general: loss_reduction_func: min model: onnx_path: ${CONFIG_PATH}/ffnnTANH__Point_9x100.onnx input_shape: [-1, 1, 28, 28] data: dataset: MNIST end: 100 specification: epsilon: 0.006 solver: batch_size: 2048 early_stop_patience: 100 alpha-crown: iteration: 100 beta-crown: lr_alpha: 0.1 lr_beta: 0.1 iteration: 25 bab: timeout: 300 branching: nonlinear_split: loose_tanh_threshold: 2