The accelerated failure time model provides an interesting alternative for survival analysis if the proportional hazards assumption is not fulfilled.
Please provide an implementation similar to that of the parametric model e.g. including prediction profiles.
Winfried Koch