unlocking-magnus-carelson-chess-playing-strategies.ipynb

{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.10.14","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"none","dataSources":[{"sourceId":83735,"databundleVersionId":9881586,"sourceType":"competition"},{"sourceId":10070193,"sourceType":"datasetVersion","datasetId":6206788}],"dockerImageVersionId":30786,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":false}},"nbformat_minor":4,"nbformat":4,"cells":[{"source":"<a href=\"https://www.kaggle.com/code/datascientistsalman/unlocking-magnus-carelson-chess-playing-strategies?scriptVersionId=210700596\" target=\"_blank\"><img align=\"left\" alt=\"Kaggle\" title=\"Open in Kaggle\" src=\"https://kaggle.com/static/images/open-in-kaggle.svg\"></a>","metadata":{},"cell_type":"markdown"},{"cell_type":"markdown","source":"**Created by:** Salman Ahmed  \n**LinkedIn:** [Salman Ahmed ](https://www.linkedin.com/in/salmanahmed84/)  \n**Portfolio:** [Salman's Portfolio](https://github.com/salman-datascientist)","metadata":{"execution":{"iopub.status.busy":"2024-12-01T18:35:21.893362Z","iopub.execute_input":"2024-12-01T18:35:21.893771Z","iopub.status.idle":"2024-12-01T18:35:21.900702Z","shell.execute_reply.started":"2024-12-01T18:35:21.893735Z","shell.execute_reply":"2024-12-01T18:35:21.899313Z"}}},{"cell_type":"markdown","source":"**Gemini 1.5 with Long-Context Windows for Analyzing Hours-Long Magnus Carlsen Chess Game Videos**\n\n***Extended Context Length:***\nVast Input Capacity: Gemini 1.5 Pro excels at processing hours-long chess game videos, managing inputs of up to 10 million tokens. This allows for seamless analysis of entire tournaments without segmenting the data into smaller chunks​.\nImproved Contextual Understanding: The extended context length enables the model to connect strategic decisions across games, identify long-term patterns, and correlate player behavior over time.\n\n***High Recall:***\nAccurate Information Retrieval: When analyzing hours of video, Gemini 1.5 Pro can precisely locate specific moves, strategies, or key moments, such as a player executing a checkmate or pivotal errors in gameplay. This high recall ensures no critical detail is missed​.\nSuperior Performance: Compared to other models, Gemini 1.5 Pro demonstrates unmatched accuracy in retrieving and reasoning over long video sequences, even outperforming specialized models like GPT-4V for video QA.\nMultimodal Capabilities:\n\n***Unified Understanding:***\nGemini 1.5 Pro integrates video, audio commentary, and text-based annotations in chess games to create a comprehensive analysis of gameplay strategies and decisions.\nEnhanced Contextual Awareness: By combining visual inputs (e.g., board state), audio (e.g., player commentary), and text (e.g., move notations), the model delivers richer insights into player strategies and decision-making processes.\nEfficiency and Scalability:\n\n***Optimized Processing:*** \nDesigned to handle extensive data, Gemini 1.5 Pro efficiently processes multiple hours of chess gameplay videos, ensuring that even tournaments spanning several days can be analyzed in one session.\nScalability: Its ability to analyze entire archives of chess games, such as championship series, makes it ideal for researchers, commentators, or AI systems studying gameplay evolution and strategy trends​.\nReduced Reliance on External Tools:\n\n***Self-Sufficiency:***\nUnlike systems that rely on external video processing or annotation tools, Gemini 1.5 Pro can directly analyze video content, extracting timestamps, moves, and strategic insights without additional preprocessing​.\n\n***Simplified Workflow:***\nThis all-in-one capability simplifies the workflow, enabling faster and more efficient analysis of chess games without requiring integration with separate retrieval or indexing systems​.","metadata":{}},{"cell_type":"markdown","source":"Install the SDK","metadata":{}},{"cell_type":"code","source":"%pip install -U -q \"google-generativeai>=0.8.3\"","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:54:12.375178Z","iopub.execute_input":"2024-12-01T22:54:12.375556Z","iopub.status.idle":"2024-12-01T22:54:45.499294Z","shell.execute_reply.started":"2024-12-01T22:54:12.375524Z","shell.execute_reply":"2024-12-01T22:54:45.497569Z"}},"outputs":[{"name":"stdout","text":"Note: you may need to restart the kernel to use updated packages.\n","output_type":"stream"}],"execution_count":1},{"cell_type":"markdown","source":"<center>\n  <h2 style=\"color: blue;\">YouTube Video: <a href=\"https://www.youtube.com/watch?v=RkBgf06G89E\" target=\"_blank\" style=\"color: blue;\">Magnus Carlsen goes 94% BERSERK in Blitz Titled Arena</a></h2>\n</center>","metadata":{"execution":{"iopub.status.busy":"2024-12-01T18:36:13.844311Z","iopub.execute_input":"2024-12-01T18:36:13.844735Z","iopub.status.idle":"2024-12-01T18:36:13.852419Z","shell.execute_reply.started":"2024-12-01T18:36:13.844698Z","shell.execute_reply":"2024-12-01T18:36:13.850699Z"}}},{"cell_type":"code","source":"from IPython.display import HTML\n\n# Display a centered YouTube video\nvideo_id = \"RkBgf06G89E\"\nHTML(f\"\"\"\n<div style=\"display: flex; justify-content: center;\">\n    <iframe width=\"1024\" height=\"576\" src=\"https://www.youtube.com/embed/{video_id}\" frameborder=\"0\" allowfullscreen></iframe>\n</div>\n\"\"\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:54:45.501983Z","iopub.execute_input":"2024-12-01T22:54:45.502532Z","iopub.status.idle":"2024-12-01T22:54:45.514086Z","shell.execute_reply.started":"2024-12-01T22:54:45.502474Z","shell.execute_reply":"2024-12-01T22:54:45.512675Z"}},"outputs":[{"execution_count":2,"output_type":"execute_result","data":{"text/plain":"<IPython.core.display.HTML