angel
A Flexible and Powerful Parameter Server for large-scale machine learning
What is angel?
A Flexible and Powerful Parameter Server for large-scale machine learning
Overview of angel
What it is
Angel is a high-performance distributed machine learning and graph computing platform based on the Parameter Server philosophy. It is jointly developed by Tencent and Peking University, and is tuned for performance with big data from Tencent. The platform partitions parameters of complex models into multiple parameter-server nodes, and implements machine learning and graph algorithms using model-updating interfaces and functions, along with a flexible consistency model for synchronization. Angel is developed with Java and Scala, and supports running on Yarn. With PS Service abstraction, it supports Spark on Angel.
Key features
- Parameter Server architecture for distributed model training
- Supports both machine learning and graph computing algorithms
- Runs on Yarn and integrates with Spark via PS Service
- Provides a variety of algorithms including traditional ML (LR, SVM, FM, GBDT, LDA) and Spark on Angel algorithms (DeepFM, DCN, GCN, GraphSage, etc.)
- Flexible consistency model for synchronization
Use cases
Angel is suitable for large-scale machine learning tasks that require handling high-dimensional models, such as recommendation systems, click-through rate prediction, graph analytics, and topic modeling. It can be used in both industry production environments and academic research.
Getting started
See the repository README for quick start examples and deployment guides.
License and pricing
The license is not stated in the provided metadata. The project is listed as open source.
Based on the project's public README and repository metadata.