reviewed replay_memory code

Former-commit-id: 95a432d

texar/agents/dqn_agent.py

@@ -104,7 +104,7 @@ def set_initial_state(self, observation):
         self.current_state = np.array(observation)
 
     def perceive(self, action, reward, is_terminal, next_observation):
-        self.replay_memory.push({
+        self.replay_memory.add({
             'state': self.current_state,
             'action': action,
             'reward': reward,
@@ -121,7 +121,7 @@ def train_qnet(self):
         """ Train the Q-Network
         :return:
         """
-        minibatch = self.replay_memory.sample(self.batch_size)
+        minibatch = self.replay_memory.get(self.batch_size)
         state_batch = np.array([data['state'] for data in minibatch])
         action_batch = np.array([data['action'] for data in minibatch])
         reward_batch = np.array([data['reward'] for data in minibatch])

texar/core/explorations.py

@@ -77,7 +77,6 @@ def get_epsilon(self, timestep):
             return self._hparams.initial_epsilon
         if timestep > et:
             return self._hparams.final_epsilon
-
         r = (timestep - st) * 1.0 / nsteps
         epsilon = (1 - r) * self._hparams.initial_epsilon + \
                 r * self._hparams.final_epsilon

texar/core/replay_memories.py

@@ -1,27 +1,46 @@
-from texar.hyperparams import HParams
+#
+"""
+TODO: docs
+"""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
 
 from collections import deque
 import random
 
+from texar.hyperparams import HParams
+
 
 class ReplayMemoryBase(object):
+    """TODO: docs
+    """
     def __init__(self, hparams=None):
         self._hparams = HParams(hparams, self.default_hparams())
 
-    def push(self, element):
+    def add(self, element):
+        """TODO: docs
+        """
         raise NotImplementedError
 
-    def sample(self, size):
+    def get(self, size):
+        """TODO: docs
+        """
         raise NotImplementedError
 
     @staticmethod
     def default_hparams():
+        """Returns a dictionary of default hyperparameters.
+        """
         return {
             'name': 'replay_memory'
         }
 
 
 class DequeReplayMemory(ReplayMemoryBase):
+    """TODO: docs
+    """
     def __init__(self, hparams=None):
         ReplayMemoryBase.__init__(self, hparams)
         self.deque = deque()
@@ -34,10 +53,11 @@ def default_hparams():
             'capacity': 80000
         }
 
-    def push(self, element):
+    def add(self, element):
         self.deque.append(element)
         if len(self.deque) > self.capacity:
             self.deque.popleft()
 
-    def sample(self, size):
+    #TODO(zhiting): is it okay to have stand alone random generator ?
+    def get(self, size):
         return random.sample(self.deque, size)