object>","text/html":"\n<div style=\"display: flex; justify-content: center;\">\n    <iframe width=\"1024\" height=\"576\" src=\"https://www.youtube.com/embed/RkBgf06G89E\" frameborder=\"0\" allowfullscreen></iframe>\n</div>\n"},"metadata":{}}],"execution_count":2},{"cell_type":"code","source":"# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the read-only \"../input/\" directory\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        print(os.path.join(dirname, filename))\n\n# You can write up to 20GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:54:45.515533Z","iopub.execute_input":"2024-12-01T22:54:45.515937Z","iopub.status.idle":"2024-12-01T22:54:45.949015Z","shell.execute_reply.started":"2024-12-01T22:54:45.515903Z","shell.execute_reply":"2024-12-01T22:54:45.947847Z"}},"outputs":[{"name":"stdout","text":"/kaggle/input/magnus-chess-blitz/magnus_chess_blitz.mp4\n/kaggle/input/chess-guides/Chess_ The Complete Guide To Chess - Master_ Chess Tactics Chess Openings and Chess Strategies.pdf\n/kaggle/input/chess-guides/1 Chess Game Instructions.pdf\n/kaggle/input/gemini-long-context/submission_instructions.txt\n","output_type":"stream"}],"execution_count":3},{"cell_type":"code","source":"import os\nimport time\nimport datetime\nimport google.generativeai as genai\nfrom google.generativeai import caching\nfrom IPython.display import HTML, Markdown, display, Video","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:54:45.951225Z","iopub.execute_input":"2024-12-01T22:54:45.951695Z","iopub.status.idle":"2024-12-01T22:54:47.336261Z","shell.execute_reply.started":"2024-12-01T22:54:45.951654Z","shell.execute_reply":"2024-12-01T22:54:47.335102Z"}},"outputs":[],"execution_count":4},{"cell_type":"markdown","source":"Get GEMINI_API_KEY stored in User Secrets.","metadata":{}},{"cell_type":"code","source":"# Get the API key from here: https://ai.google.dev/tutorials/setup\n# Create a new secret called \"GEMINI_API_KEY\" via Add-ons -> Secrets in the top menu, and attach it to this notebook.\n\nfrom kaggle_secrets import UserSecretsClient\n\nGOOGLE_API_KEY = UserSecretsClient().get_secret(\"GOOGLE_API_KEY\")\ngenai.configure(api_key=GOOGLE_API_KEY)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:54:47.337684Z","iopub.execute_input":"2024-12-01T22:54:47.338246Z","iopub.status.idle":"2024-12-01T22:54:47.555189Z","shell.execute_reply.started":"2024-12-01T22:54:47.338184Z","shell.execute_reply":"2024-12-01T22:54:47.553912Z"}},"outputs":[],"execution_count":5},{"cell_type":"markdown","source":"\n**Choice of Gemini Model:**\n\ngemini-1.5-flash, alias that points to gemini-1.5-flash-002\n\n* Best for: Image understanding, Video understanding, Audio understanding\n* Use case: Process 3,000 images at a time, Look through 1 hour long videos, Listen to hours of audio\ned here.","metadata":{}},{"cell_type":"markdown","source":"!Testing ! - Run your first prompt\nIn this step, you will test that your API key is set up correctly by making a request. The gemini-1.5-flash model has been selected here.","metadata":{}},{"cell_type":"code","source":"# found 'gemini-1.5-flash-001' response to be vastly different and more useful, but still went with latest version of gemini flash\n#flash = genai.GenerativeModel('gemini-1.5-flash-001')\n# Create the model\ngeneration_config = {\n    \"temperature\": 0.7,\n    \"top_p\": 0.75,\n    \"top_k\": 10,\n    \"max_output_tokens\": 8192\n}\n\n\nflash = genai.GenerativeModel(model_name='gemini-1.5-flash-002', generation_config=generation_config)\nresponse = flash.generate_content(\"Explain AI to me like I'm a kid.\")\nprint(response.text)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:54:47.557025Z","iopub.execute_input":"2024-12-01T22:54:47.557492Z","iopub.status.idle":"2024-12-01T22:54:49.722845Z","shell.execute_reply.started":"2024-12-01T22:54:47.557441Z","shell.execute_reply":"2024-12-01T22:54:49.721594Z"}},"outputs":[{"name":"stdout","text":"Imagine you have a super smart puppy.  You teach it tricks, like \"sit\" and \"fetch.\"  At first, it doesn't know what those words mean, but you show it, and it learns.  \n\nArtificial Intelligence, or AI, is like teaching a computer to be that super smart puppy.  We give the computer lots and lots of information, like pictures of cats and dogs, or sentences in different languages.  The computer learns from this information, just like your puppy learns from your training.\n\nThen, we can ask the computer things, like \"Is this a cat or a dog?\" or \"Translate this sentence into Spanish.\"  Because it learned from all that information, it can answer!  It's not really *thinking* like you and me, but it can follow instructions and solve problems based on what it has learned.\n\nSo, AI is basically a really smart computer program that can learn and do things that usually only people can do, like understanding speech, recognizing faces, and even playing games!\n\n","output_type":"stream"}],"execution_count":6},{"cell_type":"markdown","source":"***When to use context caching***\n\r\nContext caching is particularly well suited to scenarios where a substantial initial context is referenced repeatedly by shorter requests. Consider using context caching for use cases such as:\r\n* Chatbots with extensive system instructions\n* Repetitive analysis of lengthy video files\n* Recurring queries against large document sets\n* Frequent code repository analysis or bug fixingtput tokens.","metadata":{}},{"cell_type":"markdown","source":"***How caching reduces costs***\n\nContext caching is a paid feature designed to reduce overall operational costs. Billing is based on the following factors:\n\n1. **Cache token count**: The number of input tokens cached, billed at a reduced rate when included in subsequent prompts.