Ask Your Question

3

Runtime API - multiple runs [closed]

energyplus
python
python-plugin
retag add tags

Hi all,

I am working on integrating E+ with OpenAI Gym to train reinforcement learning agents. When using the new runtime API, I am getting an error when trying to run multiple simulations. image description

The documentation says to run clear_callbacks before making multiple runs - but it doesn't make any effect.

Some people ran the simulation in a different process and communicated through pipes but I would like to use new API callback features to simplify the process.

Any hint about what is going on?

Thanks in advance,

asked 2020-09-02 13:25:13 -0500

updated 2020-10-08 12:16:27 -0500

edit flag offensive 0 remove flag reopen merge delete

Closed for the following reason "question is not relevant or outdated" by Sergio Ortiz 2020-09-06 05:53:57 -0500

Comments

Are you running these simulations in parallel?

I am doing RL work too, one simulation at a time on my CPU. And I would like to run multiple simulations at once to speed up training and tuning.

mechyai (2022-03-14 17:15:26 -0500) edit

Yes. With more recent E+ versions I don't get further errors caused by threading.

Sergio Ortiz (2022-03-25 13:04:01 -0500) edit

add a comment see more comments

1 Answer

Sort by

4

After updating to version EnergyPlus 9.4.0 IO Freeze, I notice the runtime API has changed introducing a new StateManager. Using this version and adapting the code, the issue disappears.

Hope this helps

answered 2020-09-06 05:53:34 -0500

edit flag offensive 0 remove flag delete link

Comments

add a comment see more comments