Merge pull request #45 from CarperAI/daveey-git-pr-13444-7615

Remove population from realm

nmmo/core/agent.py

@@ -1,13 +1,7 @@
 
-from nmmo.lib import colors
-
-
 class Agent:
   policy   = 'Neural'
 
-  color    = colors.Neon.CYAN
-  pop      = 0
-
   def __init__(self, config, idx):
     '''Base class for agents
 
@@ -17,7 +11,6 @@ def __init__(self, config, idx):
     '''
     self.config = config
     self.iden   = idx
-    self.pop    = Agent.pop
 
   def __call__(self, obs):
     '''Used by scripted agents to compute actions. Override in subclasses.

nmmo/core/config.py

@@ -388,9 +388,6 @@ class Combat:
   COMBAT_SYSTEM_ENABLED              = True
   '''Game system flag'''
 
-  COMBAT_FRIENDLY_FIRE               = True
-  '''Whether agents with the same population index can hit each other'''
-
   COMBAT_SPAWN_IMMUNITY              = 20
   '''Agents older than this many ticks cannot attack agents younger than this many ticks'''
 
@@ -518,8 +515,8 @@ class Item:
   ITEM_INVENTORY_CAPACITY             = 12
   '''Number of inventory spaces'''
 
-  ITEM_GIVE_TO_FRIENDLY               = True
-  '''Whether agents with the same population index can give gold/item to each other'''
+  ITEM_ALLOW_GIFT               = True
+  '''Whether agents can give gold/item to each other'''
 
   @property
   def INVENTORY_N_OBS(self):

nmmo/core/env.py

@@ -18,11 +18,7 @@
 
 
 class Env(ParallelEnv):
-  '''Environment wrapper for Neural MMO using the Parallel PettingZoo API
-
-  Neural MMO provides complex environments featuring structured observations/actions,
-  variably sized agent populations, and long time horizons. Usage in conjunction
-  with RLlib as demonstrated in the /projekt wrapper is highly recommended.'''
+  # Environment wrapper for Neural MMO using the Parallel PettingZoo API
 
   def __init__(self,
     config: Default = nmmo.config.Default(), seed=None):
@@ -390,14 +386,9 @@ def _compute_rewards(self, agents: List[AgentID], dones: Dict[AgentID, bool]):
       agent = self.realm.players.get(agent_id)
       assert agent is not None, f'Agent {agent_id} not found'
 
-      infos[agent_id] =  {'population': agent.population}
-
-      if agent.diary is None:
-        rewards[agent_id] = 0
-        continue
-
-      rewards[agent_id] = sum(agent.diary.rewards.values())
-      infos[agent_id].update(agent.diary.rewards)
+      if agent.diary is not None:
+        rewards[agent_id] = sum(agent.diary.rewards.values())
+        infos[agent_id].update(agent.diary.rewards)
 
     return rewards, infos
 

nmmo/core/log_helper.py

@@ -146,7 +146,4 @@ def _player_stats(self, player: Agent) -> Dict[str, float]:
       for achievement in player.diary.achievements:
         stats["Achievement_{achievement.name}"] = float(achievement.completed)
 
-    # Used for SR
-    stats['PolicyID'] = player.population
-
     return stats

nmmo/core/observation.py

@@ -209,15 +209,11 @@ def _make_attack_mask(self):
     else:
       spawn_immunity = np.ones(self.entities.len, dtype=np.int8)
 
-    if not self.config.COMBAT_FRIENDLY_FIRE:
-      population = self.entities.values[:,EntityState.State.attr_name_to_col["population_id"]]
-      no_friendly_fire = population != agent.population_id # this automatically masks self
-    else:
-      # allow friendly fire but no self shooting
-      no_friendly_fire = np.ones(self.entities.len, dtype=np.int8)
-      no_friendly_fire[self.entities.index(agent.id)] = 0 # mask self
+    # allow friendly fire but no self shooting
+    not_me = np.ones(self.entities.len, dtype=np.int8)
+    not_me[self.entities.index(agent.id)] = 0 # mask self
 
-    return np.concatenate([within_range & no_friendly_fire & spawn_immunity,
+    return np.concatenate([within_range & not_me & spawn_immunity,
       np.zeros(self.config.PLAYER_N_OBS - self.entities.len, dtype=np.int8)])
 
   def _make_use_mask(self):
@@ -287,10 +283,11 @@ def _make_give_target_mask(self):
     entities_pos = self.entities.values[:, [EntityState.State.attr_name_to_col["row"],
                                             EntityState.State.attr_name_to_col["col"]]]
     same_tile = utils.linf(entities_pos, (agent.row, agent.col)) == 0
-    same_team_not_me = (self.entities.ids != agent.id) & (agent.population_id == \
-                self.entities.values[:, EntityState.State.attr_name_to_col["population_id"]])
+    not_me = self.entities.ids != self.agent_id
+    player = (self.entities.values[:,EntityState.State.attr_name_to_col["npc_type"]] == 0)
 