\n2. **Storage duration**: The amount of time cached tokens are stored (TTL), billed based on the TTL duration of cached token count. There are no minimum or maximum bounds on the TTL.\n3. **Other factors**: Other charges apply, such as for non-cached input tokens and output tokens.\n","metadata":{}},{"cell_type":"code","source":"magnus_video1 = '/kaggle/input/magnus-chess-blitz/magnus_chess_blitz.mp4'\nVideo(magnus_video1, embed=True, width=640, height=480)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:58:39.310646Z","iopub.execute_input":"2024-12-01T22:58:39.311127Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"chess_game_instructions_1 = '/kaggle/input/chess-guides/1 Chess Game Instructions.pdf'","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"chess_game_instructions_2 = '/kaggle/input/chess-guides/Chess_ The Complete Guide To Chess - Master_ Chess Tactics Chess Openings and Chess Strategies.pdf'","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"**Clearing Cached Content**\r\n\nCllear any existing cached content to ensure a fresh start.","metadata":{}},{"cell_type":"code","source":"# Clear cached content to avoid using old data\n\nfor content in genai.caching.CachedContent.list():\n    genai.caching.CachedContent.delete(content)","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Upload the video using the Files API\nvideo_file = genai.upload_file(path=magnus_video1)\n\n# Wait for the file to finish processing\nwhile video_file.state.name == 'PROCESSING':\n  print('Waiting for video to be processed.')\n  time.sleep(2)\n  video_file = genai.get_file(video_file.name)\n\nprint(f'Video processing complete: {video_file.uri}')\n","metadata":{"trusted":true},"outputs":[{"name":"stdout","text":"Waiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\nWaiting for video to be processed.\n","output_type":"stream"}],"execution_count":null},{"cell_type":"code","source":"pdf_file_1 = genai.upload_file(path=chess_game_instructions_1)\n\npdf_file_2 = genai.upload_file(path=chess_game_instructions_2)","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"\n# Create a cache with a 60 minute TTL\ncache = caching.CachedContent.create(\n    model='models/gemini-1.5-flash-002',\n    display_name='magnus chess blitz of 51 games', # used to identify the cache\n    system_instruction=(\n        'You are an AI expert chess games video analyst and audio commentator analyst for player Magnus Carlsen (DrNykterstein) who plays 51 blitz chess games, and your job is to answer '\n        'the user\\'s query based on the video file you have access to.'\n        'Each game is of 3 minutes or 180 seconds by looking at the clock at top right is played in a sequence where Magnus Carlsen (DrNykterstein) plays his pieces from the bottom'\n        'The commentator explains when one game finishes and next games starts by commenting next game, game 1, game 2, game 3, game 4, moving on, game resign, resignation, or when the king box is red'\n        'You can read the two pdf_files to understand how chess is played and the various strategies used for winning'\n        'Here is an interview of Magnus Carlsen (DrNykterstein) explaining how he plays'\n        'I don\\'t calculate much or try looking for the right move.'\n        'I always say that i can feel the correct move i.e. intuition.'\n        'Many people have been asking how I trained my intuition to become so powerful that I don\\'t need to look for the right moves, but rather feel it.'\n        'Well here is my secret. I have memorised over 10,000 great games and to this very day I still remember it.'\n        'I dont just memorise the games at first. I analyse them, make my own notes, sleep on my notes, and sometimes use an engine to assist me.'\n        'I memorise the game to the point that I have dreams of it.'\n        'When I don\\'t have a chess board in front of me i visualize it and replay the game in my head and along side it i make make more mental notes or futher improve on it.'\n        'Not only does this help me in blindfold chess but it also makes it a whole hell of a lot easier to calculate.'\n        'I repeat these steps until I can literally play the game without having to concentrate on moves.'\n        'I do this for every great game that has been played.'\n        'Now you might be asking how does this relate to my super strong unrivaled intuition.'\n        'It\\'s easy. Seeing as I have flawlessly memorised 10,000 games the correct moves are always in the back of my head.'\n        'If in a game I am in an uncharted territory (which is rare) I will replay a similar game to the one I\\'m playing at the board and get ideas from it.'\n        'Having all of these moves memorised makes it much easier for me to recognise the correct move.'\n        'Thus practicing this skill took my intuition on a whole another level.'\n        'It\\'s like I\\'m cheating because I have access to my own database in my head.' \n        'Hahaha but I\\'m not technically cheating. Hard work always pays off.'