Multiple jobs on a single GPU

Hi, we are running cryosparc on a quadro rtx 8000 with 48 GB memory. Since cryosparc needs far less memory even with large box sizes, we were wondering if we can oversubscribe the GPU with at least two jobs using slurm and mps. Is there a way to cap the GPU memory on the cryosparc side for a job to avoid crashes due to potential memory spikes (which seem to occur sometimes)?