Stable-baselines: Update of the learning parameters after the restore of a checkpoint.

[AGPX]

Hi,

I have to train a network (via PPO2) in different sessions, so I use the checkpoint to load and restore the model. However, in subsequent sessions, it may be necessary to update some learning parameters (such as learning_rate), but unfortunately they are saved with the model. Is there a way to update them after loading the model (simply assigning them to the model doesn't seem to work)?

[Miffyli]

Try using load_parameter function (example in docs). I.e. Create a new agent with new learning rate, and then use load_parameters to load the parameters from previous session.