seanpoyner/smolcode-coder-sql-3b-tools
The seanpoyner/smolcode-coder-sql-3b-tools model is a 3.1 billion parameter LoRA fine-tune of Qwen2.5-Coder-1.5B-Instruct, specifically designed to enable native function call emission. This model is optimized for agentic coding loops, allowing small coder models to effectively drive tool-use. It excels at generating structured tool calls for integration with runtimes like Ollama and llama.cpp, addressing a common limitation in smaller code generation models.
Loading preview...
Overview
This model, seanpoyner/smolcode-coder-sql-3b-tools, is a LoRA fine-tune of the Qwen2.5-Coder-1.5B-Instruct base model, specifically engineered to facilitate native <tool_call> function call generation. It was developed for the smolcode agentic coding assistant during the Hugging Face "Build Small" hackathon.
Key Capabilities
- Native Tool Call Emission: Unlike standard small Qwen-Coder models that output plain-text JSON for tool calls, this fine-tune enables the model to emit the native
<tool_call>format, which is directly parsed by runtimes like Ollama and llama.cpp. - Agentic Loop Integration: Designed to allow tiny (≤2B parameter) coder models to effectively drive agentic coding loops by correctly interpreting and generating tool-use commands.
- Optimized Training: Trained using bf16 LoRA with assistant-only loss on attention and MLP projections, focusing on tool calls and final answers. The training data includes
NousResearch/hermes-function-calling-v1and syntheticsmolcodetool-use trajectories, ensuring byte-identical training and inference templates.
Use Cases
This model is ideal for developers building agentic coding assistants or applications that require small language models to interact with external tools via structured function calls. It addresses the challenge of integrating smaller code generation models into tool-using workflows, making them more functional and efficient for specific tasks.