Cori GPU provides several modules to different machine learning and deep learning tools and libraries.
cuDNN is NVIDIA's library of GPU-accelerated primitives for deep neural networks. It is available as the
cudnn module, and is loaded automatically when a
cuda module is loaded.
TensorFlow is an open-source toolkit for machine learning. It is available as the module
tensorflow/gpu-<version>. Users are also encouraged to use GPU-optimized TensorFlow containers provided on NVIDIA's NGC container library. NGC containers can run on Cori GPU using Shifter.
Refer to the full NERSC TensorFlow documentation for more details.
Refer to the full NERSC PyTorch documentation for more details.