Useful commands

Run Gemma 4 locally (e.g. for query generation):

docker build . -f Dockerfile.serve -t paperwerk-serve:latest

docker run --gpus all  --shm-size=16g -p 8000:8000 paperwerk-serve:latest /venv/bin/vllm serve google/gemma-4-E4B-it     --max-model-len 65536     --enable-auto-tool-choice       --tool-call-parser gemma4     --tensor-parallel-size 2

or

docker run --gpus all --shm-size=16g -p 8000:8000 \
  -e PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True \
  paperwerk-serve:latest \
  /venv/bin/vllm serve google/gemma-4-31B-it-qat-w4a16-ct \
    --tensor-parallel-size 2 \
    --dtype bfloat16 \
    --max-model-len 32768 \
    --gpu-memory-utilization 0.85 \
    --enforce-eager \
    --max-num-seqs 8 \
    --enable-auto-tool-choice --tool-call-parser gemma4