diff --git a/common/vec_env/subproc_vec_env.py b/common/vec_env/subproc_vec_env.py
index 2952ba0..0259c28 100644
--- a/common/vec_env/subproc_vec_env.py
+++ b/common/vec_env/subproc_vec_env.py
@@ -23,6 +23,7 @@ def worker(remote, map_name, nscripts, i):
   with sc2_env.SC2Env(
       agent_interface_format=[agent_format],
       map_name=map_name,
+      players=[sc2_env.Agent(sc2_env.Race.terran)],
       step_mul=2) as env:
     available_actions = []
     result = None
diff --git a/train_defeat_zerglings.py b/train_defeat_zerglings.py
index 0fba7e3..f144540 100644
--- a/train_defeat_zerglings.py
+++ b/train_defeat_zerglings.py
@@ -75,6 +75,11 @@ def main():
       map_name="DefeatZerglingsAndBanelings",
       step_mul=step_mul,
       visualize=True,
+      players=[sc2_env.Agent(sc2_env.Race.terran)],
+      agent_interface_format=sc2_env.AgentInterfaceFormat(
+        feature_dimensions=sc2_env.Dimensions(
+        screen=64,
+        minimap=64)),
       game_steps_per_episode=steps * step_mul) as env:
 
     model = deepq.models.cnn_to_mlp(