Caffe2 is a deep learning framework enabling simple and flexible deep learning. Built on the original Caffe, Caffe2 is designed with expression, speed, and modularity in mind, allowing for a more flexible way to organize computation.

Caffe2 aims to provide an easy and straightforward way for you to experiment with deep learning by leveraging community contributions of new models and algorithms. Caffe2 comes with native Python and C++ APIs that work interchangeably so you can prototype quickly now, and easily optimize later. Caffe2 is fine tuned from the ground up to take full advantage of the latest NVIDIA Deep Learning SDK libraries, cuDNN , cuBLAS and NCCL , to deliver high-performance, multi-GPU acceleration for desktop, data centers, and embedded edge devices

Getting Caffe2

Caffe2 has been designed from the ground up to take full advantage of the NVIDIA GPU deep learning platforms. Get started with Caffe2 on your desktop, cloud or datacenter GPU solution:

Caffe2 in the cloud:

Caffe2 on DGX-1

Users of NVIDIA DGX-1 AI supercomputer can download DGX optimized Caffe2 containers via NVIDIA DGX-1 Container Registry.

Caffe2 on desktop and datacenter

To make it easy to install Caffe2 from source, locally on your desktop or datacenter, follow the step-by-step instruction in the Caffe2 GPU-Ready App Quick Start Guide. GPU-Ready Apps guides provides installation recipes that helps you get up and running fast on GPUs.

Visit Caffe2 installation page to learn more about for other ways to get Caffe2, including pre-compiled binaries, docker images.

Caffe2 performance

Caffe2 features built-in distributed training using the NCCL multi-GPU communications library. This means that you can very quickly scale up or down without refactoring your design. Caffe2 delivers near-linear scaling of deep learning training achieving up to 7.7x speed up with 8 GPUs, compared to a single GPU training.

Learn More

For a technical overview of Caffe2 from the authors of Caffe2 check out the following Parallel ForAll blogpost: Caffe2: a High-Performance, Mobile-Focused Deep Learning Open Source Framework from Facebook