\n        'So my advice to you is to; analyse, annotate, memorise every great game played. Sleep on the games, read your notes the first thing in the morning, play the games in your head (whenever you\\'re bored i.e. on the bus etc.)'\n        'And always hit the gym. Your physical health is just as important as your mental health.'\n        'Sometimes when two powerful opponents clash off on the board the end result is determined by the one who is in better shape and can keep calculating.'\n        'Always return markdown.'\n        ),\n    contents=[video_file, pdf_file_1, pdf_file_2],\n    ttl=datetime.timedelta(minutes=60),\n    safety_settings={\n        'HATE': 'BLOCK_NONE',\n        'HARASSMENT': 'BLOCK_NONE',\n        'SEXUAL' : 'BLOCK_NONE',\n        'DANGEROUS' : 'BLOCK_NONE'\n        }\n)\n\n# Cinfiguration\ngeneration_config = {\n    \"temperature\": 0.7,\n    \"top_p\": 0.75,\n    \"top_k\": 10,\n    \"max_output_tokens\": 8192\n}\n\n# Construct a GenerativeModel which uses the created cache.\n#model = genai.GenerativeModel.from_cached_content(cached_content=cache, generation_config=generation_config)\nmodel = genai.GenerativeModel.from_cached_content(cached_content=cache, generation_config=generation_config)","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"from contextlib import contextmanager\n\n@contextmanager\ndef measure_time():\n    start_time = time.perf_counter()\n    yield\n    end_time = time.perf_counter()\n    elapsed_time = end_time - start_time\n    print(f\"Elapsed time: {elapsed_time:.4f} seconds\")","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Query the model\n\nwith measure_time():\n  response = model.generate_content(\"provide a summary for the video\")\n  print(response.usage_metadata)","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"Markdown(response.text)","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Query the model\nprompt = f\"I am learning chess. Can you summarize the strategies from watching DrNykterstein (aka Magnus Carlsen) moves: {video_file}\"\nresponse = model.generate_content([video_file, prompt])\n\nprint(response.usage_metadata)\n\n# The output should look something like this:\n#\n# prompt_token_count: 696219\n# cached_content_token_count: 696190\n# candidates_token_count: 214\n# total_token_count: 696433","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"print(response.text)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:16:47.548891Z","iopub.execute_input":"2024-12-01T22:16:47.549311Z","iopub.status.idle":"2024-12-01T22:16:47.556571Z","shell.execute_reply.started":"2024-12-01T22:16:47.549254Z","shell.execute_reply":"2024-12-01T22:16:47.55528Z"}},"outputs":[{"name":"stdout","text":"Analyzing Magnus Carlsen's games to distill overarching strategies is challenging because his strength lies in his incredibly deep understanding and adaptability.  He doesn't rigidly adhere to one style, but rather employs a wide range of strategies based on the specific position. However, some recurring themes emerge from observing his play:\n\n**1. Proactive and Precise Development:** Carlsen prioritizes rapid, efficient development of his pieces, aiming for a harmonious structure early on. He rarely wastes tempo. His development is often geared towards controlling key squares and initiating a plan, rather than just getting pieces out.\n\n**2. Subtle Strategic Understanding:**  He excels at identifying subtle imbalances and exploiting them. This involves a deep understanding of pawn structures, piece activity, and king safety.  He's masterful at finding small advantages and gradually converting them into winning positions.  He often sacrifices material for positional advantages if the compensation is significant enough.\n\n**3. Dynamic Play and Initiative:** While capable of positional maneuvering, Carlsen often seeks to create dynamic imbalances and seize the initiative. He's not afraid to launch attacks, even if they involve some risk, if he sees a path to significant advantage.\n\n**4. Exceptional Calculation:**  Carlsen's calculation abilities are legendary. He's able to accurately assess complex tactical variations, often several moves deep, to identify winning lines or avoid traps. This is crucial for both attack and defense.\n\n**5. Adaptability and Flexibility:**  He adjusts his strategy based on his opponent's moves and the evolving position. He's not bound by rigid opening principles or plans; he's always looking for the best move in the current situation.  He's equally comfortable playing aggressively or passively, depending on what the position demands.\n\n**6. King Safety:** While aiming for dynamic play, he always maintains a keen awareness of his king's safety.  He rarely allows his king to be exposed unnecessarily.\n\n**7. Deep Understanding of Endgame Principles:**  Carlsen's endgame technique is exceptional. He's able to convert seemingly small advantages into wins with precise maneuvering and understanding of pawn structures and piece activity in the endgame.\n\n**In short:**  Learning from Carlsen isn't about copying specific moves but about adopting his mindset:  **proactive development, deep strategic understanding, precise calculation, adaptability, and a relentless pursuit of even the smallest advantage.**  Focus on understanding *why* he makes the moves he does, rather than just memorizing them.  Analyzing his games with a strong chess engine and commentary will be invaluable for understanding his thought process.