Assume the weight has been updated by n times, then the learning rate will
 be base_lr * factor^^(floor(n/step))
- step int, schedule learning rate after n updates
- factor number, the factor for reducing the learning rate