A collection of various resources, examples, and executables for the general NREL HPC user community's benefit. Use the following website for accessing documentation.
Following some discussion with Haley and Olivia, removing CUDA_VISIBLE_DEVICES to avoid confusion as it is generally unnecessary with Slurm. SLURM_JOB_GPUS will reflect the real GPU device IDs in a given job if are multiple users running jobs on one node, so that variable should be preferred anyway.