\n\n","output_type":"stream"}],"execution_count":51},{"cell_type":"markdown","source":"Define methods to upload large data/video files.","metadata":{}},{"cell_type":"code","source":"# Query the model\nresponse = model.generate_content([(\n    'Can you list any or all the popular strategies used by DrNykterstein (aka Magnus Carlsen) from watching DrNykterstein (aka Magnus Carlsen) moves and explain the pices for those moves')])\n\nprint(response.usage_metadata)\n\n# The output should look something like this:\n#\n# prompt_token_count: 696219\n# cached_content_token_count: 696190\n# candidates_token_count: 214\n# total_token_count: 696433","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:17:04.732999Z","iopub.execute_input":"2024-12-01T22:17:04.733383Z","iopub.status.idle":"2024-12-01T22:17:09.247148Z","shell.execute_reply.started":"2024-12-01T22:17:04.733351Z","shell.execute_reply":"2024-12-01T22:17:09.2461Z"}},"outputs":[{"name":"stdout","text":"prompt_token_count: 3010615\ncandidates_token_count: 643\ntotal_token_count: 3011258\ncached_content_token_count: 3010570\n\n","output_type":"stream"}],"execution_count":56},{"cell_type":"code","source":"print(response.text)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:17:09.248932Z","iopub.execute_input":"2024-12-01T22:17:09.249255Z","iopub.status.idle":"2024-12-01T22:17:09.254754Z","shell.execute_reply.started":"2024-12-01T22:17:09.249223Z","shell.execute_reply":"2024-12-01T22:17:09.253645Z"}},"outputs":[{"name":"stdout","text":"Analyzing Magnus Carlsen's games to pinpoint specific, named \"strategies\" is difficult because his strength lies in his incredibly deep understanding of chess principles, rather than relying on a set of easily categorized strategies.  He adapts his approach constantly based on the position, opponent, and tournament context.  However, we can identify recurring *themes* and *positional preferences* that contribute to his success:\n\n**1. Prophylactic Play & Prevention:** Carlsen excels at anticipating his opponent's plans and subtly preventing them before they become dangerous. This isn't a flashy strategy, but it's incredibly effective.  He often makes moves that seem passive or even slightly disadvantageous at first glance, but they subtly neutralize threats or limit opponent's options, giving him a long-term positional advantage.  The pieces involved vary wildly depending on the position, but it often involves controlling key squares or preventing pawn breaks.\n\n**2. Subtle Pawn Structures & Weaknesses:**  Carlsen is a master of manipulating pawn structures. He's adept at creating weaknesses in his opponent's pawn structure while maintaining his own, or even creating weaknesses that he can exploit later in the game.  This often involves slow maneuvering of pawns and pieces to create imbalances.  The pieces involved are usually pawns themselves, but also knights and bishops positioned to exploit the created weaknesses.\n\n**3. Dynamic Piece Play:** While known for positional understanding, Carlsen is also a powerful tactical player. He often sacrifices material or creates complex tactical positions to gain a decisive advantage.  These aren't always \"strategies\" in the sense of a pre-planned approach, but rather arise from his deep understanding of the position and his ability to calculate variations accurately.  All pieces can be involved, but often rooks and queens are key players in these dynamic moments.\n\n**4. King Safety & Endgame Technique:** Carlsen prioritizes king safety early in the game and demonstrates exceptional endgame technique.  He's often willing to sacrifice material to improve his king's safety or to gain a positional advantage that will be decisive in the endgame.  This involves careful pawn structures, piece placement around the king, and precise calculation in the endgame.  All pieces play a role, but the king's safety and the pawns are crucial.\n\n**5. Psychological Warfare & Opponent's Psychology:** While not a chess strategy per se, Carlsen is a master of psychological warfare. He understands his opponents' playing styles and can subtly influence their decisions through his moves and demeanor.  This isn't something that can be easily described in terms of piece movement, but it's a significant factor in his success.\n\n\nIt's important to note that these are *themes* and not rigid strategies.  To truly understand Carlsen's play, you need to study his games in detail, paying attention to his reasoning behind each move, rather than trying to fit his play into pre-defined strategic boxes.  Many of his moves appear simple, but their effectiveness comes from the profound understanding of the position that underlies them.  He doesn't follow a checklist; he adapts to the specifics of each game.\n\n","output_type":"stream"}],"execution_count":57},{"cell_type":"code","source":"# Query the model\nprompt = f\"how may games did DrNykterstein (aka Magnus Carlsen) win in the content files called video_file, there  are 51 games in total: {video_file}\"\n\nresponse = model.generate_content([video_file, prompt])\n\nprint(response.usage_metadata)\n\n# The output should look something like this:\n#\n# prompt_token_count: 696219\n# cached_content_token_count: 696190\n# candidates_token_count: 214\n# total_token_count: 696433","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:47:12.839581Z","iopub.status.idle":"2024-12-01T22:47:12.839938Z","shell.execute_reply.started":"2024-12-01T22:47:12.83977Z","shell.execute_reply":"2024-12-01T22:47:12.839787Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"print(response.text)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:47:12.841079Z","iopub.status.idle":"2024-12-01T22:47:12.841464Z","shell.execute_reply.started":"2024-12-01T22:47:12.841245Z","shell.execute_reply":"2024-12-01T22:47:12.841261Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Generate