Make agent examples compatible with python 3 (#150)

* make cem agen exaple compatible with python 2 and 3

* make the keyboard_agent example compatible with python 2 and 3

Changing `xrange` to `range` should not impact performance unless we're
generating millions of elements (currently only 1000).

* remove algorithm_id from the upload call

examples/agents/cem.py

@@ -1,7 +1,13 @@
+from __future__ import print_function
+
 import gym
 import logging
 import numpy as np
-import json, sys, cPickle, os
+try:
+    import cPickle as pickle
+except ImportError:
+    import pickle
+import json, sys, os
 from os import path
 from _policies import BinaryActionLinearPolicy # Different file so it can be unpickled
 import argparse
@@ -77,10 +83,10 @@ if __name__ == '__main__':
     # Train the agent, and snapshot each stage
     for (i, iterdata) in enumerate(
         cem(noisy_evaluation, np.zeros(env.observation_space.shape[0]+1), **params)):
-        print 'Iteration %2i. Episode mean reward: %7.3f'%(i, iterdata['y_mean'])
+        print('Iteration %2i. Episode mean reward: %7.3f'%(i, iterdata['y_mean']))
         agent = BinaryActionLinearPolicy(iterdata['theta_mean'])
         if args.display: do_rollout(agent, env, 200, render=True)
-        writefile('agent-%.4i.pkl'%i, cPickle.dumps(agent, -1))
+        writefile('agent-%.4i.pkl'%i, str(pickle.dumps(agent, -1)))
 
     # Write out the env at the end so we store the parameters of this
     # environment.
@@ -89,4 +95,4 @@ if __name__ == '__main__':
     env.monitor.close()
 
     logger.info("Successfully ran RandomAgent. Now trying to upload results to the scoreboard. If it breaks, you can always just try re-uploading the same results.")
-    gym.upload(outdir, algorithm_id='cem')
+    gym.upload(outdir)

examples/agents/keyboard_agent.py

@@ -1,11 +1,13 @@
 #!/usr/bin/env python
+from __future__ import print_function
+
 import sys, gym
 
 #
 # Test yourself as a learning agent! Pass environment name as a command-line argument.
 #
 
-env = gym.make('LunarLander-v0' if len(sys.argv)<2 else sys.argv[1])
+env = gym.make('LunarLander-v2' if len(sys.argv)<2 else sys.argv[1])
 
 ACTIONS = env.action_space.n
 ROLLOUT_TIME = 1000
@@ -40,9 +42,9 @@ def rollout(env):
     human_wants_restart = False
     obser = env.reset()
     skip = 0
-    for t in xrange(ROLLOUT_TIME):
+    for t in range(ROLLOUT_TIME):
         if not skip:
-            #print "taking action {}".format(human_agent_action)
+            #print("taking action {}".format(human_agent_action))
             a = human_agent_action
             skip = SKIP_CONTROL
         else:
@@ -57,9 +59,9 @@ def rollout(env):
             import time
             time.sleep(0.1)
 
-print "ACTIONS={}".format(ACTIONS)
+print("ACTIONS={}".format(ACTIONS))
-print "Press keys 1 2 3 ... to take actions 1 2 3 ..."
+print("Press keys 1 2 3 ... to take actions 1 2 3 ...")
-print "No keys pressed is taking action 0"
+print("No keys pressed is taking action 0")
 
 while 1:
     rollout(env)

examples/agents/random_agent.py

@@ -55,4 +55,4 @@ if __name__ == '__main__':
     # Upload to the scoreboard. We could also do this from another
     # process if we wanted.
     logger.info("Successfully ran RandomAgent. Now trying to upload results to the scoreboard. If it breaks, you can always just try re-uploading the same results.")
-    gym.upload(outdir, algorithm_id='random')
+    gym.upload(outdir)

gym/envs/box2d/car_racing.py

@@ -12,7 +12,7 @@ from gym.utils import colorize, seeding
 import pyglet
 from pyglet.gl import *
 
-from car_dynamics import Car
+from .car_dynamics import Car
 
 # Easiest continuous control task to learn from pixels, a top-down racing environment.
 # Discreet control is reasonable in this environment as well, on/off discretisation is