UntitledAITrainingGame 0.0.0.1

UntitledGame 0.0.0.1 - First Test Release

This is the first public test build of an experimental character-and-AI
playground. Build characters, set up battles, play a character yourself, or
train a neural network AI how to control it.

This is an early preview, not a finished game. Feedback on usability, bugs,
training behaviour, and the manual is very welcome.

What You Can Try

• Create characters and customise their stats and actions.
• Build playgrounds with multiple teams and code-controlled, human-controlled,
  or AI-controlled characters.
• Record human play and train a policy using behaviour cloning (BC).
• Train a policy using reinforcement learning (RL/PPO), with training and
  evaluation playgrounds and configurable rewards.
• Inspect training progress, keep checkpoints, and continue compatible models
  with either BC or RL.
• Export ONNX models and assign them to characters without requiring a Python
  training environment to run them.

Download And Run

Download UntitledGame.zip from the release assets. Extract the entire
archive and run test.exe.

Read Manual0.0.0.1_en.pdf for the illustrated walkthrough. README_FIRST.txt
contains a shorter installation guide and the save locations.

Optional AI Training Setup

Playing and running existing ONNX format AI models do not require this setup.

To enable BC/RL training, close the game, open rl_environment, and run one
installer:

• install_pinned.bat: CPU training. Start here if you are unsure.
• install_gpu_pinned.bat: the packaged NVIDIA CUDA training environment,
  for a compatible GPU and driver.

Installation downloads Python and packages, so internet access is needed.
After it finishes successfully, restart the game. The game checks the
environment before enabling the AI training menu; GPU training is available
only when its GPU check also passes.

Do not rename rl_environment, copy a virtual environment from another
computer, or run both installers at the same time.

Practical Notes

• There is no in-game tutorial yet. Use the manual and parameter tooltips.
• This is an early release. Compatibility with future builds is not guaranteed.

Release Scope And Credits

See LICENSE_DECLARATION.md, THIRD_PARTY_NOTICES.md, and the accompanying
licenses folder for component-specific licenses and credits. Third-party
assets are not covered by the project's original-code license.

The project was developed with assistance from ChatGPT/Codex.

Feedback

Please report problems through this release repository's Issues page, or email
cchx7461@gmail.com.

For permission or copyright concerns, contact cchx7461@gmail.com.