#!/bin/bash
set -uo pipefail
root=/home/lab/bench/ornith/modular/shared
export CUDA_VISIBLE_DEVICES=''
python3 "$root/quality.py" --base http://127.0.0.1:8120/v1 --model ornith-local --out "$root/quality-llama.json"
/home/lab/venvs/vllm/bin/python "$root/bench.py" --base http://127.0.0.1:8120/v1 --model ornith-local --out "$root/llama-baseline" --lengths 4096,32768,131072,262144 --concurrency 1,3 --repeats 5 --gpu GPU-56b9b8f1-048e-1abc-752c-729769819c5b
