Currently, the model training is only executed on a single CPU core. However, in general, stable_baselines3 and gymnasium support parallelization of the learning process. Apparently, masking makes it a bit more complex (see, eg.g., https://github.com/Stable-Baselines-Team/stable-baselines3-contrib/issues/49). Nevertheless, this would could lower the training time significantly and should be explored more.
What's the problem this feature will solve?
Currently, the model training is only executed on a single CPU core. However, in general,
stable_baselines3
andgymnasium
support parallelization of the learning process. Apparently, masking makes it a bit more complex (see, eg.g., https://github.com/Stable-Baselines-Team/stable-baselines3-contrib/issues/49). Nevertheless, this would could lower the training time significantly and should be explored more.Describe the solution you'd like
Implement a parallel training procedure.