a summary\nprompt = f\"From the video file, can you summarize the main strategies Magnus Carlsen used in his games?: {video_file}\"\nresponse = model.generate_content([video_file, prompt])\n\nprint(response.usage_metadata)\n\n# The output should look something like this:\n#\n# prompt_token_count: 696219\n# cached_content_token_count: 696190\n# candidates_token_count: 214\n# total_token_count: 696433","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:45:12.565711Z","iopub.execute_input":"2024-12-01T22:45:12.566168Z","iopub.status.idle":"2024-12-01T22:47:12.836472Z","shell.execute_reply.started":"2024-12-01T22:45:12.566129Z","shell.execute_reply":"2024-12-01T22:47:12.834363Z"}},"outputs":[{"traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mInvalidArgument\u001b[0m                           Traceback (most recent call last)","Cell \u001b[0;32mIn[73], line 3\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[38;5;66;03m# Generate a summary\u001b[39;00m\n\u001b[1;32m      2\u001b[0m prompt \u001b[38;5;241m=\u001b[39m \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mFrom the video file, can you summarize the main strategies Magnus Carlsen used in his games?: \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mvideo_file\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m\"\u001b[39m\n\u001b[0;32m----> 3\u001b[0m response \u001b[38;5;241m=\u001b[39m \u001b[43mmodel\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mgenerate_content\u001b[49m\u001b[43m(\u001b[49m\u001b[43m[\u001b[49m\u001b[43mvideo_file\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mprompt\u001b[49m\u001b[43m]\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m      5\u001b[0m \u001b[38;5;28mprint\u001b[39m(response\u001b[38;5;241m.\u001b[39musage_metadata)\n\u001b[1;32m      7\u001b[0m \u001b[38;5;66;03m# The output should look something like this:\u001b[39;00m\n\u001b[1;32m      8\u001b[0m \u001b[38;5;66;03m#\u001b[39;00m\n\u001b[1;32m      9\u001b[0m \u001b[38;5;66;03m# prompt_token_count: 696219\u001b[39;00m\n\u001b[1;32m     10\u001b[0m \u001b[38;5;66;03m# cached_content_token_count: 696190\u001b[39;00m\n\u001b[1;32m     11\u001b[0m \u001b[38;5;66;03m# candidates_token_count: 214\u001b[39;00m\n\u001b[1;32m     12\u001b[0m \u001b[38;5;66;03m# total_token_count: 696433\u001b[39;00m\n","File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/google/generativeai/generative_models.py:331\u001b[0m, in \u001b[0;36mGenerativeModel.generate_content\u001b[0;34m(self, contents, generation_config, safety_settings, stream, tools, tool_config, request_options)\u001b[0m\n\u001b[1;32m    329\u001b[0m         \u001b[38;5;28;01mreturn\u001b[39;00m generation_types\u001b[38;5;241m.\u001b[39mGenerateContentResponse\u001b[38;5;241m.\u001b[39mfrom_iterator(iterator)\n\u001b[1;32m    330\u001b[0m     \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m--> 331\u001b[0m         response \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_client\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mgenerate_content\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    332\u001b[0m \u001b[43m            \u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    333\u001b[0m \u001b[43m            \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mrequest_options\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    334\u001b[0m \u001b[43m        \u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    335\u001b[0m         \u001b[38;5;28;01mreturn\u001b[39;00m generation_types\u001b[38;5;241m.\u001b[39mGenerateContentResponse\u001b[38;5;241m.\u001b[39mfrom_response(response)\n\u001b[1;32m    336\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m google\u001b[38;5;241m.\u001b[39mapi_core\u001b[38;5;241m.\u001b[39mexceptions\u001b[38;5;241m.\u001b[39mInvalidArgument \u001b[38;5;28;01mas\u001b[39;00m e:\n","File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/google/generativeai/client.py:267\u001b[0m, in \u001b[0;36m_ClientManager.make_client.<locals>.add_default_metadata_wrapper.<locals>.call\u001b[0;34m(metadata, *args, **kwargs)\u001b[0m\n\u001b[1;32m    265\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mcall\u001b[39m(\u001b[38;5;241m*\u001b[39margs, metadata\u001b[38;5;241m=\u001b[39m(), \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs):\n\u001b[1;32m    266\u001b[0m     metadata \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mlist\u001b[39m(metadata) \u001b[38;5;241m+\u001b[39m \u001b[38;5;28mlist\u001b[39m(\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdefault_metadata)\n\u001b[0;32m--> 267\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mf\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mmetadata\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mmetadata\u001b[49m\u001b[43m)\u001b[49m\n","File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/google/ai/generativelanguage_v1beta/services/generative_service/client.py:830\u001b[0m, in \u001b[0;36mGenerativeServiceClient.generate_content\u001b[0;34m(self, request, model, contents, retry, timeout, metadata)\u001b[0m\n\u001b[1;32m    827\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_validate_universe_domain()\n\u001b[1;32m    829\u001b[0m \u001b[38;5;66;03m# Send the request.