flush logger

agent/DDPG_agent.py

@@ -67,8 +67,8 @@ class DDPGAgent:
 
             # tensorboard logging
             suffix = 'test_' if deterministic else ''
-            config.logger.scalar_summary(suffix+'action', action, self.total_steps)
-            config.logger.scalar_summary(suffix+'reward', reward, self.total_steps)
+            config.logger.scalar_summary(suffix + 'action', action, self.total_steps)
+            config.logger.scalar_summary(suffix + 'reward', reward, self.total_steps)
             for key in info:
                 config.logger.scalar_summary('info_' + key, info[key], self.total_steps)
 
@@ -125,4 +125,6 @@ class DDPGAgent:
 
                 self.soft_update(self.target_network, self.worker_network)
 
+        config.logger.writer.file_writer.flush()
+
         return total_reward, steps

async_worker/dpg.py

@@ -135,4 +135,6 @@ class DeterministicPolicyGradient:
         self.shared_reward_normalizer.offline_stats.merge(self.reward_normalizer.online_stats)
         self.reward_normalizer.online_stats.zero()
 
+        config.logger.writer.file_writer.flush()
+
         return steps, total_reward

async_worker/ppo.py

@@ -167,4 +167,6 @@ class ProximalPolicyOptimization:
                 self.critic_opt.step()
                 config.total_epochs.value += 1
 
+        config.logger.writer.file_writer.flush()
+
         return batched_steps, batched_rewards