FBTT-Embedding：基于Tensor Train的嵌入表压缩库

FBTT-Embedding：基于Tensor Train的嵌入表压缩库

FBTT-Embedding This is a Tensor Train based compression library to compress sparse embedding tables used in large-scale machine learning models such as recommendation and natural language processing. We showed this library can reduce the total model size by up to 100x in Facebook’s open sourced DLRM model while achieving same model quality. Our implementation is faster than the state-of-the-art implementations. Existing the state-of-the-art library also decompresses the whole embedding tables on the fly therefore they do not provide memory reduction during runtime of the training. Our library decompresses only the requested rows therefore can provide 10,000 times memory footprint reduction per embedding table. The library also includes a software cache to store a portion of the entries in the table in decompressed format for faster lookup and process. 项目地址: https://gitcode.com/gh_mirrors/fb/FBTT-Embedding

1. 项目基础介绍及编程语言

FBTT-Embedding 是由Facebook研究团队开发的一个开源项目，旨在为大规模机器学习模型中的稀疏嵌入表提供基于Tensor Train（张量乘积）的压缩方案。该库可用于推荐系统和自然语言处理等领域。项目主要使用C++进行核心算法实现，同时提供Python接口以方便用户使用。

2. 核心功能

• 压缩稀疏嵌入表：通过Tensor Train结构，FBTT-Embedding能够显著减小嵌入表的大小，据报道在Facebook的开源DLRM模型中可以达到100倍的压缩率。
• 保持模型质量：在压缩的同时，能够保持模型的原始质量不受影响。
• 内存和速度优化：与现有技术相比，该库在运行时只解压请求的行，大幅减少了内存占用，并且提高了处理速度。
• 软件缓存：包括一个软件缓存机制，用于存储部分最频繁访问的解压嵌入向量，以加快查找和处理速度。

3. 最近更新的功能

由于项目存档，最新的更新可能不会反映在官方文档中。但根据可用的信息，以下是一些可能包含在最近更新中的功能：

• 性能优化：持续的性能优化，确保在最新的硬件和软件环境中达到最佳运行效果。
• 易用性改进：对API和文档的改进，使得库更加易于使用和理解。
• 错误修复：修复了在之前版本中发现的问题，提高了库的稳定性和可靠性。

请注意，项目的具体更新内容应参考官方的Release Notes或ChangeLog，这里提供的内容是基于项目描述和文档的概述。

FBTT-Embedding This is a Tensor Train based compression library to compress sparse embedding tables used in large-scale machine learning models such as recommendation and natural language processing. We showed this library can reduce the total model size by up to 100x in Facebook’s open sourced DLRM model while achieving same model quality. Our implementation is faster than the state-of-the-art implementations. Existing the state-of-the-art library also decompresses the whole embedding tables on the fly therefore they do not provide memory reduction during runtime of the training. Our library decompresses only the requested rows therefore can provide 10,000 times memory footprint reduction per embedding table. The library also includes a software cache to store a portion of the entries in the table in decompressed format for faster lookup and process. 项目地址: https://gitcode.com/gh_mirrors/fb/FBTT-Embedding