Feature/python3dot10 gym compatibility update

lbforaging.py

@@ -1,9 +1,9 @@
+'''Basic flow to see if the base install worked over one environment.'''
 import argparse
 import logging
-import random
 import time
-import gym
 import numpy as np
+import gymnasium as gym
 import lbforaging
 
 
@@ -13,7 +13,7 @@
 def _game_loop(env, render):
     """
     """
-    obs = env.reset()
+    _, _ = env.reset()
     done = False
 
     if render:
@@ -24,7 +24,7 @@ def _game_loop(env, render):
 
         actions = env.action_space.sample()
 
-        nobs, nreward, ndone, _ = env.step(actions)
+        _, nreward, ndone, _, _ = env.step(actions)
         if sum(nreward) > 0:
             print(nreward)
 
@@ -38,9 +38,11 @@ def _game_loop(env, render):
 
 def main(game_count=1, render=False):
     env = gym.make("Foraging-8x8-2p-2f-v2")
-    obs = env.reset()
 
-    for episode in range(game_count):
+    _, info = env.reset()
+    assert info == {}
+
+    for _ in range(game_count):
         _game_loop(env, render)
 
 
@@ -54,3 +56,5 @@ def main(game_count=1, render=False):
 
     args = parser.parse_args()
     main(args.times, args.render)
+
+    print("Done. NO RUNTIME ERRORS.")

lbforaging/__init__.py

@@ -1,5 +1,6 @@
-from gym.envs.registration import registry, register, make, spec
+from gymnasium.envs.registration import registry, register, make, spec
 from itertools import product
+from lbforaging import foraging
 
 sizes = range(5, 20)
 players = range(2, 20)

lbforaging/agents/__init__.py

@@ -1,5 +1 @@
-# from lbforaging.agents.random_agent import RandomAgent
-# from lbforaging.agents.heuristic_agent import H1, H2, H3, H4
-# from lbforaging.agents.q_agent import QAgent
-# from lbforaging.agents.monte_carlo import MonteCarloAgent
-# from lbforaging.agents.hba import HBAAgent
+from lbforaging.agents import *

lbforaging/agents/hba.py

@@ -1,8 +1,8 @@
-from . import QAgent
-from foraging import Env
+from lbforaging.agents.q_agent import QAgent
+from lbforaging.foraging.environment import ForagingEnv as Env
 import random
 import numpy as np
-from agents import H1, H2, H3, H4
+from lbforaging.agents.heuristic_agent import H1, H2, H3, H4
 from itertools import product
 from collections import defaultdict
 from functools import reduce

lbforaging/agents/heuristic_agent.py

@@ -1,7 +1,7 @@
 import random
 import numpy as np
-from foraging import Agent
-from foraging.environment import Action
+from lbforaging.foraging.agent import Agent
+from lbforaging.foraging.environment import Action
 
 
 class HeuristicAgent(Agent):
@@ -28,7 +28,7 @@ def _move_towards(self, target, allowed):
             raise ValueError("No simple path found")
 
     def step(self, obs):
-        raise NotImplemented("Heuristic agent is implemented by H1-H4")
+        raise NotImplementedError("Heuristic agent is implemented by H1-H4")
 
 
 class H1(HeuristicAgent):

lbforaging/agents/monte_carlo.py

@@ -8,7 +8,8 @@
 import plotly.graph_objs as go
 from networkx.drawing.nx_pydot import graphviz_layout
 
-from foraging import Agent, Env
+from lbforaging.foraging.agent import Agent
+from lbforaging.foraging.environment import ForagingEnv as Env
 
 MCTS_DEPTH = 15
 

lbforaging/agents/nn_agent.py

@@ -1,6 +1,6 @@
 import random
 
-from foraging import Agent
+from lbforaging.foraging.agent import Agent
 
 
 class NNAgent(Agent):

lbforaging/agents/q_agent.py

@@ -4,9 +4,10 @@
 import numpy as np
 import pandas as pd
 
-from agents import H1
-from lbforaging import Agent, Env
-from lbforaging.environment import Action
+from lbforaging.agents.heuristic_agent import H1
+from lbforaging.foraging.agent import Agent
+from lbforaging.foraging.environment import Action
+from lbforaging.foraging.environment import ForagingEnv as Env
 
