🎓 Download AI Trainer
Train your own Afrika Korps AI models on your machine, then share them with the community!
Linux x86_64
Complete package with pre-compiled C++ for 10-20x speedup!
Requirements:
- Python 3.11, 3.12, or 3.13
- 8GB+ RAM
- NVIDIA GPU (recommended)
✅ C++ engine pre-compiled - no build required!
# Extract and install unzip akorp-trainer-complete-*.zip cd akorp-trainer-complete pip install . # Then run with C++ acceleration! akorp-trainer login akorp-trainer serve
Windows x64
Complete package with pre-compiled C++ for Windows!
Requirements:
- Python 3.11, 3.12, or 3.13 from python.org
- 8GB+ RAM
- NVIDIA GPU + CUDA Toolkit
✅ C++ engine pre-compiled - no Visual Studio needed!
# Extract ZIP (use Explorer or PowerShell) Expand-Archive akorp-trainer-complete-*.zip cd akorp-trainer-complete # Install and run with C++ acceleration! pip install . akorp-trainer serve
From Source
Clone from GitHub for development.
# Clone repository git clone https://github.com/stevetrabold/Akorp-Python.git cd Akorp-Python # Install dependencies pip install -r requirements.txt # Run trainer CLI python -m games.afrika_korps.ai.cli
Requirements:
- Git installed
- Python 3.11+
- Internet connection
🪟 Windows Setup Guide
Install Python 3.11+
Download from python.org. Check "Add Python to PATH" during install.
Install CUDA (for GPU training)
Download CUDA Toolkit 12.x. Restart after install.
Install PyTorch with CUDA
pip install torch --index-url https://download.pytorch.org/whl/cu128
Use cu124 or cu121 for older NVIDIA drivers
Extract & Install Trainer
Extract ZIP, open folder in PowerShell: pip install .
Start Training!
akorp-trainer serve then open localhost:5050
🖱️ Enable One-Click Launching
After installing the trainer, run this command once to enable launching directly from the browser:
# Register the akorp:// protocol handler akorp-setup-protocol
This registers the akorp:// protocol so you can start your local trainer with one click from any page on this site!
✨ What's Included
Hierarchical AI Training
Train strategic and tactical policies using state-of-the-art reinforcement learning.
Personality Presets
Choose from presets like "Blitzer", "Iron Turtle", or create custom personalities.
Training Dashboard
Monitor progress with live charts showing rewards, win rates, and learning curves.
Model Lineage
Train from community models and track your model's heritage.
Easy Upload
One command to export and upload your trained model to the community.
ELO Rankings
Compete on the leaderboard and watch your AI climb the ranks!
🚀 Training Workflow
Download & Install
Download the trainer for your platform and follow the setup instructions.
Login to Your Account
Run akorp-trainer login and enter your credentials.
Download Parent Model (Optional)
Run akorp-trainer download <model-id> to use an existing model as a starting point.
Start Training
Run akorp-trainer train --name "My AI" --timesteps 500000
Upload Your Model
Run akorp-trainer upload my_model to share with the community!
💻 CLI Commands
# Login to your account akorp-trainer login # List available community models akorp-trainer list --sort elo --limit 10 # Download a model to use as parent akorp-trainer download <model-id> # Train L1 Foundation model (from scratch) akorp-trainer train --tier L1 --name "Foundation v1" --timesteps 2000000 # Train L2 Personality model (with personality preset) akorp-trainer train --tier L2 --parent my_l1_model --personality blitzer # Train L3 Faction-specific model akorp-trainer train --tier L3 --parent my_l2_model --faction axis # Export to ONNX and upload akorp-trainer upload my_model --name "Desert Fox AI" --description "Aggressive Axis strategy" # Check training status akorp-trainer status