Skip to content

Swift applications#

Some optimized versions of common applications are provided for the Swift cluster. Below is a list of how to utilize these applications and the optimizations for Swift.


Many are available as part of the Modules setup.


TensorFlow has been built for the AMD architecture on Swift. This was done by using the following two build flags.


This version of TensorFlow can be installed from a wheel file:

pip install --upgrade --no-deps --force-reinstall /nopt/nrel/apps/wheels/tensorflow-2.4.2-cp38-cp38-linux_x86_64-cpu.whl

Currently, this wheel is not built with NVIDIA CUDA support for running on GPU.

TensorFlow installed on Swift with Conda may be significantly slower than the optimized version