run_eval_alpaca_eval.sh

export DEBUG=1

reference_models="microsoft/WizardLM-2-8x22B,Qwen/Qwen1.5-110B-Chat,Qwen/Qwen1.5-72B-Chat,meta-llama/Llama-3-70b-chat-hf,mistralai/Mixtral-8x22B-Instruct-v0.1,databricks/dbrx-instruct"

python generate_for_alpaca_eval.py \
    --model="Qwen/Qwen1.5-72B-Chat" \
    --output-path="outputs/Qwen-72B-round-1_MoA-Lite.json" \
    --reference-models=${reference_models} \
    --rounds 1 \
    --num-proc 1

alpaca_eval --model_outputs outputs/Qwen-72B-round-1_MoA-Lite.json --reference_outputs alpaca_eval/results/gpt4_1106_preview/model_outputs.json --output_path leaderboard