Dataset Viewer
Auto-converted to Parquet Duplicate
training_summary
dict
agent_comparison
dict
export_timestamp
stringclasses
1 value
system_info
dict
{ "total_games": 0, "total_moves": 11, "training_time": "00:00:12", "current_game": 2, "training_active": false }
{ "black_win_rate": "0.0", "green_win_rate": "0.0", "draws_rate": "0.0" }
2026-01-02T22:48:02.103Z
{ "user_agent": "Mozilla/5.0 (X11; Linux x86_64; rv:146.0) Gecko/20100101 Firefox/146.0", "platform": "Linux x86_64", "screen_resolution": "1920x1080" }

Chess RL Training Export

This dataset was created using synthetic AI vs AI training app found in /generator/. The app simulates games of chess between 2 web workers in a front-end page to train RL datasets. This dataset is for testing and UNDER DEVELOPMENT as we attempt to enhance and make more datasets using this app.

Export Information

  • Export Date: 2026-01-02T22:48:02.105Z
  • Total Training Games: 0
  • Total Moves: 11
  • Training Time: 00:00:12

Files Included

1. training_games.json

Complete dataset of all chess games played during training. Each game includes:

  • Full PGN notation
  • Move-by-move records
  • Game result and metadata
  • Agent parameters for each game

2. training_games.csv

Same data as JSON but in CSV format for easy import into spreadsheets or databases.

3. black_agent_model.json

Black Agent (Policy Network) configuration and statistics:

  • Neural network architecture
  • Hyperparameters (learning rate, exploration rate, etc.)
  • Training statistics (wins, losses, draws)
  • Model metadata

4. green_agent_model.json

Green Agent (Value Network) configuration and statistics:

  • Neural network architecture
  • Hyperparameters
  • Training statistics
  • Model metadata

5. training_statistics.json

Overall training summary and statistics including:

  • Training duration
  • Win rates for both agents
  • System information
  • Export metadata

Training System

Generated by ANN Chess RL Trainer v3.0 - A web-based reinforcement learning system for chess AI development.

Usage

These files can be used to:

  • Continue training from this point
  • Analyze the learning progress
  • Import into other machine learning frameworks
  • Share with the research community

Notes

  • All data is in standard JSON/CSV formats
  • Compatible with Hugging Face datasets
  • Can be compressed with GZIP, ZSTD, BZ2, LZ4, or LZMA for upload
Downloads last month
-