It’s time we sped things up!
A key part of the hurdle has been the technical limitation that popular open-source deep learning software frameworks do not run effectively across multiple servers.
Today, IBM Research unveiled breakthrough software and posted record test results showing a leap forward in deep learning performance.
This is a milestone in making deep learning much more practical at scale — to train AI models using millions of photos, drawings or even medical images — by dramatically increasing the speed and making significant gains in image recognition accuracy.
The IBM Research software — “Distributed Deep Learning” library (DDL for AI Developers) — was designed to solve severe performance bottlenecks caused by increasingly fast GPUs trying to synch with each other at once.
“We just invented the jet engine for deep learning.“ Sumit Gupta, HPC & AI Offerings Executive, IBM Cognitive Systems.
The IBM researchers expect this breakthrough will reduce the wait-time associated with deep learning training from days or hours to minutes or seconds, and enable many advances, such as faster, more accurate analysis of medical images or improved fraud detection or enhanced speech recognition in mobile phones and AI assistants.
No longer are the days where only an elite group of AI developers have access to accurate, high-level trained models.
DDL for AI Developers reduces the amount of time it takes to train an AI model; thus, dramatically improving productivity and accuracy of the model AI developers and data scientists can create. With the incorporation of distributed deep learning into IBM PowerAI, IBM is democratizing the availability of high-speed training to enable all AI developers to develop higher accuracy AI models.
IBM Research is delivering a beta version of the deep learning software to IBM Systems, which will make the technology available to IBM PowerAI v4 developers and customers.