# pick a tag from the Run it column of Table 1
IMG=systalyze/systalyze-serve:gemma-4-26b-a4b-it-g7
$ docker pull $IMG
$ docker run -d --name optimized --network host --gpus all \
-v /data/models:/models $IMG
$ curl localhost:8000/health # ready when this returns# finds saturation, then steps load up from 30% (30–90 min)$ docker run --rm --network host -v /data/models:/models \
-v $PWD/results:/results $IMG bench --port 8000 --label optimized
$ docker stop optimized# BF16 checkpoint on stock vLLM, default settings$ docker run --rm $IMG print-baseline > baseline.sh
# add -e VLLM_HTTP_TIMEOUT_KEEP_ALIVE=60 to each docker run in baseline.sh$ bash baseline.sh # serves on :8300$ docker run --rm --network host -v /data/models:/models \
-v $PWD/results:/results $IMG bench --port 8300 --label baseline --max-num-seqs 256# prints the same four measures as Table 1$ docker run --rm -v $PWD/results:/results $IMG \
compare /results/optimized-* /results/baseline-*