hiyouga/LlamaFactory ? reverse-engineered prompt

I want to build a comprehensive toolkit for fine-tuning large language models. The goal is to make it super easy for people without deep coding skills to customize their own AI. It should have a clean web interface, maybe using Gradio, where you can select a base model, choose a dataset, and start the training process with just a few clicks.

The tool needs to support a wide range of popular models like Llama, Gemma, Mistral, and Qwen. It should also include various efficient fine-tuning methods like LoRA and QLoRA so it can run on consumer-grade GPUs. It's not just about training; it should also let me test the model afterwards and maybe even deploy it with an OpenAI-compatible API endpoint. You can call it "LLaMA Factory".