benchmark/llm_text_generation/chat/pegasus/H100/queue_2gpu.yaml

- command:
    - "python scripts/benchmark_one_model.py --model {{ model }} --result-root results/joule --gpu-ids {{ gpu }} --backends vllm --server-images mlenergy/vllm:v0.5.4-openai --request-rate inf --power-limits 700 --max-num-seqs 1536 1280 1024 768 512 320 256 192 128 64 32 --data-dup-factor 4"
  model:
    - google/gemma-2-27b-it