Request: GPU handling (GPU count sanity check)

Would it be possible to add some sort of sanity check on GPU count asked for in the main UI?

I just got asked by a postdoc why a job wasn’t starting (wish they’d asked me yesterday!) and it was because said postdoc typo’d in the number of GPUs box and asked for 19 GPUs on a 10 GPU system. Rather than throw an error, it just sat waiting for GPUs that would never be…

@rbs_sci Thanks for raising this issue. Did it arise on a CryoSPARC instance that uses the CryoSPARC-builtin scheduler, not an external workload manager?

Yes, CryoSPARC’s built in scheduler.