Develop Cognitive Applications with PowerAI

Big News from SuperComputing 2016 this morning! I am excited to share with you that IBM and NVIDIA teamed up to announce the release of PowerAI, the world's fastest deep learning enterprise solution. PowerAI is an optimized, integrated offering including everything to get started with cognitive applications using Deep Learning, including optimized CPU and GPU high performance libraries and prebuilt optimized versions of the most popular Deep Learning frameworks for GPU accelerated Power servers. The PowerAI release also includes a binary version of IBM Caffe, the popular Caffe deep learning framework enhanced with new algorithms designed by IBM’s Research labs to deliver the most scalable deep learning in the industry. In addition, this offering includes support for scripting languages such as Lua JIT and Python notebooks for ease of use and productive data scientist use.

The new PowerAI solution is the first generation of the Machine Learning and Deep Learning toolkit released for Ubuntu 16.04 and CUDA8 to support the S822LC for HPC server and its high-performance Pascal 100 GPUs with CPU-GPU NVLink offering a high bandwidth, low latency link between CPUs and GPUs. Building on the first releases of the MLDL toolkit, the PowerAI toolkit continues to support GPU accelerated Power servers with PCIe-attached GPUs. Release 2 of the Power MLDL toolkit continues to support easy to install Deep Learning Frameworks for servers with PCIe-attached GPUs running Ubuntu 14.04 and CUDA 7.5.

• ?Caffe, a dedicated artificial neural network (ANN) training environment developed by the Berkeley Vision and Learning Center (BVLC) at the University of California at Berkeley is now available in three versions: the leading edge Caffe development version from Berkeley's BVLC with performance enhancements developed at IBM's research labs, an unmodified BVLC Caffe, and Nvidia's NVCaffe.
• Torch, a framework consisting of several ANN modules built on an extensible mathematics library
• Theano, another framework consisting of several ANN modules built on an extensible mathematics library
• TensorFlow a deep learning Framework developed at Google, reflecting its growing popularity in the Deep Learning Community. (available shortly)

In addition to these binary releases, we are enabling Machine Learning and Deep Learning in the Cloud. We are partnering with NIMBIX to make PowerAI available as a service in the cloud. The Power MLDL solution on NIMBIX allows users to get started with PowerAI with preconfigured ready-to-use Deep Learning engines immediately and while you install servers in your data center. We are also expanding the SuperVessel container cloud for academic users.

We continue to expand the open source ecosystem for Linux on Power for Machine Learning and Deep Learning. We have published code updates reflecting the Power enablement to community repositories and made additional enhancements available on github. We have also published build instructions for the tuned open source Deep Learning frameworks included in PowerAI for those looking to build and further enhance these frameworks on Power.

In addition to the prebuilt optimized releases included in the PowerAI toolkit, we have ported and published NVIDIA docker containers for Power to simplify deploying GPU-accelerated application stacks on Power, additional Deep Learning frameworks (such as Chainer), and tools such as GUIs and scripting languages for Power which can be easily installed from source. 

Get started with PowerAI to develop cognitive applications on Power today and share how you are unleashing the power of deep learning to transform the future of computing in the comments section.