NVIDIA and Facebook are joining forces to advance the world of artificial intelligence. New to developers’ arsenal is the open-source deep learning framework Caffe2, which NVIDIA and Facebook will use to create large-scale distributed training scenarios, as well as build machine learning applications for today’s devices.
Facebook has been working with the open source community to build a framework that can handle the challenges that come with training and deploying AI models associated with massive amounts of data or supercomputers. Facebook said it developed Caffe2 to overcome these challenges, and the company open sourced its first production-ready release of the framework yesterday.
Caffe2 is a deep learning framework which delivers “near-linear scaling of deep learning training with 57x throughput acceleration on eight networked Facebook Big Basin AI servers with 64 NVIDIA Tesla P100 GPU accelerators,” according to an NVIDIA blog. With this, developers should be able to “train and iterate AI models faster than ever,” the NVIDIA team wrote.
As part of Facebook and NVIDIA’s collaboration, the NVIDIA DGX-1 AI supercomputer will be the first AI system to offer Caffe2 with a software stack for deep learning. The collaboration also means Caffe2 takes full advantage of the NVIDIA GPU deep learning platform and it uses the latest NVIDIA deep learning SDK libraries. These libraries include cuDNN, cuBLAS and NCCL. With these libraries, users can focus on developing their AI powered applications, according to NVIDIA.