seanpoyner/smolcode-coder-sql-1.5b-tools
The seanpoyner/smolcode-coder-sql-1.5b-tools is a 1.5 billion parameter LoRA fine-tune of Qwen2.5-Coder-1.5B-Instruct, optimized for generating native function calls. This model, with a 32768 token context length, enables agentic coding loops by correctly emitting tool calls, addressing a common limitation in small coder models. It is specifically designed to drive agentic coding assistants like smolcode, making it highly effective for tool-use scenarios in code generation.
Loading preview...
Overview
This model, seanpoyner/smolcode-coder-sql-1.5b-tools, is a 1.5 billion parameter LoRA fine-tune of the Qwen2.5-Coder-1.5B-Instruct base model. Its primary purpose is to enable small language models (SLMs) to emit native <tool_call> function calls, which is crucial for driving agentic coding loops. Unlike out-of-the-box Qwen-Coder models that often produce plain-text JSON for tool calls, this fine-tune ensures compatibility with runtimes like Ollama and llama.cpp by generating the correct <tool_call> format.
Key Capabilities
- Native Tool Call Generation: Emits
<tool_call>{"name": ..., "arguments": ...}</tool_call>directly, facilitating seamless integration into agentic workflows. - Agentic Coding Support: Specifically built for
smolcode, an SLM-optimized agentic coding assistant, enhancing its ability to interact with tools. - Efficient Fine-tuning: Utilizes bf16 LoRA (r=16, α=32) with assistant-only loss, focusing training on tool calls and final answers.
- Data-driven Accuracy: Trained on
NousResearch/hermes-function-calling-v1and syntheticsmolcodetool-use trajectories, ensuring high precision for its intended use.
Good For
- Developers building agentic coding assistants that require precise tool-call generation from small models.
- Use cases where a 1.5B parameter model needs to reliably interact with external tools or APIs through structured function calls.
- Environments where efficient, small-footprint models are preferred for code-related tasks involving tool utilization.