Pretrained model with different number of output

Hi all, I am using mxnet to train landmark detectionusing my custom network (mx.gluon.nn.HybridSequential()). I trained it on widerface dataset that has 5 outputs. Now, I want to use the same network but different number of output likes 68 outputs (last FC layer) to finetune on my dataset. Is it possible ? How can I do it. Just using lenet arch as example