Can cryoSPARC tolerate a worker node with a more recent CUDA version than the master node? Thanks.
The constraint is that on each worker node, the installed CUDA toolkit must be consistent with the contents of the
cryosparc_worker directory tree that is used to run jobs on that worker node.
If the master node in your case is also a worker node with a supported version
x.y of the CUDA toolkit, but the connected worker has toolkit version
x.z, then the connected worker needs its own
cryosparc_worker tree that was installed (or updated via
cryosparcw newcuda) pointing to toolkit version
OK, thanks. I’ll try installing a worker node with a latter version of CUDA.