RL SAC Training Restoration in Matlab 2024b

Hi, I have a question for RL SAC training in Matlab 2024b.

I trained a model and stopped training after 10000 episode.

Then, I loaded the last model and restarted training.

In this case, the training graph goes down in the beginning and goes up later.

I want to restore the original training status.

To fix this issue, I tried the following approaches:

- Generate replay buffer from the last trained agent and reload it for new training.

- Enable InfoToSave=true, and load the original training options.

But the issue still happened.

Could you please investigate this and let me know the approach to restore the previous training status?