def main():
gamegrid = GameGrid()
gamegrid.hide()
agent = DNNAgent(None,
waitTime=0,
trainDataPickle="ULRD_train_2000_20.pickle")
gamegrid.setAgent(agent)
with open(('ULRD_trained_model_20_game_layers_64.pickle'), 'wb') as f:
pickle.dump(agent, f)
print("Train data stored in {}".format(f))
class Runner:
def __init__(self, numEpochs, numItterations, agentCode):
self._numEpochs = numEpochs
self._numItterations = numItterations
self._agentCode = agentCode
self._gamegrid = GameGrid()
self._gamegrid.hide()
self._trainingRecord = []
def createAgent(self,
gameSessionFile=None,
trainName=None,
trainData=None,
trainDataPickle=None,
existingAgent=None):
if (self._agentCode == 0):
self._gamegrid.setAgent(RandomAgent(None, waitTime=0))
elif (self._agentCode == 1):
self._gamegrid.setAgent(PatternAgentULRD(None, waitTime=0))
elif (self._agentCode == 2):
self._gamegrid.setAgent(PatternAgentLURD(None, waitTime=0))
elif (self._agentCode == 3):
self._gamegrid.setAgent(ManualAgent(None, waitTime=0))
elif (self._agentCode == 4):
self._gamegrid.setAgent(
DNNAgent(None,
waitTime=0,
gameSessionFile=gameSessionFile,
trainName=trainName,
trainData=trainData,
trainDataPickle=trainDataPickle))
self._gamegrid.getAgent().setGameGrid(self._gamegrid)
def refreshGameGrid(self):
agent = self._gamegrid.getAgent()
self._gamegrid = GameGrid()
self._gamegrid.hide()
agent.reset()
self._gamegrid.setAgent(agent)
self._gamegrid.getAgent().setGameGrid(self._gamegrid)
def runTraining(self):
for epochNum in range(0, self._numEpochs):
for itterNum in range(0, self._numItterations):
self.refreshGameGrid()
self._gamegrid.setAgent(self._agent)
print("Epoch: ", epochNum, " Iteration: ", itterNum)
self._gamegrid.mainloop()
# print(gamegrid.matrix)
print("Score: ", self._gamegrid.scoreMatrix())
self._agent.setScore(self._gamegrid.scoreMatrix())
# The current code running AI games needs to know the current epochNum for encoding filename
(boards, moves, score) = self._agent.getGameRecord()
# self._agent.pikPakGame()
self._trainingRecord.append(
(epochNum, itterNum, boards, moves, score))
return self._trainingRecord
def main():
existingAgent1 = None
with open("TrainingPartialCountRunner_100_20_4_5.pickle", 'rb') as f:
existingAgent0 = pickle.load(f)
with open("ULRD_trained_model_20_game_layers_32_16.pickle", 'rb') as f:
existingAgent1 = pickle.load(f)
with open("ULRD_trained_model_20_game_layers_64_16.pickle", 'rb') as f:
existingAgent2 = pickle.load(f)
with open("ULRD_trained_model_20_game_layers_64_16_8.pickle", 'rb') as f:
existingAgent3 = pickle.load(f)
with open("ULRD_trained_model_20_game_layers_64_32_8.pickle", 'rb') as f:
existingAgent4 = pickle.load(f)
with open("ULRD_trained_model_20_game_layers_64.pickle", 'rb') as f:
existingAgent5 = pickle.load(f)
agentDict = {
1: RandomAgent(None, waitTime=0),
2: PatternAgentULRD(None, waitTime=0),
0: existingAgent0,
3: DNNAgent(None, waitTime=0, trainName="ULRD_train.pickle"),
4: existingAgent1,
5: existingAgent2,
6: existingAgent2,
7: existingAgent2,
8: existingAgent2
}
agentDescription = {
1: "Random",
2: "Up-Left-Right-Down",
0: "Online learning NN",
3: "DNN Agent",
4: "DNN Agent with layers [32, 16]",
5: "DNN Agent with layers [64, 16]",
6: "DNN Agent with layers [64, 16, 8]",
7: "DNN Agent with layers [64, 32, 8]",
8: "DNN Agent with layers [64]"
}
agentScoreDict = {
1: [],
2: [],
0: [],
3: [],
4: [],
5: [],
6: [],
7: [],
8: []
}
agentColors = {
1: "b",
2: "r",
0: "#1f004d",
3: "g",
4: "c",
5: "m",
6: "y",
7: "k",
8: "#3CFE6E"
}
gameIDs = []
for i in range(0, 15):
gameIDs.append(i)
random.seed(i)
for (agentKey, agent) in agentDict.items():
gamegrid = GameGrid()
gamegrid.hide()
gamegrid.setAgent(agent)
agent.setGameGrid(gamegrid)
gamegrid.mainloop()
agentScoreDict[agentKey].append(sumScoreMatrix(gamegrid.matrix))
print(agentScoreDict[agentKey])
agent.reset()
plotTrainingRecord(gameIDs, agentDict, agentDescription, agentScoreDict,
agentColors)