The frameworks and tools being developed for deep learning, machine learning and AI are vast and being updated at a dramatically fast pace. In many cases, these frameworks require different versions and combinations of nVidia drivers, CUDA, and CUDNN. To top this off, GPU drivers and Cuda combinations are difficult to maintain, install and verify.

AI-AMI's intent is to provide an offering of minimalistic, tested and verified AMIs that encompass a breadth of driver, cuda, cudnn and framework combinations. Users can stop trying to build and maintain a base AMI with the required GPU combinations and focus on developing and deploying an AMI ready to develop and train on.