Hi, I am wondering if there is a way to parallelize & speed up scVI if I have multiple GPUs available?
Thanks,
Sam
Hi, I am wondering if there is a way to parallelize & speed up scVI if I have multiple GPUs available?
Thanks,
Sam
Not at the moment, it’s something we’re looking into though. If you have a really large dataset, you might try increasing batch_size=1024
. We’ve seen results that were comparable there, and it will be relatively faster.
Gotcha, good to know. On my setup scvi is already pretty fast for ~400k nuclei, but I have several GPUs and it would have been nice to take advantage of all of them.
Thanks,
Sam