I’ve spent the last few months working on MadLab. I wanted something that felt as polished as LM Studio but focused entirely on the training side of the house.
It’s a standalone desktop app (Windows/Linux/macOS) that automates the entire environment setup—GPU detection, PyTorch wheel selection, and venv creation—so you can go from zero to training in a few minutes. Under the hood, it’s managing the trainer logic for LoRA, QLoRA, and DoRA. One thing I’m experimenting with is a built-in Chat Assistant that acts as a fine-tuning coach. It doesn't just run commands; it helps recommend hyperparameters based on your model size and hardware limits. It's 100% open-source. I’d love to get some feedback from this community on the environment automation and the training UI. I'll be around to answer any technical questions!
Show HN: MadLab – A standalone desktop app for local LLM fine-tuning | Heykuki News