priorityExponent=0.0,
batchSize=64,
maxEpisodes=4096,
trainSteps=1024,
minStepsBeforeTraining=4096,
rewardScaling=(10.0 ** -0.75),
actionShift=0.0,
stepsPerUpdate=1,
render=True,
showGraphs=False,
saveModel=False,
saveModelToS3=False,
restoreModel=True,
train=False,
testSteps=1024,
maxMinutes=360,
targetEntropy=-4.0,
maxGradientNorm=5.0,
meanRegularizationConstant=0.0,
varianceRegularizationConstant=0.0,
randomStartSteps=0,
gradientSteps=1,
initialExtraNoise=0,
extraNoiseDecay=0,
evaluationEvery=25,
numFinalEvaluations=10,
maxTrainSteps = 1000000
)
print("Total Reward: "+str(agent.execute()))
batchSize=64,
nStep=3,
frameSkip=2,
maxEpisodes=4096,
trainSteps=1024,
maxTrainSteps=6000000,
minStepsBeforeTraining=10000,
rewardScaling=(10.0**-0.75),
actionShift=0.0,
stepsPerUpdate=1,
render=False,
showGraphs=True,
saveModel=True,
saveModelToS3=False,
restoreModel=False,
train=True,
testSteps=1024,
maxMinutes=360,
targetEntropy=-4.0,
maxGradientNorm=5.0,
meanRegularizationConstant=0.0,
varianceRegularizationConstant=0.0,
randomStartSteps=10000,
gradientSteps=1,
initialExtraNoise=0,
extraNoiseDecay=0,
evaluationEvery=25,
numFinalEvaluations=10)
results = agent.execute()
qNetworkSize=[256, 256],
valueNetworkSize=[256, 256],
entropyCoefficient=entropyCoefficient,
valueNetworkLearningRate=learningRate,
policyNetworkLearningRate=learningRate,
qNetworkLearningRate=learningRate,
tau=0.005,
gamma=0.99,
maxMemoryLength=int(1e6),
priorityExponent=0,
batchSize=256,
maxGradientNorm=5,
maxEpisodes=1024,
trainSteps=1024,
minStepsBeforeTraining=4096,
rewardScaling=rewardScaling,
actionScaling=actionScaling,
actionShift=0.0,
stepsPerUpdate=1,
render=True,
showGraphs=True,
meanRegularizationConstant=weightRegularizationConstant,
varianceRegularizationConstant=weightRegularizationConstant,
testSteps=1024,
maxMinutes=600)
result = agent.execute()
print("Result: " + result)
except:
print("Error evaluating parameters")
result = -20000