Repolode
Submit
stochasticai

stochasticai/xTuring

A Python library for fine-tuning, evaluating and running open-source LLMs such as LLaMA, GPT-J and GPT-OSS on your own data, locally or in a private cloud.

View on GitHubWebsite
Stars
2,677
Forks
209
Open issues
10
Last commit
yesterday

What is it?

xTuring is a Python library for fine-tuning open-source large language models on your own data. It covers data preparation, training and inference through a single high-level API, so developers can adapt models without writing their own training loops.

It supports a range of model families, including GPT-OSS, LLaMA and LLaMA 2, Qwen3, Mistral, MiniMax M2, GPT-J, GPT-2, DistilGPT-2 and Mamba. Fine-tuning can use LoRA or low-precision INT8 and INT4 loading to reduce memory and compute requirements. Models can be evaluated with perplexity, and a Gradio-based UI and a CLI are provided for chatting with fine-tuned models.

The library is designed to run locally or inside a private VPC, so training data does not need to leave the user's infrastructure.

Key features

  • 01Unified API for data preparation, fine-tuning and generation
  • 02LoRA, INT8 and INT4 variants to reduce memory and compute cost
  • 03Perplexity-based evaluation for causal language models
  • 04Dataset generation from self-instruct and commercial model APIs
  • 05CLI chat command and Gradio UI playground for fine-tuned models

Use cases

  • Adapting an open-source LLM to a company's private documents without sending data to third parties
  • Fine-tuning a small model like Qwen3 0.6B on a laptop for quick experiments
  • Comparing fine-tuning approaches such as LoRA versus INT8 on the same dataset
  • Generating instruction datasets from a local model for further training

Quick start

  1. 1

    Install xTuring from PyPI

    pip install xturing
  2. 2

    Run the small CPU-friendly fine-tuning example with the Qwen3 0.6B LoRA model

Things to know

  • Fine-tuning larger models such as GPT-OSS requires significant GPU resources.
  • xTuring requires transformers>=4.36.0 and must not be upgraded to transformers 5.x yet.

Topics

#llm#fine-tuning#lora#quantization#gen-ai#language-model#peft#deep-learning#adapter#finetuning

FAQ

What is xTuring?
A Python library for fine-tuning, evaluating and running open-source LLMs such as LLaMA, GPT-J and GPT-OSS on your own data, locally or in a private cloud. xTuring is a Python library for fine-tuning open-source large language models on your own data. It covers data preparation, training and inference through a single high-level API, so developers can adapt models without writing their own training loops.
Is xTuring free to use? What license does it have?
xTuring is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is xTuring still maintained?
Actively maintained. The last commit to xTuring was on Nov 5, 2026.
What can I use xTuring for?
Adapting an open-source LLM to a company's private documents without sending data to third parties Fine-tuning a small model like Qwen3 0.6B on a laptop for quick experiments Comparing fine-tuning approaches such as LoRA versus INT8 on the same dataset Generating instruction datasets from a local model for further training
How do I install xTuring?
Install xTuring from PyPI: pip install xturing Run the small CPU-friendly fine-tuning example with the Qwen3 0.6B LoRA model
Where can I get xTuring?
The source code and installation instructions are on GitHub: https://github.com/stochasticai/xTuring

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects