decreasing the runtime for pairtree clustervars

[ahgillmo]

Hello.

I was wondering if there are any suggestions to decrease the runtime of pairtree clustervars.

I have increased the parallel count from 2-20 and have changed the --cpus-per-task (on slurm HPC) but have not decreased the runtime. Any other suggestions would be welcome.

Thanks for your time, Aaron

[ethanumn]

Hi Aaron -

There are a few ways you can decrease the runtime of the Pairtree clustervars program. Some Pairtree specific ideas to do this are:

• Set the argument --model equal to linfreq, as this clustering method does not require computing pairwise relationships before clustering can begin
• Lower the number of --chains (Gibbs sampling chains)
• Lower the number of --iterations (number of samples each Gibbs sampling chain takes)

Changing these parameters can impact how well the clustering performs.

Please let me know if this helps, Ethan