cleaning

jimkon · Jan 19, 2018 · 02f8e5c · 02f8e5c
1 parent edd59c7
commit 02f8e5c
Show file tree

Hide file tree

Showing 3 changed files with 20 additions and 17 deletions.
diff --git a/src/main.py b/src/main.py
@@ -12,7 +12,7 @@
 AUTO_SAVE_AFTER_EPISODES = 500
 
 
-def run(episodes=2500,
+def run(episodes=250,
         collecting_data=False,
         experiment='InvertedPendulum-v1',
         max_actions=1e3,
@@ -114,6 +114,10 @@ def run(episodes=2500,
 
     data_fetcher.save()
 
+    # printing average times for running and training steps
+    data_fetcher.print_times(other_keys=data_fetcher.get_keys('run'))
+    data_fetcher.print_times(other_keys=data_fetcher.get_keys('agent_'), total_time_field='count')
+
 
 if __name__ == '__main__':
     run()
diff --git a/src/show_results.py b/src/show_results.py
@@ -13,20 +13,18 @@ def show():
     experiment = 'InvertedPendulum-v1'
     v = 3
 
-    name = '{}data_{}_Wolp{}_{}k{}_{}'.format(folder,
-                                              episodes,
-                                              v,
-                                              actions,
-                                              k,
-                                              experiment
-                                              )
+    name = '{}data_{}_Wolp{}_{}k{}_{}_shrinked'.format(folder,
+                                                       episodes,
+                                                       v,
+                                                       actions,
+                                                       k,
+                                                       experiment
+                                                       )
 
     fd = Agent_data(name)
 
     fd.load()
-    # printing average times for running and training steps
-    # fd.print_times(other_keys=fd.get_keys('run'))
-    # fd.print_times(other_keys=fd.get_keys('agent_'), total_time_field='count')
+
     plot_rewards(fd)
 
     plot_average_reward(fd)

diff --git a/src/util/agent_data.py b/src/util/agent_data.py
@@ -283,12 +283,13 @@ def get_full_episode_data(self, ep):
     def get_number_of_episodes(self):
         return len(self.get_data('rewards'))
 
-    def get_adaption_episode(self, reward_threshold=50):
-        eps = self.get_episodes_with_reward_greater_than(reward_threshold)
-        if len(eps) > 0:
-            return eps[0]
-        else:
-            return -1
+    def get_adaption_episode(self, reward_threshold=10):
+        rewards = self.get_data('rewards')
+        total = 0
+        for i in range(len(rewards)):
+            total += rewards[i]
+            if total / (i + 1) > reward_threshold:
+                return i
 
     def get_adaption_time(self, reward_threshold=50):
         first_increase = self.get_adaption_episode(reward_threshold)