-    return np.concatenate([same_tile & same_team_not_me,
+    return np.concatenate([
+      (same_tile & player & not_me),
       np.zeros(self.config.PLAYER_N_OBS - self.entities.len, dtype=np.int8)])
 
   def _make_give_gold_mask(self):

nmmo/core/realm.py

@@ -119,7 +119,7 @@ def packet(self):
   }
 
   @property
-  def population(self):
+  def num_players(self):
     """Number of player agents"""
     return len(self.players.entities)
 

nmmo/entity/entity.py

@@ -14,7 +14,7 @@
 EntityState = SerializedState.subclass(
   "Entity", [
     "id",
-    "population_id",
+    "npc_type", # 1 - passive, 2 - neutral, 3 - aggressive
     "row",
     "col",
 
@@ -48,7 +48,7 @@
 EntityState.Limits = lambda config: {
   **{
     "id": (-math.inf, math.inf),
-    "population_id": (-3, config.PLAYER_POLICIES-1),
+    "npc_type": (0, 4),
     "row": (0, config.MAP_SIZE-1),
     "col": (0, config.MAP_SIZE-1),
     "damage": (0, math.inf),
@@ -211,7 +211,7 @@ def packet(self):
 
 # pylint: disable=no-member
 class Entity(EntityState):
-  def __init__(self, realm, pos, entity_id, name, color, population_id):
+  def __init__(self, realm, pos, entity_id, name):
     super().__init__(realm.datastore, EntityState.Limits(realm.config))
 
     self.realm = realm
@@ -222,11 +222,9 @@ def __init__(self, realm, pos, entity_id, name, color, population_id):
     self.repr = None
 
     self.name = name + str(entity_id)
-    self.color = color
 
     self.row.update(pos[0])
     self.col.update(pos[1])
-    self.population_id.update(population_id)
     self.id.update(entity_id)
 
     self.vision = self.config.PLAYER_VISION_RADIUS
@@ -259,10 +257,6 @@ def packet(self):
       'name': self.name,
       'level': self.attack_level,
       'item_level': self.item_level.val,
-      'color': self.color.packet(),
-      'population': self.population,
-      # FIXME: Don't know what it does. Previous nmmo entities all returned 1
-      # 'self': self.self.val,
     }
 
     return data
@@ -338,7 +332,3 @@ def attack_level(self) -> int:
     mage = self.skills.mage.level.val
 
     return int(max(melee, ranged, mage))
-
-  @property
-  def population(self):
-    return self.population_id.val

nmmo/entity/entity_manager.py

@@ -7,7 +7,7 @@
 from nmmo.entity.entity import Entity
 from nmmo.entity.npc import NPC
 from nmmo.entity.player import Player
-from nmmo.lib import colors, spawn
+from nmmo.lib import spawn
 from nmmo.systems import combat
 
 
@@ -126,7 +126,6 @@ def actions(self, realm):
 class PlayerManager(EntityGroup):
   def __init__(self, realm):
     super().__init__(realm)
-    self.palette = colors.Palette()
     self.loader  = self.realm.config.PLAYER_LOADER
     self.agents = None
     self.spawned = None
@@ -137,9 +136,9 @@ def reset(self):
     self.spawned = OrderedSet()
 
   def spawn_individual(self, r, c, idx):
-    pop, agent = next(self.agents)
+    agent = next(self.agents)
     agent      = agent(self.config, idx)
-    player     = Player(self.realm, (r, c), agent, self.palette.color(pop), pop)
+    player     = Player(self.realm, (r, c), agent)
     super().spawn(player)
 
   def spawn(self):

nmmo/entity/npc.py

@@ -3,7 +3,6 @@
 
 from nmmo.entity import entity
 from nmmo.io import action as Action
-from nmmo.lib.colors import Neon
 from nmmo.systems import combat, droptable
 from nmmo.systems.ai import policy
 from nmmo.systems import item as Item
@@ -46,15 +45,17 @@ def packet(self):
     return packet
 
 
+# pylint: disable=no-member
 class NPC(entity.Entity):
-  def __init__(self, realm, pos, iden, name, color, pop):
-    super().__init__(realm, pos, iden, name, color, pop)
+  def __init__(self, realm, pos, iden, name, npc_type):
+    super().__init__(realm, pos, iden, name)
     self.skills = skill.Combat(realm, self)
     self.realm = realm
     self.last_action = None
     self.droptable = None
     self.spawn_danger = None
     self.equipment = None
+    self.npc_type.update(npc_type)
 
   def update(self, realm, actions):
     super().update(realm, actions)
@@ -161,21 +162,21 @@ def is_npc(self) -> bool:
 
 class Passive(NPC):
   def __init__(self, realm, pos, iden):
-    super().__init__(realm, pos, iden, 'Passive', Neon.GREEN, -1)
+    super().__init__(realm, pos, iden, 'Passive', 1)
 
   def decide(self, realm):
     return policy.passive(realm, self)
 
