Hugging Face Overview
Hammer-1.5b: Function Calling for On-Device Agents
Hammer-1.5b is a 1.5 billion parameter model from the Hammer series by MadeAgents, specifically engineered to enhance the function calling capabilities of AI agents. Unlike models that primarily focus on data refinement, Hammer optimizes performance through advanced training techniques, including function masking, and is built upon the Qwen 2.0 series architecture.
Key Capabilities
- Robust Function Calling: Hammer-1.5b is highly proficient in interpreting user queries and accurately mapping them to available tools, enabling complex agentic behaviors.
- On-Device Optimization: Designed with on-device applications in mind, it provides strong performance in a lightweight package, making it suitable for personalized, local AI agents.
- Advanced Training: The model was fine-tuned using the APIGen Function Calling Datasets (60,000 samples) and supplemented with xlam-irrelevance-7.5k, demonstrating exceptional performance across various function calling benchmarks.
- Generalization: Evaluations on academic benchmarks beyond function calling indicate the model's stable and robust generalization ability.
Good For
- Developers building on-device agentic applications that require precise tool use.
- Scenarios demanding reliable function calling from a compact language model.
- Integrating AI agents into environments where computational resources are limited but sophisticated interaction is needed.