Hammer2.1-7b
| Property | Value |
|---|---|
| Model Size | 7B parameters |
| Base Model | Qwen 2.5 coder series |
| Hugging Face | MadeAgents/Hammer2.1-7b |
| Requirements | transformers>=4.47.0 |
What is Hammer2.1-7b?
Hammer2.1-7b is a state-of-the-art Large Action Model specifically designed for advanced function calling capabilities. Built upon the Qwen 2.5 coder series, it represents the latest iteration in the Hammer series, offering enhanced functionality while maintaining efficient performance in a 7B parameter model size.
Implementation Details
The model utilizes sophisticated function masking techniques and can be deployed using either vLLM for efficient serving or Hugging Face Transformers. It supports both single-turn and multi-turn interactions, with particular emphasis on complex function calling scenarios.
- Built on Qwen 2.5 coder series architecture
- Implements advanced function masking techniques
- Supports multiple deployment options (vLLM and Hugging Face)
- Optimized for both efficiency and functionality
Core Capabilities
- Multi-Step Function Calling: Can perform multiple internal function calls within a single request
- Multi-Turn Function Calling: Maintains context awareness across multiple exchanges
- Enhanced Irrelevant Information Inspection: Improved ability to identify and handle irrelevant function calls
- Strong performance on Berkeley Function-Calling Leaderboard (BFCL-v3)
Frequently Asked Questions
Q: What makes this model unique?
Hammer2.1-7b stands out for its specialized function calling capabilities, particularly its ability to handle multi-step and multi-turn interactions while maintaining high accuracy in function selection and execution. It's specifically optimized for practical applications requiring complex function calling sequences.
Q: What are the recommended use cases?
The model is ideal for applications requiring sophisticated function calling capabilities, such as chatbots, automated assistants, and systems requiring complex API interactions. It's particularly well-suited for scenarios requiring multiple sequential function calls or maintaining context across multiple interaction turns.
