Does 'traingd' function in nntraintool implement plain vanilla gradient descent? Or does it change the learning rate adaptively?

hiddenlayersize = [5 5];
net = fitnet(hiddenlayersize,'traingd');