Best Model Selection During Training

I am currently training a model for 2000 epochs and would like to inquire about the best practices for determining the optimal model. Should I validate the model at certain intervals throughout training, and if so, what would be an effective strategy for identifying the best model?