      <td>0.236514</td>\n","    </tr>\n","    <tr>\n","      <th>4</th>\n","      <td>2</td>\n","      <td>1</td>\n","      <td>2</td>\n","      <td>0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.266661</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["   X_1  X_2  X_3  X_4    a    y     probs\n","0    1    3    0    0  0.0  0.0  0.266661\n","1    3    1    3    1  1.0  0.0  0.236514\n","2    3    0    0    1  1.0  0.0  0.236514\n","3    0    3    1    0  1.0  0.0  0.236514\n","4    2    1    2    0  0.0  0.0  0.266661"]},"metadata":{},"execution_count":71}]},{"cell_type":"markdown","source":["#### LinTS dynamic by steps"],"metadata":{"id":"h9x6zOMx1qdr"}},{"cell_type":"code","source":["num_experiments = 10\n","batch_size1 = 30\n","batch_size2 = 100\n","env_info = {'pickle_file': dataset}\n","\n","agent1_info = {'alpha': 1,\n","              'num_actions': 3,\n","              'seed': 1,\n","              'batch_size': batch_size1,\n","              'replay_buffer_size': 100000}\n","agent2_info = {'alpha': 1,\n","                    'num_actions': 3,\n","                    'seed': 1,\n","                    'batch_size': batch_size2,\n","                    'replay_buffer_size': 100000}\n","experiment_parameters = {\"num_runs\": num_experiments}"],"metadata":{"id":"SDmmh_Js1qag"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["agent = LinTSAgent\n","environment = ReplayEnvironment\n","\n","result1 = run_experiment(environment, agent, env_info, agent1_info, experiment_parameters, False)\n","result2 = run_experiment(environment, agent, env_info, agent2_info, experiment_parameters, False)\n","\n","smoothed_leveled_result1 = smooth(result1, 100)\n","smoothed_leveled_result2 = smooth(result2, 100)\n","\n","mean_smoothed_leveled_result1 = np.mean(smoothed_leveled_result1, axis=0)\n","mean_smoothed_leveled_result2 = np.mean(smoothed_leveled_result2, axis=0)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":81,"referenced_widgets":["79e9b193d9284e4b898cb4344db92ef3","8ea51857d7cf48c1beddb0a9bd3b764b","7b0e4960186b4307ae594b269043c16d","aa39361015a84c158e922f8e0b8dca9a","63acbf6a68e6455c91cbea40169d1ee9","e2c2139f96b345768e075ef3593fac06","3dcbffa6176e4824b9a48cccfaf024cc","740c9d9905b542d4a0edef78f1d8ca30","cd2dc9d9c7504cd4b118a33b7c24f683","33707c49e4fd42ad932dd74e59f4a060","60745768aedc4942bad8883d708b0e25","48077097754d4b96b1345fb1c06560b8","760943cea30f4236a0715b61a047fd99","3bf9f4be47424b13b1db937f67894360","28c1b766f56c41179840f2140da35b44","487beca1cfc847c0a41a414968097b6b","c28b5aecd2eb4353b99534e8eef2380c","f167c63d8ac54b07b31b261c77a98391","4a8e550188ac481fa2838df0996d3b0b","50dfc1413ebd4298af9fc979de4ebb63","f36eb7ce107741159cfc86169724580f","5b4edf526246433d8cca04eba8be3229"]},"id":"Sp9wGAPS1-V-","executionInfo":{"status":"ok","timestamp":1639150020202,"user_tz":-330,"elapsed":182996,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"b1fc5a37-fb85-44dc-cb62-57e3a9813b63"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"79e9b193d9284e4b898cb4344db92ef3","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/10 [00:00<?, ?it/s]"]},"metadata":{}},{"output_type":"display_data","data":{"application/vnd.jupyter.widget-view+json":{"model_id":"48077097754d4b96b1345fb1c06560b8","version_minor":0,"version_major":2},"text/plain":["  0%|          | 0/10 [00:00<?, ?it/s]"]},"metadata":{}}]},{"cell_type":"code","source":["plt.plot(mean_smoothed_leveled_result1, label='batch size = ' + str(batch_size1))\n","plt.plot(mean_smoothed_leveled_result2, label='batch size = ' + str(batch_size2))\n","plt.legend()\n","plt.xlabel('time steps')\n","plt.title(\"Smooth Cumulative Reward averaged over {} runs\".format(num_experiments))\n","plt.ylabel('smoothed conversion rate')\n","plt.grid(b=True, which='major', linestyle='--', alpha=0.5)\n","plt.minorticks_on()\n","plt.grid(b=True, which='minor', linestyle=':', alpha=0.2)\n","plt.show()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":295},"id":"t2CezR4D2AeW","executionInfo":{"status":"ok","timestamp":1639150021342,"user_tz":-330,"elapsed":1192,"user":{"displayName":"Sparsh Agarwal","photoUrl":"https://lh3.googleusercontent.com/a/default-user=s64","userId":"13037694610922482904"}},"outputId":"dedc8712-71e2-4c3e-f805-6c5902a2de89"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","source":["#### LinTS dynamic by batches"],"metadata":{"id":"B3W58Zqv2CnH"}},{"cell_type":"code","source":["num_experiments = 20\n","env_info = {'pickle_file': dataset}\n","experiment_parameters = {\"num_runs\": num_experiments}\n","\n","agent = LinTSAgent\n","environment = ReplayEnvironment\n","\n","# run batch agent\n","batch_sizes = np.logspace(1.0, 2.7, num=20).astype(int)\n","actual_regret = []\n","for batch in batch_sizes:\n","    agent_info_batch = {'alpha': 1,\n","                        'num_actions': 3,\n","                        'seed': 1,\n","                        'batch_size': batch,\n","                        'replay_buffer_size': 100000}\n","    batch_result = run_experiment(environment, agent, env_info, agent_info_batch,\n","                                  experiment_parameters, False)\n","    # smooth and average the result\n","    smoothed_leveled_result1 = smooth(batch_result, 100)\n","    mean_smoothed_leveled_result1 = np.mean(smoothed_leveled_result1, axis=0)\n","    mean_smoothed_leveled_result1 = mean_smoothed_leveled_result1[~np.isnan(mean_smoothed_leveled_result1)]\n","\n","    actual_regret.append(mean_smoothed_leveled_result1[-1])"],"metadata":{"id":"Vs_O6HE92Ckv"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["plt.plot(batch_sizes, actual_regret, label='actual regret')\n","plt.legend()\n","plt.title(\"Reward as a f-n of batch size (each point is averaged over {} runs)\".format(num_experiments))\n","plt.xlabel('batch size (log scale)')\n","plt.ylabel('reward')\n","plt.grid(b=True, which='major', linestyle='--', alpha=0.5)\n","plt.minorticks_on()\n","plt.grid(b=True, which='minor', linestyle=':', alpha=0.2)\n","plt.show()"],"metadata":{"id":"4iErUrNt2CiB"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":[""],"metadata":{"id":"ajBoPT1U2CfL"},"execution_count":null,"outputs":[]}]}