File size: 400 Bytes
b10121d
 
 
 
 
1
2
3
4
5
6
- 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:
    - mistralai/Mixtral-8x7B-Instruct-v0.1
    - meta-llama/Meta-Llama-3.1-70B-Instruct