Add scripted baseline agents

agents/scripted_baseline_agent.py

+from utils.base_agent import NeuralMMOAgent
+from forge.trinity.scripted import baselines
+import projekt
+
+class BaselineForageAgent(NeuralMMOAgent):
+    def __init__(self):
+        self.agent = getattr(baselines, 'Forage')(projekt.config.SmallMaps())
+
+    def register_reset(self, observations):
+        action = self.agent(observations)
+        return action
+
+    def compute_action(self, observations, info=None):
+        action = self.agent(observations)
+        return action
+
+class BaselineCombatAgent(NeuralMMOAgent):
+    def __init__(self):
+        self.agent = getattr(baselines, 'Combat')(projekt.config.SmallMaps())
+
+    def register_reset(self, observations):
+        action = self.agent(observations)
+        return action
+
+    def compute_action(self, observations, info=None):
+        action = self.agent(observations)
+        return action
+
+class BaselineRandomAgent(NeuralMMOAgent):
+    def __init__(self):
+        self.agent = getattr(baselines, 'Random')(projekt.config.SmallMaps())
+
+    def register_reset(self, observations):
+        action = self.agent(observations)
+        return action
+
+    def compute_action(self, observations, info=None):
+        action = self.agent(observations)
+        return action

env/gym-neuralmmo/gym_neuralmmo/envs/neuralmmo_eval.py

@@ -40,7 +40,7 @@ class NeuralMMOEval(gym.Env):
         self.available_agents = self.eval_agents[:]
         self.assign_agents()
         self.actions = self.get_agent_actions()
-        return self.observations
+        return self.observations[self.player_idx]
 
     def get_available_agent(self):
         try:
@@ -64,7 +64,7 @@ class NeuralMMOEval(gym.Env):
         return actions
 
     def step(self,action):
-        self.observations, dones, rewards, _ = self.env.step(self.actions)
+        self.observations, dones, rewards, _ = self.env.step(self.actions, preprocessActions = False)
         for agent in dones:
           if dones[agent] == -1:
             self.dead_agents.append(agent)

env/gym-neuralmmo/setup.py

 from setuptools import setup
 
-setup(name="gym_neuralmmo", version="0.0.1", install_requires=["gym"])
+setup(name="gym_neuralmmo", version="0.0.2", install_requires=["gym"])

players.yaml

@@ -4,21 +4,21 @@
 # Max number of opponent agents is 127
 
 player_agent:
-  file: random_agent
-  agent_class: RandomNeuralMMOAgent
+  file: scripted_baseline_agent
+  agent_class: BaselineForageAgent
 
 opponent_agents:
   agent_1:
-    file: random_agent
-    agent_class: RandomNeuralMMOAgent
+    file: scripted_baseline_agent
+    agent_class: BaselineForageAgent
     num_agents: 50
 
   agent_2:
-    file: random_agent
-    agent_class: RandomNeuralMMOAgent
+    file: scripted_baseline_agent
+    agent_class: BaselineForageAgent
     num_agents: 51
 
   agent_3:
-    file: random_agent
-    agent_class: RandomNeuralMMOAgent
-    num_agents: 101
+    file: scripted_baseline_agent
+    agent_class: BaselineForageAgent
+    num_agents: 26

rollout.py

@@ -20,7 +20,7 @@ def main():
         done = False
         while done == False:
             obs, dones, rewards, _ = env.step(action)
-            action = player_agent.compute_action(action)
+            action = player_agent.compute_action(obs["player"])
             total_rewards += rewards["player"]
             done = dones["player"]