Unsloth
Fine-tune LLMs 2× faster with up to 70% less memory.
- Star 数
- 78K
- Fork 数
- 7.2K
- 未关闭 Issue
- 368
- 最近提交
- 2026-10-15
数据来自 GitHub API · 更新于 2026-10-15
Unsloth 是什么？
Unsloth provides hand-optimized kernels for LoRA/QLoRA fine-tuning and reinforcement learning of popular open LLMs, with free Colab notebooks for most models.
主要功能
- Faster LoRA/QLoRA fine-tuning
- GRPO reinforcement learning
- Export to GGUF and Ollama
- Free Colab notebooks
系统要求
- 平台
- Linux · Windows · Colab
- 显卡
- NVIDIA
- 显存
- 3 GB+ depending on model size
如何安装 Unsloth
根据项目 README 整理。开源项目更新很快，请以 GitHub 页面上的最新说明为准。
pip
- 1Install.
pip install unsloth
Google Colab
- 1Open one of the notebooks linked in the README and run all cells.
优势
- Huge speed and memory savings
- Great notebooks
不足
- Multi-GPU support limited in the open version
Unsloth 替代品对比
| 软件 | Star 数 | 许可证 | 平台 | 显存 |
|---|---|---|---|---|
| Unsloth Fine-tune LLMs 2× faster with up to 70% less memory. | 78K | Apache-2.0 | 3 GB+ depending on model size | |
| Axolotl Config-driven LLM fine-tuning for many models and techniques. | 12.6K | Apache-2.0 | — | |
| LLaMA-Factory Unified fine-tuning for 100+ LLMs and VLMs with a no-code web UI. | 75.5K | Apache-2.0 | — | |
| TRL Hugging Face’s library for post-training LLMs: SFT, DPO, GRPO. | 19.5K | Apache-2.0 | — | |
| Kohya’s GUI GUI for training LoRA, DreamBooth and fine-tunes for SD, SDXL and FLUX. | 12.6K | Apache-2.0 | 8 GB+ for SD1.5 LoRA; 12–24 GB for SDXL/FLUX | |
| OneTrainer One-stop trainer for diffusion models with a desktop UI. | 3.2K | AGPL-3.0 | 8 GB+ |
常见问题
Unsloth 免费吗？
是的——Unsloth 是开源软件，采用 Apache-2.0 许可证。商用前请确认软件及模型权重的许可条款。
Unsloth 支持 Windows、Mac 和 Linux 吗？
Unsloth 支持：Linux, Windows, Colab。
运行 Unsloth 需要什么配置？
NVIDIA. VRAM: 3 GB+ depending on model size
如何安装 Unsloth？
1. Install.
Unsloth 有哪些好的替代品？
常见替代品包括 Axolotl, LLaMA-Factory, TRL, Kohya’s GUI，详见本页对比表。