A CUDA Extension of NNabla

This repository provides an official CUDA/cuDNN-accelerated extension of the NNabla deep learning framework.

In order to use it, the default context needs to be changed from 'cpu' to 'cuda.cudnn':

from nnabla.contrib.context import extension_context

ctx = extension_context('cuda.cudnn', device_id=args.device_id)
nn.set_default_context(ctx)

Currently, the installation documentation and usage is integrated into the base NNabla.

FAQ

• Coming soon