===================
input_array:[0. 1. 2. 3.]
===================
shape:(4,)
===================
输入层:model.add(Dense(125,input_shape=(train_x.shape[1],)))
#其中train_x.shape[1]=4
===================
推理部分:y_train_pred_nn = model.predict(input_array)
然后抱错,通过修改shape完成debug,这里模型的(m,n)与numpy的(n,m)反的
# predict_x = input_array.reshape(1,train_x.shape[1])
===================
参考资料