Merge pull request #9 from islamelnabarawy/fix-datawriter-condition

Fix error in the way learners check for the presence of a summary writer
yukezhu · Oct 31, 2017 · 173a04f · 173a04f
2 parents f3f2321 + b5bf6ae
commit 173a04f
Show file tree

Hide file tree

Showing 4 changed files with 4 additions and 4 deletions.
diff --git a/rl/neural_q_learner.py b/rl/neural_q_learner.py
@@ -193,7 +193,7 @@ def updateModel(self):
         next_state_mask[k] = 1
 
     # whether to calculate summaries
-    calculate_summaries = self.train_iteration % self.summary_every == 0 and self.summary_writer is not None
+    calculate_summaries = self.summary_writer is not None and self.train_iteration % self.summary_every == 0
 
     # perform one update of training
     cost, _, summary_str = self.session.run([

diff --git a/rl/pg_actor_critic.py b/rl/pg_actor_critic.py
@@ -190,7 +190,7 @@ def updateModel(self):
       discounted_rewards[t] = r
 
     # whether to calculate summaries
-    calculate_summaries = self.train_iteration % self.summary_every == 0 and self.summary_writer is not None
+    calculate_summaries = self.summary_writer is not None and self.train_iteration % self.summary_every == 0
 
     # update policy network with the rollout in batches
     for t in range(N-1):

diff --git a/rl/pg_ddpg.py b/rl/pg_ddpg.py
@@ -205,7 +205,7 @@ def updateModel(self):
         next_state_mask[k] = 1
 
     # whether to calculate summaries
-    calculate_summaries = self.train_iteration % self.summary_every == 0 and self.summary_writer is not None
+    calculate_summaries = self.summary_writer is not None and self.train_iteration % self.summary_every == 0
 
     # compute a = u(s)
     policy_outs = self.session.run(self.policy_outputs, {

diff --git a/rl/pg_reinforce.py b/rl/pg_reinforce.py
@@ -174,7 +174,7 @@ def updateModel(self):
     discounted_rewards /= np.std(self.all_rewards)
 
     # whether to calculate summaries
-    calculate_summaries = self.train_iteration % self.summary_every == 0 and self.summary_writer is not None
+    calculate_summaries = self.summary_writer is not None and self.train_iteration % self.summary_every == 0
 
     # update policy network with the rollout in batches
     for t in range(N-1):