feat: release 06-09-2021

generational/generation.py

@@ -174,7 +174,7 @@ class GenerationalEnvironment:
                     self.infos_list = checkpoint["infos_list"]
         pass
 
-    def visualize_graphs(self):
+    def visualize_graphs(self, version:int, files: str = 'files'):
         turns_list_length = []
         i = 0
         for item in self.turns_list:
@@ -185,20 +185,72 @@ class GenerationalEnvironment:
         plt.title("turns_taken_per_epoch")
         plt.show()
 
-        for agent_id in self.game_networks:
-            agent_rewards_epoch = []
-            for item in self.epoch_rewards_list:
-                agent_rewards_epoch.append(item[agent_id])
-            agent_losses_epoch = []
-            for item in self.epoch_loss_list:
-                agent_losses_epoch.append(item[agent_id])
-
-            plt.plot(agent_rewards_epoch)
-            plt.title(agent_id + "_epoch_rewards")
-            plt.show()
-            plt.plot(agent_losses_epoch)
-            plt.title(agent_id + "_epoch_loss")
-            plt.show()
+        agents_medium_rewards_list = []
+        agents_medium_loss_list = []
+
+
+        agent_rewards_epoch = {agent: [] for agent in self.game_networks }
+        for item in self.epoch_rewards_list:
+            for agent_id in self.game_networks:
+                agent_rewards_epoch[agent_id].append(item[agent_id])
+            median = 0
+            for value in item.values():
+                median += value
+            median = median / self.game_networks.__len__()
+            agents_medium_rewards_list.append(median)
+
+
+        agent_losses_epoch = {agent: [] for agent in self.game_networks }
+        for item in self.epoch_loss_list:
+            for agent_id in self.game_networks:
+                agent_losses_epoch[agent_id].append(item[agent_id])
+            median = 0
+            for value in item.values():
+                median += value
+            median = median / self.game_networks.__len__()
+            agents_medium_loss_list.append(median)
+
+            # plt.title(agent_id + "_epoch_rewards")
+            # plt.show()
+            # plt.plot(agent_losses_epoch)
+            # plt.title(agent_id + "_epoch_loss")
+            # plt.show()
+
+        fig, axs = plt.subplots(3, 2, sharey=True)
+
+        axs[0, 0].plot(agent_rewards_epoch['agent_1'], 'g-')
+        axs[0, 0].set_title('Agent 1')
+        axs[0, 1].plot(agent_rewards_epoch['agent_2'], 'c-')
+        axs[0, 1].set_title('Agent 2')
+        axs[1, 0].plot(agent_rewards_epoch['agent_3'], 'm-')
+        axs[1, 0].set_title('Agent 3')
+        axs[1, 1].plot(agent_rewards_epoch['agent_4'], 'y-')
+        axs[1, 1].set_title('Agent 4')
+        axm = plt.subplot2grid((3, 3), (2, 0), colspan=3)
+        axm.plot(agents_medium_rewards_list, 'r-')
+        axm.set_title('Median')
+        plt.title("epoch_rewards")
+        plt.gcf().set_size_inches((15, 15))
+        plt.savefig(fname="{file}/{version}/rewards.pdf".format(file=files, version=str(version)))
+        # plt.show()
+        plt.close()
+
+        fig, axs = plt.subplots(3, 2, sharey=True)
+        axs[0, 0].plot(agent_losses_epoch['agent_1'], 'g-')
+        axs[0, 0].set_title('Agent 1')
+        axs[0, 1].plot(agent_losses_epoch['agent_2'], 'c-')
+        axs[0, 1].set_title('Agent 2')
+        axs[1, 0].plot(agent_losses_epoch['agent_3'], 'm-')
+        axs[1, 0].set_title('Agent 3')
+        axs[1, 1].plot(agent_losses_epoch['agent_4'], 'y-')
+        axs[1, 1].set_title('Agent 4')
+        axm = plt.subplot2grid((3, 3), (2, 0), colspan=3)
+        axm.plot(agents_medium_loss_list, 'r-')
+        axm.set_title('Median')
+        plt.title("epoch_loss")
+        plt.gcf().set_size_inches((15, 15))
+        plt.savefig(fname="{file}/{version}/loss.pdf".format(file=files, version=str(version)))
+        # plt.show()
 
             # agent_losses = []
             # for item in self.loss_list:

main.py

@@ -72,21 +72,20 @@ class sticky_actions_vcustom(StepAltWrapper):
         super().step(action)
         
 
-def show_progress(version: int):
-    file = "files"
+def show_progress(version: int, file: str = "files"):
     tron = GenerationalTronEnvironment(env=env_creator(), device=device, render=True, training=False)
-    tron.load_agents("{}/{version}/".format(file, version=str(version)))
+    tron.load_agents("{file}/{version}/".format(file=file, version=str(version)))
 
 
-    tron.visualize_graphs()
-    rewards_list, loss_list, turns_list, infos_list = tron.play_loop(1)
-    infos = infos_list[-1]
-    for agent, info in infos.items():
-        if agent == "__iteration__":
-            continue
-        plt.imshow(info, cmap='hot', interpolation='nearest')
-        plt.savefig(fname="{}/{version}/{agent}_iteration_{iteration}.png".format(file, version=str(version), agent=agent, iteration=str(infos["__iteration__"])))
-    pass
+    tron.visualize_graphs(version=version)
+    # rewards_list, loss_list, turns_list, infos_list = tron.play_loop(1)
+    # infos = infos_list[-1]
+    # for agent, info in infos.items():
+    #     if agent == "__iteration__":
+    #         continue
+    #     plt.imshow(info, cmap='hot', interpolation='nearest')
+    #     plt.savefig(fname="{}/{version}/{agent}_iteration_{iteration}.png".format(file, version=str(version), agent=agent, iteration=str(infos["__iteration__"])))
+    # pass
 
 
 def create_params(competitive: bool = True, cooperative: bool = False, death_punish: bool = True):
@@ -528,11 +527,11 @@ if __name__ == '__main__':
     # parallel_test_env(num_cycles=2500000, env=atari_pong_no_config(False))
     # parallel_test_env(env=atari_pong({}), num_cycles=100000)
 
-    # show_progress(47)
+    show_progress(47)
     # train()
     # train2p()
     # train_tron_v2(cooperative=False, competitive=True, death_punish=True)
-    vis_graphic_tron("files", 48, 2)
+    # vis_graphic_tron("files", 48, 3)
 
     # AutoROM.main(True, "D:/Programmieren/Python/Rom/")