SimionZoo
SimionZoo copied to clipboard
simionsoft
A workbench for online model-free Reinforcement Learning on continuous control problems
**Description** Out of Memory exception logged **To Reproduce** Run the last version of the online/offline experiment
Both installers of HerdAgent (Linux/Windows) could and should be built/tested in the Azure pipeline
HerdAgent uses int32.Parse() when receiving a file. This limits the maximum size to a little above 2Gb, which is not ideal and should be very easy to fix
Right now, the implementation clones the online networks periodically as in Deep-CACLA and DQN. We need to implement softUpdate yet
It would be best to save n-dimensional sampled functions for later analysis: -Add n-Dimensional Function Samplers in RLSimion -Add an axis selection and export to image functionalities in SimionLogViewer
Metadata
Owner
Metadata
A workbench for online model-free Reinforcement Learning on continuous control problems