distil-lfm25-shellper
A fine-tuned version of LiquidAI/LFM2.5-350M for multi-turn shell command execution via tool calling, trained using the distil labs platform.
This model converts natural language requests into bash commands via structured tool calls, based on the Berkeley Function Calling Leaderboard Gorilla file system task.
Results
| Metric | Teacher (120B) | LFM2.5-350M Base | LFM2.5-350M Tuned |
|---|---|---|---|
| Tool Call Equivalence | 97.03% | 61.4% | 98.0% |
| ROUGE | 94.42% | 91.8% | 99.4% |
The tuned 350M model exceeds the 120B teacher by nearly a full percentage point.
Training Details
| Parameter | Value |
|---|---|
| Base model | LiquidAI/LFM2.5-350M |
| Teacher model | GPT-oss-120B |
| Task type | Multi-turn tool calling (closed-book) |
| Training data | distil-labs/distil-SHELLper |
| Training method | SFT with LoRA |
| Platform | distil labs |
Training Progress
| Epoch | Tool Call Equivalence |
|---|---|
| 0 (base) | 61.4% |
| 1 | 98.0% |
| 2 | 97.0% |
| 3 | 98.0% |
| 4 | 98.0% |
Usage
This model uses the LFM2.5 tool calling format with <|tool_call_start|> and <|tool_call_end|> tags:
<|tool_call_start|>[function_name(arg1="value1", arg2=42)]<|tool_call_end|>
Deployment
The model works with Ollama, vLLM, llama.cpp, or any inference runtime that supports Safetensors. For quantized deployment, use the GGUF, ONNX, or MLX variants of the base model as a starting point.
Blog Post
For the full writeup, see: Fine-Tuning Liquid's LFM2.5: Accurate Tool Calling at 350M Parameters
License
This model is licensed under the LFM Open Model License v1.0.