## Run benchmark

### Benchmark sglang
```
python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port 30000
```

```
python3 bench_sglang.py --num-questions 200
```