benchmark/llm_text_generation/code/pegasus/H100/queue_1gpu.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-rates inf --power-limits 700 --max-num-seqs 1536 1280 1024 768 512 320 256 192 128 64 32 --data-dup-factor 10"
  model:
    - codellama/CodeLlama-7b-hf
    - codellama/CodeLlama-13b-hf
    - bigcode/starcoder2-3b
    - bigcode/starcoder2-7b
    - bigcode/starcoder2-15b
    - google/codegemma-1.1-2b
    - google/codegemma-7b