search

nrontsis / PILCO

Bayesian Reinforcement Learning in Tensorflow
MIT License
313 stars 84 forks source link

local variable 'model' referenced before assignment #18

Closed wonchul-kim closed 5 years ago

wonchul-kim commented 5 years ago

When I run inverted_pendulum.py, this below error came up.

-----Learned models------ ---Lengthscales--- GP0 GP1 GP2 GP3 0 7531.310 5412.474 42419.628 14051.520 1 8967.154 5576.802 42.032 4.307 2 6.223 18.859 143.548 8.425 3 9088.658 42.614 165237.366 178.836 4 28.919 58.716 18.404 15.033 ---Variances--- GP0 GP1 GP2 GP3 0 0.027 0.595 14.387 38.815 ---Noises--- GP0 GP1 GP2 GP3 0 1.000e-06 1.000e-06 1.000e-06 1.000e-06 Controller's optimization: done in 22.0 seconds with reward=34.581. No of ops: 5245

Rollout: 1 Traceback (most recent call last): File "inverted_pendulum.py", line 57, in pilco.optimizemodels() File "/home/wonchul/Desktop/PILCO-master/pilco/models/pilco.py", line 57, in optimizemodels self.mgpr.optimize(restarts=restarts) File "/home/wonchul/Desktop/PILCO-master/pilco/models/mgpr.py", line 51, in optimize best_parameters = model.read_values(session=session) UnboundLocalError: local variable 'model' referenced before assignment

How can I fix it?