\u001b[39;00m\n\u001b[0;32m--> 830\u001b[0m response \u001b[38;5;241m=\u001b[39m \u001b[43mrpc\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    831\u001b[0m \u001b[43m    \u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    832\u001b[0m \u001b[43m    \u001b[49m\u001b[43mretry\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mretry\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    833\u001b[0m \u001b[43m    \u001b[49m\u001b[43mtimeout\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mtimeout\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    834\u001b[0m \u001b[43m    \u001b[49m\u001b[43mmetadata\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mmetadata\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    835\u001b[0m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    837\u001b[0m \u001b[38;5;66;03m# Done; return the response.\u001b[39;00m\n\u001b[1;32m    838\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m response\n","File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/google/api_core/gapic_v1/method.py:131\u001b[0m, in \u001b[0;36m_GapicCallable.__call__\u001b[0;34m(self, timeout, retry, compression, *args, **kwargs)\u001b[0m\n\u001b[1;32m    128\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_compression \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[1;32m    129\u001b[0m     kwargs[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mcompression\u001b[39m\u001b[38;5;124m\"\u001b[39m] \u001b[38;5;241m=\u001b[39m compression\n\u001b[0;32m--> 131\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mwrapped_func\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n","File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/google/api_core/retry/retry_unary.py:293\u001b[0m, in \u001b[0;36mRetry.__call__.<locals>.retry_wrapped_func\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m    289\u001b[0m target \u001b[38;5;241m=\u001b[39m functools\u001b[38;5;241m.\u001b[39mpartial(func, \u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs)\n\u001b[1;32m    290\u001b[0m sleep_generator \u001b[38;5;241m=\u001b[39m exponential_sleep_generator(\n\u001b[1;32m    291\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_initial, \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_maximum, multiplier\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_multiplier\n\u001b[1;32m    292\u001b[0m )\n\u001b[0;32m--> 293\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mretry_target\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    294\u001b[0m \u001b[43m    \u001b[49m\u001b[43mtarget\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    295\u001b[0m \u001b[43m    \u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_predicate\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    296\u001b[0m \u001b[43m    \u001b[49m\u001b[43msleep_generator\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    297\u001b[0m \u001b[43m    \u001b[49m\u001b[43mtimeout\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_timeout\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    298\u001b[0m \u001b[43m    \u001b[49m\u001b[43mon_error\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mon_error\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    299\u001b[0m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m\n","File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/google/api_core/retry/retry_unary.py:153\u001b[0m, in \u001b[0;36mretry_target\u001b[0;34m(target, predicate, sleep_generator, timeout, on_error, exception_factory, **kwargs)\u001b[0m\n\u001b[1;32m    149\u001b[0m \u001b[38;5;66;03m# pylint: disable=broad-except\u001b[39;00m\n\u001b[1;32m    150\u001b[0m \u001b[38;5;66;03m# This function explicitly must deal with broad exceptions.\u001b[39;00m\n\u001b[1;32m    151\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mException\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m exc:\n\u001b[1;32m    152\u001b[0m     \u001b[38;5;66;03m# defer to shared logic for handling errors\u001b[39;00m\n\u001b[0;32m--> 153\u001b[0m     \u001b[43m_retry_error_helper\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    154\u001b[0m \u001b[43m        \u001b[49m\u001b[43mexc\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    155\u001b[0m \u001b[43m        \u001b[49m\u001b[43mdeadline\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    156\u001b[0m \u001b[43m        \u001b[49m\u001b[43msleep\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    157\u001b[0m \u001b[43m        \u001b[49m\u001b[43merror_list\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    158\u001b[0m \u001b[43m        \u001b[49m\u001b[43mpredicate\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    159\u001b[0m \u001b[43m        \u001b[49m\u001b[43mon_error\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    160\u001b[0m \u001b[43m        \u001b[49m\u001b[43mexception_factory\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    161\u001b[0m \u001b[43m        \u001b[49m\u001b[43mtimeout\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    162\u001b[0m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    163\u001b[0m     \u001b[38;5;66;03m# if exception not raised, sleep before next attempt\u001b[39;00m\n\u001b[1;32m    164\u001b[0m     time\u001b[38;5;241m.\u001b[39msleep(sleep)\n","File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/google/api_core/retry/retry_base.py:212\u001b[0m, in \u001b[0;36m_retry_error_helper\u001b[0;34m(exc, deadline, next_sleep, error_list, predicate_fn, on_error_fn, exc_factory_fn, original_timeout)\u001b[0m\n\u001b[1;32m    206\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m predicate_fn(exc):\n\u001b[1;32m    207\u001b[0m     final_exc, source_exc \u001b[38;5;241m=\u001b[39m exc_factory_fn(\n\u001b[1;32m    208\u001b[0m         error_list,\n\u001b[1;32m    209\u001b[0m         RetryFailureReason\u001b[38;5;241m.