|Apache SINGA Reaches Top Level Status|
|Written by Kay Ewbank|
|Tuesday, 12 November 2019|
Apache SINGA has been promoted to a top level project. SINGA is an open source distributed, scalable machine learning library originally developed at the National University of Singapore.
SINGA was submitted to the Apache Incubator in March 2015, but has now been accepted as a top level project. It is the first distributed deep learning Apache project to become a Top Level Project.
SINGA is designed to enable the training of large-scale machine learning (especially deep learning) models over a cluster of machines with large datasets. It has a programming model based on the layer abstraction, and a variety of popular deep learning models are supported, namely feed-forward models including convolutional neural networks (CNN), energy models like restricted Boltzmann machine (RBM), and recurrent neural networks (RNN). SINGA can be used to run synchronous, asynchronous and hybrid training frameworks, and has features to run training in parallel by partitioning on batch dimension, feature dimension or hybrid partitioning.
The developers have incorporated a range of optimization techniques to make the training faster and to enable it to be scaled out, with optimizations on efficiency, memory, communication and synchronization. The developers are currently working on SINGA-lite for deep learning on edge devices with 5G, and SINGA-easy for making AI usable by domain experts (without deep AI background).
The SINGA team is working to make the library easier to use, and regularly adds new features. For example, SINGA has a sub-component called SINGA-auto, which provides AutoML features like automatic hyper-parameter tuning.
Beng Chin Ooi, Distinguished Professor of National University of Singapore, who was one of the original team that developed SINGA, says:
"It is essential to scale deep learning via distributed computing as the deep learning models are typically large and trained over big datasets, which may take hundreds of days using a single GPU."
SINGA is available on GitHub.
or email your comment to: firstname.lastname@example.orgAp