LLM-OS-Models/LFM2.5-1.2B-Terminal-SFT-2Epoch-Unsloth

Overview

LLM-OS-Models/LFM2.5-1.2B-Terminal-SFT-2Epoch-Unsloth is a 1.2 billion parameter model, fine-tuned from LiquidAI/LFM2.5-1.2B-Instruct using Unsloth SFT over 2 epochs. Its primary function is to automate terminal operations by generating JSON-formatted commands based on user requests and current terminal status. The model is optimized for cost-effective and fast inference in specific acceleration paths.

Key Capabilities

• Terminal Automation: Generates subsequent terminal commands in JSON format, including analysis, plan, and keystrokes, to automate tasks.
• JSON Output: Designed to produce structured JSON output for terminal actions, though it has a 47.2% valid JSON rate, requiring parsing verification.
• Conservative Command Generation: Tends to issue fewer incorrect commands, prioritizing accuracy over recall.
• Efficiency: Achieves a fast inference speed of approximately 0.083 seconds per step, making it suitable for iterative evaluation and RL experiments.

Good For

• Terminal Automation Development: Ideal for developers building tools that require automated terminal interaction and command generation.
• RL Experimentation: Its efficiency and base performance make it a strong candidate for reinforcement learning (RL) experiments focused on improving terminal command accuracy and JSON validity.
• Cost-Efficient Inference: Suitable for scenarios where fast inference at a specific size and cost point is critical.

Limitations

• Lower Recall: May omit some necessary commands due to a relatively low recall score.
• JSON Format Failures: Requires parsing validation and potential retries due to a significant percentage of invalid JSON outputs.
• Specialized Use: Not intended for general conversation or broad reasoning tasks; performance is guaranteed only for terminal automation.