Training progress is not visible

I am training the network, but the progress is not shown in the plot. However, the variables during the reward function calculation are displayed on the screen as shown in the figure below.

doTraining = true;
if doTraining
    % Train the agent.
    trainingStats = train(agent,env,trainOpts);
else
    % Load the pretrained agent for the example.
    load("PretrainedDDPGAgent.mat","agent")
end

Answer 1

[Shubham]

Hey Sania,

For plotting the training progress, you can use the "Plots" Name-Value arguments while providing the training options to train the network. For example:

trainOpts = rlTrainingOptions(...
    MaxEpisodes=1000,...
    MaxStepsPerEpisode=1000,...
    StopTrainingCriteria="AverageReward",...
    StopTrainingValue=480,...
    Verbose=true,...
    Plots="training-progress")

You can refer to the following documentation for more information on the arguments:

https://www.mathworks.com/help/reinforcement-learning/ref/rl.option.rltrainingoptions.html#d126e86442

I hope this helps!

[Sania Gul]

Tnks Shubham, now graph is plotted in the training monitor window. But still only the Episode Q0 is plotted. No episode or average reward is visible as shown in pic attached. My training options are

dotraining = true;
 
trainOpts = rlTrainingOptions(...
    MaxEpisodes = 150, ...
    MaxStepsPerEpisode = maxStepsPerEpisode, ...
    ScoreAveragingWindowLength = 5,...    
    Verbose = true, ...
    Plots = "training-progress",...
    StopTrainingCriteria = "AverageReward",...
    StopTrainingValue = 2)

Please tell me how I can save the trained model. I shall be very thankful to you.

[Shubham]

For episode and average reward, could you share the relevant files so that I can reproduce the issue?

In order to save the trained agent, you can either use training options or use the "save" function: https://www.mathworks.com/help/reinforcement-learning/ug/train-reinforcement-learning-agents.html#mw_0727b1a5-1162-4517-ae6d-bdb00ef7d5db

[Sania Gul]

Tnk u soooooooooo mch for this coperation. I have increased the value of StopTrainingValue and now all 3 curves(dark, light blue and orange) are visible.

[Shubham]

Glad the issue is resolved!