

In [2]:

    
car = NAF()
car.main()









    



[2016-07-19 13:45:16,351] Making new env: MountainCarContinuous-v0






    



True action space: [ 0.], [ 2.]
True state space: [-1.2  -0.07], [ 0.6   0.07]
Filtered action space: [-1.], [ 1.]
Filtered state space: [-2.02205285 -2.88607976], [ 3.3058819   2.88506784]
state dim 2
action dim 1






    



[2016-07-19 13:45:17,944] Action '[[ 1.18974927]]' is not contained within action space 'Box(1,)'.
[2016-07-19 13:45:17,947] Observation '[[[ -4.55458212e-01]]

 [[ -3.20103608e-04]]]' is not contained within observation space 'Box(2,)'.






    



writing output
result after minibatch no. 0 : mean squared error: 0.809489488602
plotting the mu() policy learned by NN






    












    



plotting the mu() policy learned by NN






    












    



plotting the Qfunction
Qfunction for action -1.0






    












    



Qfunction for action -0.5






    












    



Qfunction for action 0.0






    












    



Qfunction for action 0.5






    












    



Qfunction for action 1.0






    












    



writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output
writing output






    



---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
<ipython-input-2-68a7539ee927> in <module>()
      1 car = NAF()
----> 2 car.main()

/home/frederik/Dokumente/DeepRL/NAF/naf_core.pyc in main(self)
    289         tf.gfile.MakeDirs(self.summaries_dir)
    290 
--> 291         self.run()
    292 
    293 

/home/frederik/Dokumente/DeepRL/NAF/naf_core.pyc in run(self, maxstep, max_episode_length)
    218 
    219             # run episode
--> 220             episode_length = self.run_episode(test_run= False, enable_render=False, limit= max_episode_length)
    221             self.train_lengths.append(episode_length)
    222 

/home/frederik/Dokumente/DeepRL/NAF/naf_core.pyc in run_episode(self, enable_render, limit, test_run)
    200                 # if (len(self.replay_memory) > self.warmup) and (self.samples_count % (self.batch_size/2) == 0):
    201                 if len(self.replay_memory) > self.warmup:
--> 202                     self.train_networks()
    203 
    204             state = state_prime

/home/frederik/Dokumente/DeepRL/NAF/naf_core.pyc in train_networks(self)
    263 
    264         dict_ = feed_dict()
--> 265         summary, _, mse_val = self.sess.run([self.log_all, self.train_step, self.loss], feed_dict= dict_)
    266 
    267         if self.step % 10 == 0:

/home/frederik/anaconda2/lib/python2.7/site-packages/tensorflow/python/client/session.pyc in run(self, fetches, feed_dict, options, run_metadata)
    370     try:
    371       result = self._run(None, fetches, feed_dict, options_ptr,
--> 372                          run_metadata_ptr)
    373       if run_metadata:
    374         proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)

/home/frederik/anaconda2/lib/python2.7/site-packages/tensorflow/python/client/session.pyc in _run(self, handle, fetches, feed_dict, options, run_metadata)
    582 
    583     # Validate and process fetches.
--> 584     processed_fetches = self._process_fetches(fetches)
    585     unique_fetches = processed_fetches[0]
    586     target_list = processed_fetches[1]

/home/frederik/anaconda2/lib/python2.7/site-packages/tensorflow/python/client/session.pyc in _process_fetches(self, fetches)
    523         try:
    524           fetch_t = self.graph.as_graph_element(subfetch, allow_tensor=True,
--> 525                                                 allow_operation=True)
    526           fetch_name = compat.as_bytes(fetch_t.name)
    527           if isinstance(fetch_t, ops.Operation):

/home/frederik/anaconda2/lib/python2.7/site-packages/tensorflow/python/framework/ops.pyc in as_graph_element(self, obj, allow_tensor, allow_operation)
   2320     """
   2321     with self._lock:
-> 2322       return self._as_graph_element_locked(obj, allow_tensor, allow_operation)
   2323 
   2324   def _as_graph_element_locked(self, obj, allow_tensor, allow_operation):

/home/frederik/anaconda2/lib/python2.7/site-packages/tensorflow/python/framework/ops.pyc in _as_graph_element_locked(self, obj, allow_tensor, allow_operation)
   2340       raise ValueError("allow_tensor and allow_operation can't both be False.")
   2341 
-> 2342     temp_obj = _as_graph_element(obj)
   2343     if temp_obj is not None:
   2344       obj = temp_obj

/home/frederik/anaconda2/lib/python2.7/site-packages/tensorflow/python/framework/ops.pyc in _as_graph_element(obj)
    121 
    122 
--> 123 def _as_graph_element(obj):
    124   """Convert `obj` to a graph element if possible, otherwise return `None`.
    125 

KeyboardInterrupt: