dsh-fine-tune: Fine-tune local small models with LoRA/QLoRA in DSH

songrui111/dsh-fine-tune (DSH plugin, model inference type, version 0.6.0) consolidates local small model fine-tuning into a unified panel: Import JSONL data → QLoRA/LoRA training → Evaluate accuracy and text similarity → Export adapter or GGUF. It has been fully verified only on Windows 11 + RTX 3060 Laptop 6 GB; Linux is unverified, and the remote SSH backend was only tested using a simulated server. This article details hardware requirements, installation, typical usage, verified test data, and known limitations step-by-step, making it suitable for developers with NVIDIA GPUs who wish to perform domain adaptation on open-source small models from the Qwen/Llama families.