Run loadtest with tabby on modal GPUs Steps: Adjust TABBY_API_HOST in run.sh to match your modal deployment url. Add models you're interested in to benchmark at end of run.sh Run run.sh, output will be appended to record.csv