{"knn_weight": 0.1, "mlp_weight": 0.3, "quantize": true}