Addresses oxwhirl#72

douglasrizzo · web-flow · commit 8042fdcfe037 · 2021-07-09T13:01:53.000-03:00
diff --git a/smac/env/starcraft2/starcraft2.py b/smac/env/starcraft2/starcraft2.py
@@ -742,7 +742,7 @@ def reward_battle(self):
                     delta_enemy += prev_health - e_unit.health - e_unit.shield
 
         if self.reward_only_positive:
-            reward = abs(delta_enemy + delta_deaths)  # shield regeneration
+            reward = max(0, delta_enemy + delta_deaths)  # shield regeneration
         else:
             reward = delta_enemy + delta_deaths - delta_ally
 
