xTuring

xTuring is an open-source library from Stochastic for fine-tuning large language models such as LLaMA, Falcon and GPT-J on your own data. It targets developers who want a short, simple API instead of writing training loops.

A few lines of Python load a base model, prepare an instruction dataset and start fine-tuning, then the same model object handles inference. It supports LoRA and low-bit quantized variants so training fits on a single GPU, and models can be saved and reloaded locally.

Documentation covers installation, dataset preparation, configuration and the list of supported models, and a playground interface lets you try a tuned model in the browser. Everything runs on your own hardware.