TLoL (Large Language Model) - League of Legends LLM Module (Integrates LLMs for Game Analysis and Game Playing)

• llm.ipynb: Initial LLM experiments on data

A brief description of datasets used for this project are listed below:

• ESPORTSTMNT02-3080905(old).db
  • context: League of Legends - Worlds 2022 Final (Game 5)
  • total files: 1
  • approx size: 403MB
  • map: Summoner's Rift
  • frames: 16,858 (~4.5 obs/sec * 2528.97 secs + many duplicate frames)
  • source: Bayes Esports (*.rofl) + TLoL-Scraper (*.db)
  • download: Google Drive

Each file is an SQLite3 database generated using TLoL-Scraper, which has scraped the game objects from a live running Leauge of Legends replay. Each database contains the following 4 tables:

• games (first table)
  • game_id (Internal Riot Game ID, ignore this for Bayes Esports replays)
  • duration (Game Duration in seconds)
• champs/missiles/objects (remaining three tables)
  • game_id (Internal Riot Game ID, ignore this for Bayes Esports replays)
  • time (Game time in seconds)
  • obj_type ()
  • etc. (Refer to this) for full specification