Change Sigmoid output to probability

Question

I already trained a neural network with the last layer using sigmoid. If I can not retrain the network with softmax, can I change the final predictions as probability? Now t