Angel是一个基于参数服务器(Parameter Server)理念开发的高性能分布式机器学习和图计算平台,它基于腾讯内部的海量数据进行了反复的调优,并具有广泛的适用性和稳定性,模型维度越高,优势越明显。 Angel由腾讯和北京大学联合开发,兼顾了工业界的高可用性和学术界的创新性。
Angel的核心设计理念围绕模型。它将高维度的大模型合理切分到多个参数服务器节点,并通过高效的模型更新接口和运算函数,以及灵活的同步协议,轻松实现各种高效的机器学习和图算法。
Angel基于Java和Scala开发,能在社区的Yarn上直接调度运行,并基于PS Service,支持Spark on Angel,集成了图计算和深度学习算法。
欢迎对机器学习、图计算有兴趣的同仁一起贡献代码,提交Issues或者Pull Requests。请先查阅: Angel Contribution Guide
- Angel or Spark On Angel?
- Algorithm Parameter Description
- Angel
- Traditional Machine Learning Methods
- Spark on Angel
- Mailing list: [email protected]
- Angel homepage in Linux FD: https://angelml.ai/
- Committers & Contributors
- Contributing to Angel
- Roadmap
- PaSca: A Graph Neural Architecture Search System under the Scalable Paradigm. WWW, 2022
- Graph Attention Multi-Layer Perceptron. KDD, 2022
- Node Dependent Local Smoothing for Scalable Graph Learning. NeurlPS, 2021
- PSGraph: How Tencent trains extremely large-scale graphs with Spark?.ICDE, 2020.
- DimBoost: Boosting Gradient Boosting Decision Tree to Higher Dimensions. SIGMOD, 2018.
- LDA*: A Robust and Large-scale Topic Modeling System. VLDB, 2017
- Heterogeneity-aware Distributed Parameter Servers. SIGMOD, 2017
- Angel: a new large-scale machine learning system. National Science Review (NSR), 2017
- TencentBoost: A Gradient Boosting Tree System with Parameter Server. ICDE, 2017