🚀 SQFT基础模型：sqft-phi-3-mini-4k-30-base

SQFT基础模型基于特定的源模型，采用了稀疏化方法进行处理，在模型性能和资源利用上有独特的优势。它为相关领域的研究和应用提供了新的选择。

🚀 快速开始

若要获取此模型，请参考 SQFT/run_command/phi-3-mini-4k-instruct/sparse_quantization.sh#11 中的命令。

✨ 主要特性

• 源模型：microsoft/Phi-3-mini-4k-instruct
• 稀疏方法：Wanda
• 稀疏率：30%
• 量化：无

📚 详细文档

模型来源

• 仓库：https://github.com/IntelLabs/Hardware-Aware-Automated-Machine-Learning/tree/main/SQFT
• 论文：
  • SQFT: Low-cost Model Adaptation in Low-precision Sparse Foundation Models
  • Low-Rank Adapters Meet Neural Architecture Search for LLM Compression

引用信息

@inproceedings{munoz-etal-2024-sqft,
    title = "{SQFT}: Low-cost Model Adaptation in Low-precision Sparse Foundation Models",
    author = "Munoz, Juan Pablo  and
      Yuan, Jinjie  and
      Jain, Nilesh",
    editor = "Al-Onaizan, Yaser  and
      Bansal, Mohit  and
      Chen, Yun-Nung",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2024",
    month = nov,
    year = "2024",
    address = "Miami, Florida, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.findings-emnlp.749",
    pages = "12817--12832",
}

致谢

感谢Wanda的工作（论文，代码），它提供了一种简单而有效的剪枝方法。

📄 许可证

本项目采用Apache-2.0许可证。