Non-uniform refinement failed after clearing memory space of Linux workstation

Hi Nonuniform refinement failed and we got the error message (see attached file). It was funny in that non-uniform refinement was working perfectly before. But our workstation ran out of the memory space. Once we cleared it and resumed work, then Non-uniform refinement started showing error message. Clearing memory space may not be related to this error but want to let you know about it as it may be related.
Thanks for your help
Best


Yuro

Hi @yurotakagi,

This error message suggests a problem with the CUDA configuration - can you please confirm what version of CUDA is installed? Please refer to our guide for additional details.

- Suhail

Hi Suhail
Thanks for your prompted reply. CUDA version is 11. Keep in mind that cryosparc was working fine a week ago. Only things I did were (i) clear out the memory space and (ii) update Centos OS.
Thanks for your help
Yuro

Hi @yurotakagi,

Because of a cryoSPARC code dependency that doesn’t yet support CUDA 11, cryoSPARC currently doesn’t support it either - we recommend running CUDA 10.2 until a cryoSPARC update with support is available.

If possible, are you able to downgrade?

- Suhail

Hi Suhail

As I told you before, we have been using the same CUDA and cryosparc was working without any problem until recently. So, your answer does not make sense to me.

Hi @yurotakagi,

What version of cryoSPARC are you running?

- Suhail

The latest version, which is V2.15.0

Hi @yurotakagi,

I’m not sure how it managed to work previously, but we don’t support CUDA 11 at the moment. You’ll be able to run jobs successfully after downgrading.

- Suhail

Dear Suhail

It looks like the workstation has CUDA 10, 10.2 and 11. Probably it was working via 10.2. But somehow lost accessing to it. I am not Linux expert. Do you have any idea what is going on?
Thanks for your help
Yuro

Dear Suhail

I completely removed all Cuda and re-installed Cuda 10.2. Non-uniform refinement as well as patch motion correction (though the error message was different) started working again. I don’t know exactly what happened. But I got the error messaging stating “Unable to handle kernel paging xxxx”. This error was gone once I reinstalled Cuda. So, it’s possible that something happened that corrupted Cuda. I am bring this up here - so someone else may have a similar experience.

Best
Yuro