1.5B Model Trained to Write Shell Commands on a Laptop CPU
User trains a 1.5B Qwen2.5-Coder model to generate shell commands. It operates on a laptop.
A user has trained a 1.5B Qwen2.5-Coder model on 125,000 natural language/command pairs to avoid searching for tar commands. The model runs on a laptop and achieves a score of 0.620 on InterCode-ALFA. Additionally, it includes a static safety checker to prevent potentially harmful commands.