def run(self):
Generator = Generate()
importNetwork = load_model(self.networkName)
rl = RLsys(4, importNetwork.input_shape[2])
rl.qnet.network = importNetwork
rl.changeEpsilon(0)
testProbs = [0.1]
draw = Drawer()
for prob in testProbs:
numberOfLogicalErrorsRL = 0
numberOfLogicalErrorsMWPM = 0
Generator.DataGeneration(prob)
humRep = np.load('ToricCodeHumanTest.npy')
comRep = np.load('ToricCodeComputerTest.npy')
size = comRep.shape[0]
for i in range(comRep.shape[2]):
state = comRep[:, :, i]
human = humRep[:, :, i]
env = Env(state, human, checkGroundState=True)
'''
blossomReward = 0
if np.count_nonzero(state) > 0:
state_ = np.copy(state)
state_ = self.labelState(state_,size)
BlossomObject = Blossom(state_)
MWPM = BlossomObject.readResult()
Benv = BEnv(state_, human, checkGroundState = True)
for element in MWPM:
error1 = element[0]+1
error2 = element[1]+1
blossomReward, _ = Benv.blossomCancel(error1, error2)
if blossomReward != Benv.correctGsR:
numberOfLogicalErrorsMWPM += 1
'''
r = env.correctGsR
numIter = 0
if self.graphix:
print(f"state before correction {i}: ")
draw.Draw(state, human)
while len(env.getErrors()) > 0:
numIter = numIter + 1
observation = env.getObservation()
a, e = rl.choose_action(observation)
r = env.moveError(a, e)
if numIter > 50:
r = -1
break
if self.graphix:
print(f"state after correction {i}: ")
draw.Draw(env.state, env.humanState)
if r != env.correctGsR:
numberOfLogicalErrorsRL += 1
if ((i + 1) % 100 == 0):
print("Iteration: ", str(i))
print("Logical error probability RL agent: ",
numberOfLogicalErrorsRL / (i + 1))
