Motion Correction Core Requirment "CPU not avaliable"

Troubleshooting
1

We are running Cryosparc v2 on an Amazon AWS EC2 instance, however I don’t have information such as linux version, cuda version, etc.

The EC2 instance is a 4 core, 64 gb ram, 1 gpu node, and has been properly configured as such during Cryosparc setup.

Importing movies works fine, as do a number of job from particle stack import up through ab-initio generation. However, as we are working with a new data source now, we have need to run motion correction on Cryosparc, and have encountered an issue there.

Any motion correction jobs queued remained queued indefinitely, with the note “CPU not available”. Looking into the metadata section of the job, we see that the job is requesting 6 cores on a 4 core machine, hence it will never start as it’s looking for 2 cores that don’t exist.

Is there a way to change the default requirements of different jobs in Cryosparc beyond the GUI option to select the number of GPUs? Alternatively, is there a work around to get Motion Correction tasks to run on a 4 core machine?

Thanks,
JR

2

Hi @jr10, unfortunately there’s way to do this in the latest version of cryoSPARC. Patch Motion requires 6 CPUs per GPU to run properly.

3

Hi nfrasser,

So, I took a solution from another user on here who had a similar issue with ram requirements (he needed 24gb but only had 16gb) and made an edit to the build.py file for patch motion correction, modifying the number of cpu cores per gpu from 6 to 4. The job ran and seems to have output good results which have classed well so far. However, in doing that, is there a potential that there are issues with the data that I have not yet seen? Thank you for the help!

Best,
jr10

1 Like