[model cards] Keep evaluation order in training logs if there's multiple evaluators

[model cards] Keep evaluation order in training logs if there's multiple evaluators #950