We have eventually decided to publish NNabla wheel package (python3.5 and python3.6) with Multi-GPU support. Run the following command to install.
pip install nnabla-ext-cuda92-nccl2-ubuntu16
If you already installed NNabla, uninstall all of it, or start from a clean environment which you create using Anaconda, virtualenv, or pyenv. You should also install OpenMPI and NCCL in addition to CUDA and CuDNN.
Note as the package name indicates, it is only available in ubuntu16.04 now, but you can build then install the Multi-GPU python wheel package in many Linux distributions on various CPU architectures, e.g., Intel and Power architectures. Look at the document for building from the sources.