h2oai/h2o-llmstudio ? reverse-engineered prompt

Build me a simple web app for fine tuning large language models, with a clean no code interface so someone who is not a developer can use it.

I want to be able to upload training data, pick a base model, choose common training settings, start a run, and then watch logs and progress in the browser. It should also let me compare runs, see evaluation results, chat with the model to test it, and export the finished model when it is done. Please include support for modern fine tuning approaches like LoRA, and make the app work well on a machine with an NVIDIA GPU.

If there are any current best practice setup steps or dependencies you need, look them up online and follow them. Keep the experience simple, polished, and easy to use, with sensible defaults so I can get started quickly.