Visualisation of the policy

[MichielKempkens]

Question ❓

Dear @AlejandroCN7,

Is there any way to visualise the policy in some way with sinergym? I want to visualise/explain the decisions the agent took in some states. I mainly use PPO and DQN.

Kind regards,
Michiel

Checklist

• I have read the documentation (required)
• I have checked that there is no similar issue in the repo (required)

📝 Please, don't forget to include more labels besides question if it is necessary.

[AlejandroCN7]

Hi @MichielKempkens,

You can load the trained agent and execute it using Sinergym logger to get a CSV file of interactions between agent and environment. Then, you can visualize that information using matplotlib (for example). Take a look in 'scripts/load_agent.py' (you don't need to use that script if you don't want, but as example to see how to load a model).

Recently we have included WandB compability, see #306 (documentation updated too) in case you want to use it to monitor training data in real time or record evaluations there.