model_name: "nm-testing/Qwen1.5-MoE-A2.7B-Chat-quantized.w4a16"
accuracy_threshold: 0.45
num_questions: 1319
num_fewshot: 5
server_args: "--enforce-eager --max-model-len 4096"