A guide to building long-term compounding knowledge infrastructure. See details on GitHub .

Tunix

Tunix is a JAX-native post-training library for LLMs providing efficient fine-tuning, RL training, and distillation tools.

Overview

Tunix is a JAX-native library for post-training large language models. It aims to streamline fine-tuning, reinforcement learning, and distillation workflows while providing scalability and TPU-native optimizations.

Key features

  • Support for full-weight fine-tuning and parameter-efficient fine-tuning (LoRA/Q-LoRA).
  • Reinforcement learning algorithms including PPO, GRPO and token-level GSPO, and preference fine-tuning with DPO.
  • Modular, composable components and examples for building reproducible training pipelines.
  • Optimizations for distributed training and TPU execution.

Use cases

  • Research and engineering for post-training LLMs and knowledge distillation.
  • Large-scale fine-tuning and RL experiments on TPUs or multi-GPU clusters.
  • Educational examples and reproducible training recipes.

Technical details

  • Built on JAX/Flax, compatible with common models and training paradigms, and provides comprehensive example notebooks.
  • Supports installation from PyPI or running directly from GitHub source, licensed under Apache-2.0.

Comments

Tunix
Resource Info
Author google
Added Date 2025-10-02
Open Source Since 2025-09-29
Tags
Open Source Framework Fine-tuning ML Platform Training