\u001b[39mNON_RETRYABLE_ERROR,\n\u001b[1;32m    210\u001b[0m         original_timeout,\n\u001b[1;32m    211\u001b[0m     )\n\u001b[0;32m--> 212\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m final_exc \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01msource_exc\u001b[39;00m\n\u001b[1;32m    213\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m on_error_fn \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[1;32m    214\u001b[0m     on_error_fn(exc)\n","File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/google/api_core/retry/retry_unary.py:144\u001b[0m, in \u001b[0;36mretry_target\u001b[0;34m(target, predicate, sleep_generator, timeout, on_error, exception_factory, **kwargs)\u001b[0m\n\u001b[1;32m    142\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m sleep \u001b[38;5;129;01min\u001b[39;00m sleep_generator:\n\u001b[1;32m    143\u001b[0m     \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[0;32m--> 144\u001b[0m         result \u001b[38;5;241m=\u001b[39m \u001b[43mtarget\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    145\u001b[0m         \u001b[38;5;28;01mif\u001b[39;00m inspect\u001b[38;5;241m.\u001b[39misawaitable(result):\n\u001b[1;32m    146\u001b[0m             warnings\u001b[38;5;241m.\u001b[39mwarn(_ASYNC_RETRY_WARNING)\n","File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/google/api_core/timeout.py:120\u001b[0m, in \u001b[0;36mTimeToDeadlineTimeout.__call__.<locals>.func_with_timeout\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m    117\u001b[0m     \u001b[38;5;66;03m# Avoid setting negative timeout\u001b[39;00m\n\u001b[1;32m    118\u001b[0m     kwargs[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mtimeout\u001b[39m\u001b[38;5;124m\"\u001b[39m] \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mmax\u001b[39m(\u001b[38;5;241m0\u001b[39m, \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_timeout \u001b[38;5;241m-\u001b[39m time_since_first_attempt)\n\u001b[0;32m--> 120\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n","File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/google/api_core/grpc_helpers.py:78\u001b[0m, in \u001b[0;36m_wrap_unary_errors.<locals>.error_remapped_callable\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m     76\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m callable_(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs)\n\u001b[1;32m     77\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m grpc\u001b[38;5;241m.\u001b[39mRpcError \u001b[38;5;28;01mas\u001b[39;00m exc:\n\u001b[0;32m---> 78\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m exceptions\u001b[38;5;241m.\u001b[39mfrom_grpc_error(exc) \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01mexc\u001b[39;00m\n","\u001b[0;31mInvalidArgument\u001b[0m: 400 Unable to submit request because the input token count is 2923625 but model only supports up to 1000000. Reduce the input token count and try again. You can also use the CountTokens API to calculate prompt token count and billable characters. Learn more: https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models"],"ename":"InvalidArgument","evalue":"400 Unable to submit request because the input token count is 2923625 but model only supports up to 1000000. Reduce the input token count and try again. You can also use the CountTokens API to calculate prompt token count and billable characters. Learn more: https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models","output_type":"error"}],"execution_count":73},{"cell_type":"code","source":"print(response.text)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-01T22:47:12.837497Z","iopub.status.idle":"2024-12-01T22:47:12.837967Z","shell.execute_reply.started":"2024-12-01T22:47:12.837782Z","shell.execute_reply":"2024-12-01T22:47:12.837803Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"**List caches**\n\r\nIt's not possible to retrieve or view cached content, but you can retrieve cache metadata (name, model, display_name, usage_metadata, create_time, update_time, and expire_time).\r\n\r\nTo list metadata for all uploaded caches, use CachedContent.list():","metadata":{}},{"cell_type":"code","source":"for c in caching.CachedContent.list():\n  print(c)","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"**Update a cache**\n\n\r\nYou can set a new ttl or expire_time for a cache. Changing anything else about the cache isn't supported.\r\n\r\nThe following example shows how to update the ttl of a cache using CachedContent.update().","metadata":{}},{"cell_type":"code","source":"cache.update(ttl=datetime.timedelta(hours=2))","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"**Delete a cache**\n\nThe caching service provides a delete operation for manually removing content from the cache. The following example shows how to delete a cache using CachedContent.delete().","metadata":{}},{"cell_type":"code","source":"cache.delete()","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Bibliography\n* https://aistudio.google.com/app/apikey\n* https://storage.googleapis.com/deepmind-media/gemini/gemini_v1_5_report.pdf\n* https://ai.google.dev/gemini-api/docs/text-generation?lang=python\n* https://ai.google.dev/api/generate-content#image","metadata":{}}]}