Dato Core Open Source

Dato Core, the fast, scalable engine of GraphLab Create is now open source.

The Dato Core project provides the complete implementation of the following:

  • SFrame v1 and v2 format reader and writers
  • SFrame query evaluator
  • SGraph
  • Graph Analytics implementations
  • The implementations of the C++ SDK (including gl_sframe, gl_sarray, and gl_sgraph)
  • Sketch implementations
  • Serialization
  • flexible_type (efficient runtime typed object)
  • C++ interprocess communication library (used for C++ to Python communication)
  • PowerGraph's RPC implementation

Dato Core C++ and Dato Core Python Client code is available under the AGPL and BSD license, respectively. See our License Policy or the Dato Core GitHub repo for details.



What is GraphLab Create?

GraphLab Create is a Python library that includes the following:

  • a simple Python interface on top of a fast C++ engine
  • out-of-core support for graphs and tabular data with SFrame and SGraph
  • scalable machine learning modules such as boosted decision trees, deep learning and text analytics
  • auto-tuning of parameters for classification and recommendation models
  • built-in integration with SPARK, Hadoop, Apache Avro and OBDC connectors
  • ability to deploy a model to a RESTful predictive service
  • seamless integration with AWS EC2 and S3 for scalable compute and storage
  • visualization of data, jobs and predictive services with GraphLab Canvas
  • developer SDK to enable extensibility in algorithms, data connectors and more on top of the C++ engine


Announcement

On October 15, 2014, at Strata + HadoopWorld, we announced the general availability of our flagship product, GraphLab Create 1.0,

We saw an opportunity to propel the field of data science forward and to make machine learning more powerful for data scientists and more accessible to software engineers. The team learned from the academic, open source PowerGraph project and built our flagship product, GraphLab Create. We took power and speed to the next level with innovations like SFrames and SGraph plus made the entire development workflow more usable with a simple Python API, advanced machine learning toolkits and the ability to deploy predictive services directly to the Cloud. With GraphLab Create, I really believe we can Unleash Data Science from inspiration to production.”


- Carlos Guestrin, CEO, Dato

What happened to GraphLab.org? GraphLab.org was a website focused on the GraphLab PowerGraph and GraphChi academic, open source projects. We have now consolidated our web properties to Dato.com and forum.dato.com . All of the great information previously available on GraphLab.org is now accessible at the GraphLab PowerGraph github repo .