GraphLab Create is an extensible machine learning framework that enables developers and data scientists to easily build and deploy intelligent applications and services at scale. It includes distributed data structures and rich libraries for data transformation and manipulation, scalable task-oriented machine learning toolkits for creating, evaluating, and improving machine learning models, data and model visualization for all aspects of development, and a client to define and deploy both distributed batch jobs to Dato Distributed™ as well as real-time machine learning services to Dato Predictive Services™. It is designed for end-to-end developer productivity, scale, and the variety and complexity of real-world data.
Create sophisticated models
Start with state-of-the-art toolkits, including implementations for deep learning, ranking-optimized factorization machines, topic modeling, graph analytics, linear models, clustering, and nearest neighbors. Then go as deep as you need, tuning exposed hyperparameters to customize your models.
Begin with your task rather than a research paper. Use human-intuitive machine learning abstractions in our toolkits. These toolkits are named for their use and offer default parameters and baseline models so your first application comes together fast.
Interact with terabytes of data, blazingly fast, even on a laptop. SFrame, the out of core tabular data structure is columnar, distributed, and on-disk making is the most scalable data frame built for machine learning.
Explore and explain everything
GraphLab Create’s visualization “Canvas” provides exploration and visualization of big data and sophisticated machine learning models. Easily explain your results.
Deploy in seconds
Deployment clients for Predictive Services and Distributed are included. Same code deploys on remote environments as a distributed batch job or as services. Eliminate the lengthy rewrites.
Extend, to infinity and beyond
API is extensible in multiple ways. Implement data parallel and graph parallel algorithms as UDFs in Python or build your own efficient algorithms using the C++ SDK. Implement your own scalable brilliance.