These are chat archives for thunder-project/thunder

Sep 2015
Sep 17 2015 14:07
I finally managed to get Thunder running on my local cluster and I was asked to benchmark it by doing PCA on a 4D dataset. The dataset being the imaged activity of multiple brain planes ( making the volume of the whole brain) changing over time. Is that even possible to do in Thunder ? I searched of it in the documentation and Thunder.PCA only works with 2D matrices right ?
Sean M
Sep 17 2015 15:48
Our local HPC cluster at NIH ( just upgraded hardware and moved from using PBS to Slurm for cluster management and job control. We have been experimenting with Thunder locally for fMRI data analysis but never really got anything to work robustly on the old Biowulf cluster. I was wondering if anyone using Thunder may have experience (positive or negative) with using magpie ( since it seems have scripts tailored for Slurm to start up Spark. Thanks.