Multi GPU support for template picker?

[olibclarke]

Hi,

It would be useful to have multi-GPU support for template picker. Currently there is a workaround - use exposure sets to split the dataset then pick each subset using a different GPU - but native multi GPU support would be a great addition!

Cheers
Oli

[rbs_sci]

@leetleyang and @DanielAsarnow were asking for the same thing 2 years ago.

Blob Picker multiple GPU parallelisation

Hi, In case it hasn’t been mentioned before, it would be nice to be able to parallelise Blob Picker over multiple GPUs, similar to the preceding preprocessing jobtypes. We often resort to splitting the input into n groups to feed into n number of picking jobs–something we also do with the MotionCor2 wrapper–before reconsolidating, but this is rather cumbersome. Cheers, Yang

Good idea for the nudge, though; been a while and no further information forthcoming.

[olibclarke]

Ah I missed that, thanks!

I don’t mind so much for Blob picker - it is fast enough that I rarely find the speed a bottleneck - but template picker can be, particularly when using volume projections as templates

[rbs_sci]

Yes, the original request was for blob, but it expanded to include templates.

When using projections, I usually feed the projections to Select 2D and pick only 3-5 projections to pick with (even with C1, with a 20A lowpass that’s usually enough). And picking with 5 views is quite a bit faster than 20-50. Although if views are wildly different (e.g.: photosystems with very different dimensions and intensity for top/side views) picking twice (one or two top views in one pick run, one or two sides in a second) and thresholding carefully works very nicely too.

[jstagno]

Agreed! This is a much needed feature.