We consider a large class of exponential regression models with censored data and propose two modified Fisher scoring methods with corresponding algorithms. These proposed methods improve the Newton-Raphson method in estimating the model parameters. The simulated and real examples are illustrated in aspect of convergence.