hamsterspeak-qwen2.5-3b

HamsterSpeak Qwen2.5-3B

A fine-tuned version of Qwen2.5-3B-Instruct specialized for HamsterSpeak scripting in the OHRRPGCE (Official Hamster Republic Role-Playing Game Construction Engine).

Model Details

• Base Model: Qwen/Qwen2.5-3B-Instruct
• Training Framework: MLX-LM (Apple Metal)
• Training Method: QLoRA (4-bit quantized LoRA)
• Training Data: 9,250 HamsterSpeak instruction-output pairs with reasoning traces

Training Results

Capabilities

This model is specialized in:

• NPC creation and control
• Cutscene scripting
• Player movement and interaction
• Text box management
• Variable and tag operations
• Battle triggers
• Custom menu systems

Usage

With MLX-LM

from mlx_lm import load, generate

model, tokenizer = load("msprout/hamsterspeak-qwen2.5-3b")
response = generate(model, tokenizer, prompt="How do I create an NPC in HamsterSpeak?")
print(response)

With Ollama

ollama run hf.co/msprout/hamsterspeak-qwen2.5-3b

Example Prompts

How do I prevent the player from moving during a cutscene?

Write a script to move the hero to the right for 3 steps.

How do I create a specific character for a cutscene?

Files

• model.safetensors - Fused model weights
• config.json - Model configuration
• tokenizer.json - Tokenizer vocabulary
• tokenizer_config.json - Tokenizer settings

License

Apache 2.0

Acknowledgments

• Qwen Team for the base model
• Apple MLX team for the MLX framework
• OHRRPGCE community for HamsterSpeak documentation