protected void RunStep()
{
trainer.Step(environment);
rewardOneEpi += environment.LastReward(0);
bool reset = trainer.Record(environment);
//training
if (trainer.Steps >= stepsBeforeTrain && trainer.Steps % trainingStepInterval == 0)
{
trainer.TrainRandomBatch(batchSize);
//log the loss
loss.AddValue(trainer.LastLoss);
if (loss.JustUpdated)
{
print("Loss:" + loss.Average);
}
}
//reset if end
if (environment.IsEnd() || (reset && training))
{
currentEpisode++;
rewardEpiAve.AddValue(rewardOneEpi);
rewardOneEpi = 0;
environment.Reset();
}
}
protected void RunStep()
{
trainer.Step(environment);
rewardLeftOneEpi += environment.LastReward(0);
bool reset = trainer.Record(environment);
//training
if (trainer.Steps >= stepsBeforeTrain && trainer.Steps % trainingStepInterval == 0)
{
trainer.TrainRandomBatch(batchSize);
//log the loss
loss.AddValue(trainer.LastLoss);
if (loss.JustUpdated)
{
print("Loss:" + loss.Average);
}
}
//reset if end
if (environment.IsEnd() || (reset && training))
{
currentEpisode++;
if (environment.GameWinPlayer == 0)
{
leftWin++;
winningRate50Left.AddValue(1);
}
else
{
rightWin++;
winningRate50Left.AddValue(0);
}
reward50EpiLeft.AddValue(rewardLeftOneEpi);
rewardLeftOneEpi = 0;
environment.Reset();
}
}