 class PassiveAggressive(NPC):
   def __init__(self, realm, pos, iden):
-    super().__init__(realm, pos, iden, 'Neutral', Neon.ORANGE, -2)
+    super().__init__(realm, pos, iden, 'Neutral', 2)
 
   def decide(self, realm):
     return policy.neutral(realm, self)
 
 class Aggressive(NPC):
   def __init__(self, realm, pos, iden):
-    super().__init__(realm, pos, iden, 'Hostile', Neon.RED, -3)
+    super().__init__(realm, pos, iden, 'Hostile', 3)
 
   def decide(self, realm):
     return policy.hostile(realm, self)

nmmo/entity/player.py

@@ -4,11 +4,10 @@
 
 # pylint: disable=no-member
 class Player(entity.Entity):
-  def __init__(self, realm, pos, agent, color, pop):
-    super().__init__(realm, pos, agent.iden, agent.policy, color, pop)
+  def __init__(self, realm, pos, agent):
+    super().__init__(realm, pos, agent.iden, agent.policy)
 
     self.agent    = agent
-    self.pop      = pop
     self.immortal = realm.config.IMMORTAL
 
     # Scripted hooks
@@ -37,7 +36,7 @@ def __init__(self, realm, pos, agent, color, pop):
 
   @property
   def serial(self):
-    return self.population_id, self.ent_id
+    return self.ent_id
 
   @property
   def is_player(self) -> bool:
@@ -92,7 +91,6 @@ def packet(self):
     data = super().packet()
 
     data['entID']     = self.ent_id
-    data['annID']     = self.population
 
     data['resource']  = self.resources.packet()
     data['skills']    = self.skills.packet()

nmmo/io/action.py

@@ -258,11 +258,6 @@ def call(realm, entity, style, target):
     if entity.ent_id == target.ent_id:
       return None
 
-    #ADDED: POPULATION IMMUNITY
-    if not config.COMBAT_FRIENDLY_FIRE and entity.is_player \
-       and entity.population_id.val == target.population_id.val:
-      return None
-
     #Can't attack out of range
     if utils.linf(entity.pos, target.pos) > style.attack_range(config):
       return None
@@ -458,10 +453,10 @@ def call(realm, entity, item, target):
     if item.equipped.val or item.listed_price.val:
       return
 
-    if not (config.ITEM_GIVE_TO_FRIENDLY and
-            entity.population_id == target.population_id and        # the same team
+    if not (config.ITEM_ALLOW_GIFT and
             entity.ent_id != target.ent_id and                      # but not self
-            utils.linf(entity.pos, target.pos) == 0):               # the same tile
+            target.is_player and
+            entity.pos == target.pos):               # the same tile
       return
 
     if not target.inventory.space:
@@ -504,10 +499,10 @@ def call(realm, entity, amount, target):
     if not (target.is_player and target.alive):
       return
 
-    if not (config.ITEM_GIVE_TO_FRIENDLY and
-            entity.population_id == target.population_id and        # the same team
+    if not (config.ITEM_ALLOW_GIFT and
             entity.ent_id != target.ent_id and                      # but not self
-            utils.linf(entity.pos, target.pos) == 0):               # the same tile
+            target.is_player and
+            entity.pos == target.pos):                              # the same tile
       return
 
     if not isinstance(amount, int):

nmmo/lib/spawn.py

@@ -4,8 +4,6 @@ class SequentialLoader:
   '''config.PLAYER_LOADER that spreads out agent populations'''
   def __init__(self, config):
     items = config.PLAYERS
-    for idx, itm in enumerate(items):
-      itm.policyID = idx
 
     self.items = items
     self.idx   = -1
@@ -15,28 +13,7 @@ def __iter__(self):
 
   def __next__(self):
     self.idx = (self.idx + 1) % len(self.items)
-    return self.idx, self.items[self.idx]
-
-class TeamLoader:
-  '''config.PLAYER_LOADER that loads agent populations adjacent'''
-  def __init__(self, config):
-    items = config.PLAYERS
-    self.team_size = config.PLAYER_N // len(items)
-
-    for idx, itm in enumerate(items):
-      itm.policyID = idx
-
-    self.items = items
-    self.idx   = -1
-
-  def __iter__(self):
-    return self
-
-  def __next__(self):
-    self.idx += 1
-    team_idx  = self.idx // self.team_size
-    return team_idx, self.items[team_idx]
-
+    return self.items[self.idx]
 
 def spawn_continuous(config):
   '''Generates spawn positions for new agents

nmmo/overlay.py

@@ -133,9 +133,8 @@ def update(self, obs):
       '''Computes a count-based exploration map by painting
       tiles as agents walk over them'''
       for ent_id, agent in self.realm.realm.players.items():
-         pop  = agent.population_id.val
          r, c = agent.pos
-         self.values[r, c][pop] += 1
+         self.values[r, c][ent_id] += 1
 
    def register(self, obs):
       colors    = self.realm.realm.players.palette.colors