GraphVite accelerates graph embedding with multiple CPUs and GPUs. Taking around 1 minute to learn node embeddings for graphs with 1 million nodes, it enables rapid iteration of algorithms and ideas.
Designed to be scalable, it is capable of processing large-scale graphs, even with limited GPU memory. With only 4 GPUs, it can train node embeddings of a billion-scale graph within one day.
Complete pipelines of node embedding, knowledge graph embedding, and graph & high-dimensional visualization are supported. It is a ready playground for models and evaluation tasks.
|Node Embedding||Knowledge Graph Embedding||Graph & High-dimensional Data Visualization|
Models and Benchmarks
A large collection of models and benchmarks are provided to facilitate fast reproducibility. Choose your favourite model and plug it into your research or development.
Supported models: DeepWalk, LINE, node2vec, TransE, DistMult, ComplEx, SimplE, RotatE, LargeVis, …
Tailored to different needs, flexible interface brings you great user experience, while minimizes the issues you do not care. It is always easy and efficient to integrate GraphVite into your environment, no matter you are using Python or C/C++.
Powerful pre-trained models are released to benefit a wide range of semantic tasks. Serving as a large encyclopedia for intelligent systems, these models are ready to enhance language understanding with factual knowledge of the world.