Hi,
I have updated my install from R2021a to R2022a. Using the RL toolbox when running the episode manager with the following code in R2022a, when I go to stop the training early, via "Stop Training" in episode manager, the training does not stop, it seems the only way to actual stop the current training early is via the "stop" button on the "run" tab in live editor. Has anyone encounered this in R2022a? and will the agent buffer still be saved via just stopping the script rather than stopping training via EM?
T = 0.066;
maxepisodes = 100;
maxsteps = ceil(T/Ts_agent)
trainingOpts = rlTrainingOptions(…
‘MaxEpisodes’,maxepisodes, …
‘MaxStepsPerEpisode’,maxsteps, …
‘StopTrainingCriteria’,"EpisodeCount",…
‘StopTrainingValue’,maxepisodes,…
‘ScoreAveragingWindowLength’,100,…
‘SaveAgentCriteria’,"EpisodeCount",…
‘SaveAgentValue’,maxepisodes,…
‘SaveAgentDirectory’,"savedAgents",…
‘Verbose’,true,…
‘Plots’,’training-progress’)

Cheers,

PatrickHi,
I have updated my install from R2021a to R2022a. Using the RL toolbox when running the episode manager with the following code in R2022a, when I go to stop the training early, via "Stop Training" in episode manager, the training does not stop, it seems the only way to actual stop the current training early is via the "stop" button on the "run" tab in live editor. Has anyone encounered this in R2022a? and will the agent buffer still be saved via just stopping the script rather than stopping training via EM?
T = 0.066;
maxepisodes = 100;
maxsteps = ceil(T/Ts_agent)
trainingOpts = rlTrainingOptions(…
‘MaxEpisodes’,maxepisodes, …
‘MaxStepsPerEpisode’,maxsteps, …
‘StopTrainingCriteria’,"EpisodeCount",…
‘StopTrainingValue’,maxepisodes,…
‘ScoreAveragingWindowLength’,100,…
‘SaveAgentCriteria’,"EpisodeCount",…
‘SaveAgentValue’,maxepisodes,…
‘SaveAgentDirectory’,"savedAgents",…
‘Verbose’,true,…
‘Plots’,’training-progress’)

Cheers,

Patrick Hi,
I have updated my install from R2021a to R2022a. Using the RL toolbox when running the episode manager with the following code in R2022a, when I go to stop the training early, via "Stop Training" in episode manager, the training does not stop, it seems the only way to actual stop the current training early is via the "stop" button on the "run" tab in live editor. Has anyone encounered this in R2022a? and will the agent buffer still be saved via just stopping the script rather than stopping training via EM?
T = 0.066;
maxepisodes = 100;
maxsteps = ceil(T/Ts_agent)
trainingOpts = rlTrainingOptions(…
‘MaxEpisodes’,maxepisodes, …
‘MaxStepsPerEpisode’,maxsteps, …
‘StopTrainingCriteria’,"EpisodeCount",…
‘StopTrainingValue’,maxepisodes,…
‘ScoreAveragingWindowLength’,100,…
‘SaveAgentCriteria’,"EpisodeCount",…
‘SaveAgentValue’,maxepisodes,…
‘SaveAgentDirectory’,"savedAgents",…
‘Verbose’,true,…
‘Plots’,’training-progress’)

Cheers,

Patrick reinforcement learning MATLAB Answers — New Questions

​