Continue training from checkpoint Centralized Critic

RLlib
1

Hello all.
I wonder which is the correct way to continue training.
the model is Multi agent Centralized Critic.
Id appreciate any help for direction.
Thanks,
Nissim