Support python3 and pytorch 0.2

agent/__init__.py

@@ -1,5 +1,5 @@
-from async_agent import *
-from DDPG_agent import *
-from DQN_agent import *
-from A2C_agent import *
-from MSDQN_agent import *
+from .async_agent import *
+from .DDPG_agent import *
+from .DQN_agent import *
+from .A2C_agent import *
+from .MSDQN_agent import *

async_worker/__init__.py

@@ -1,6 +1,6 @@
-from actor_critic import *
-from continuous_actor_critic import *
-from n_step_q import *
-from one_step_sarsa import *
-from one_step_q import *
-from ppo import *
+from .actor_critic import *
+from .continuous_actor_critic import *
+from .n_step_q import *
+from .one_step_sarsa import *
+from .one_step_q import *
+from .ppo import *

component/__init__.py

@@ -1,5 +1,5 @@
-from atari_wrapper import *
-from policy import *
-from replay import *
-from task import *
-from random_process import *
+from .atari_wrapper import *
+from .policy import *
+from .replay import *
+from .task import *
+from .random_process import *

component/task.py

@@ -6,7 +6,7 @@
 import gym
 import sys
 import numpy as np
-from atari_wrapper import *
+from .atari_wrapper import *
 
 class BasicTask:
     def __init__(self):

network/__init__.py

@@ -1,3 +1,3 @@
-from conv_network import *
-from shallow_network import *
-from continuous_action_network import *
+from .conv_network import *
+from .shallow_network import *
+from .continuous_action_network import *

network/continuous_action_network.py

@@ -4,7 +4,7 @@
 # declaration at the top                                              #
 #######################################################################
 
-from network import *
+from .base_network import *
 
 class DeterministicActorNet(nn.Module, BasicNet):
     def __init__(self,

network/conv_network.py

@@ -4,7 +4,7 @@
 # declaration at the top                                              #
 #######################################################################
 
-from network import *
+from .base_network import *
 
 # Network for pixel Atari game with value based methods
 class NatureConvNet(nn.Module, VanillaNet):

network/shallow_network.py

@@ -4,7 +4,7 @@
 # declaration at the top                                              #
 #######################################################################
 
-from network import *
+from .base_network import *
 
 # Network for CartPole with value based methods
 class FCNet(nn.Module, VanillaNet):

utils/__init__.py

@@ -1,7 +1,8 @@
-from config import *
-from normalizer import *
-from run import *
+from .config import *
+from .normalizer import *
+from .run import *
+
 try:
-    from tf_logger import Logger
+    from .tf_logger import Logger
 except:
-    from vanilla_logger import Logger
+    from .vanilla_logger import Logger

utils/tf_logger.py

@@ -3,6 +3,7 @@
 import tensorflow as tf
 import numpy as np
 import scipy.misc
+import logging
 
 try:
     from StringIO import StringIO  # Python 2.7
@@ -18,6 +19,7 @@ class Logger(object):
         self.debug = vanilla_logger.debug
         self.warning = vanilla_logger.warning
         self.skip = skip
+        logging.info('')
 
     def scalar_summary(self, tag, value, step):
         if self.skip:

utils/vanilla_logger.py

@@ -1,4 +1,5 @@
 import numpy as np
+import logging
 
 class Logger(object):
     def __init__(self, log_dir, vanilla_logger, skip=False):
@@ -7,6 +8,7 @@ class Logger(object):
         self.debug = vanilla_logger.debug
         self.warning = vanilla_logger.warning
         self.skip = skip
+        logging.info('')
 
     def scalar_summary(self, tag, value, step):
         if self.skip: