DynMoE-Phi-2-2.7B

<h2 align="center"> <a href="https://arxiv.org/abs/2405.14297">Dynamic Mixture of Experts: An Auto-Tuning Approach for Efficient Transformer Models</a></h2> <h5 align="center"> If our project helps you, please give us a star ⭐ on <a href="https://github.com/LINs-lab/DynMoE">GitHub</a> and cite our paper!</h5> <h5 align="center">

📰 News

• [2025.01.23]: 🎉 Our paper is accepted to ICLR 2025!
• [2024.05.25] 🔥 Our checkpoints are available now!
• [2024.05.23] 🔥 Our paper is released!

😎 What's Interesting?

Dynamic Mixture of Experts (DynMoE) incorporates (1) a novel gating method that enables each token to automatically determine the number of experts to activate. (2) An adaptive process automatically adjusts the number of experts during training.

Top-Any Gating

<video controls src="https://i.imgur.com/bLgNaoH.mp4" title="Top-Any Gating"></video>

Adaptive Training Process

💡 Model Details

• 🤔 DynMoE-Phi-2 is a MoE model with dynamic top-k gating, finetuned on LanguageBind/MoE-LLaVA-Phi2-Stage2.
• 🚀 Our DynMoE-Phi-2-2.7B has totally 5.3B parameters, but only 3.4B are activated! (average top-k = 1.68)
• ⌛ With the DynMoE tuning stage, we can complete training on 8 A100 GPUs within 2 days.

👍 Acknowledgement

We are grateful for the following awesome projects:

• tutel
• DeepSpeed
• GMoE
• EMoE
• MoE-LLaVA
• GLUE-X

🔒 License

This project is released under the Apache-2.0 license as found in the LICENSE file.

✏️ Citation

@misc{guo2024dynamic,
      title={Dynamic Mixture of Experts: An Auto-Tuning Approach for Efficient Transformer Models}, 
      author={Yongxin Guo and Zhenglin Cheng and Xiaoying Tang and Tao Lin},
      year={2024},
      eprint={2405.14297},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

Star History