benchmark/llm_text_generation/code/pegasus/A100/queue_8gpu.yaml

- command:
    - "python scripts/benchmark_one_model.py --model {{ model }} --result-root results/joule --gpu-ids 0 1 2 3 4 5 6 7 --backends vllm --server-images mlenergy/vllm:v0.5.4-openai --request-rates inf --power-limits 400 --max-num-seqs 1536 1280 1024 768 512 320 256 192 128 64 32 --data-dup-factor 20"
  model:
    - codellama/CodeLlama-70b-hf