After we update some cryosparc instance to v2.15.0, we found some strange error that cryosparc can only assign job on #0 GPU.
The first one is a standalone machine, we found although we assigned 4GPU on the job, the job only assign #0 gpu.
The second one is a cluster, just update from v2.9.0 to v2.15.0 these days. They assigned jobs via slurm queue on 2 nodes, however multipule jobs are submitted to one node one gpu card, and run on specific gpu not work.
Is there any suggestions about that issue? Thanks!
In your first screenshot, how soon after launching the job did you take the screenshot? Did the job eventually allocate the other four GPU’s? Can you send the entire contents of the overview tab for the first job?
Similarly, for the second screenshot of jobs, can you also send the entire contents of the overview tab for each job? This will help me determine what resources cryoSPARC asked for, and when the respective jobs initially created a GPU context on the machines.
Can you confirm that this was working in previous versions of cryoSPARC?