 _CACHE = None
 

lbforaging/agents/random_agent.py

@@ -1,6 +1,6 @@
 import random
 
-from lbforaging import Agent
+from lbforaging.foraging.agent import Agent
 
 
 class RandomAgent(Agent):

lbforaging/foraging/agent.py

@@ -1,5 +1,4 @@
 import logging
-
 import numpy as np
 
 _MAX_INT = 999999
@@ -30,7 +29,7 @@ def _step(self, obs):
         return action
 
     def step(self, obs):
-        raise NotImplemented("You must implement an agent")
+        raise NotImplementedError("You must implement an agent")
 
     def _closest_food(self, obs, max_food_level=None, start=None):
 

lbforaging/foraging/environment.py

@@ -2,9 +2,9 @@
 from collections import namedtuple, defaultdict
 from enum import Enum
 from itertools import product
-from gym import Env
-import gym
-from gym.utils import seeding
+from gymnasium import Env
+import gymnasium as gym
+from gymnasium.utils import seeding
 import numpy as np
 
 
@@ -93,7 +93,7 @@ def __init__(
         self.field = np.zeros(field_size, np.int32)
 
         self.penalty = penalty
-        
+
         self.max_food = max_food
         self._food_spawned = 0.0
         self.max_player_level = max_player_level
@@ -109,13 +109,14 @@ def __init__(
         self._grid_observation = grid_observation
 
         self.action_space = gym.spaces.Tuple(tuple([gym.spaces.Discrete(6)] * len(self.players)))
-        self.observation_space = gym.spaces.Tuple(tuple([self._get_observation_space()] * len(self.players)))
+        self.observation_space = gym.spaces.Tuple(
+            tuple([self._get_observation_space()] * len(self.players)))
 
         self.viewer = None
 
         self.n_agents = len(self.players)
 
-    def seed(self, seed=None):
+    def seed(self, seed=0):
         self.np_random, seed = seeding.np_random(seed)
         return [seed]
 
@@ -159,7 +160,15 @@ def _get_observation_space(self):
             min_obs = np.stack([agents_min, foods_min, access_min])
             max_obs = np.stack([agents_max, foods_max, access_max])
 
-        return gym.spaces.Box(np.array(min_obs), np.array(max_obs), dtype=np.float32)
+        low_obs = np.array(min_obs)
+        high_obs = np.array(max_obs)
+        assert len(low_obs) == len(high_obs)
+        composed_obs_space = gym.spaces.Box(
+            low=low_obs,
+            high=high_obs,
+            shape=[len(low_obs)],
+            dtype=np.float32)
+        return composed_obs_space
 
     @classmethod
     def from_obs(cls, obs):
@@ -170,7 +179,15 @@ def from_obs(cls, obs):
             player.score = p.score if p.score else 0
             players.append(player)
 
-        env = cls(players, None, None, None, None)
+        env = cls(
+            players=players,
+            max_player_level=None,
+            field_size=None,
+            max_food=None,
+            sight=None,
+            max_episode_steps=50,
+            force_coop=False
+        )
         env.field = np.copy(obs.field)
         env.current_step = obs.current_step
         env.sight = obs.sight
@@ -202,6 +219,14 @@ def _gen_valid_moves(self):
             for player in self.players
         }
 
+    def test_gen_valid_moves(self) -> bool:
+        ''' Wrapper around a private method to test if the generated moves are valid. '''
+        try:
+            self._gen_valid_moves()
+        except Exception as _:
+            return False
+        return True
+
     def neighborhood(self, row, col, distance=1, ignore_diag=False):
         if not ignore_diag:
             return self.field[
@@ -253,8 +278,8 @@ def spawn_food(self, max_food, max_level):
 
         while food_count < max_food and attempts < 1000:
             attempts += 1
-            row = self.np_random.randint(1, self.rows - 1)
-            col = self.np_random.randint(1, self.cols - 1)
+            row = self.np_random.integers(1, self.rows - 1)
+            col = self.np_random.integers(1, self.cols - 1)
 
             # check if it has neighbors:
             if (
@@ -269,7 +294,7 @@ def spawn_food(self, max_food, max_level):
                 if min_level == max_level
                 # ! this is excluding food of level `max_level` but is kept for
                 # ! consistency with prior LBF versions
-                else self.np_random.randint(min_level, max_level)
+                else self.np_random.integers(min_level, max_level)
             )
             food_count += 1
         self._food_spawned = self.field.sum()
@@ -290,12 +315,12 @@ def spawn_players(self, max_player_level):
             player.reward = 0
 
             while attempts < 1000:
-                row = self.np_random.randint(0, self.rows)
-                col = self.np_random.randint(0, self.cols)
+                row = self.np_random.integers(0, self.rows)
+                col = self.np_random.integers(0, self.cols)
                 if self._is_empty_location(row, col):
                     player.setup(
                         (row, col),
-                        self.np_random.randint(1, max_player_level + 1),
+                        self.np_random.integers(1, max_player_level + 1),
                         self.field_size,
                     )
                     break
@@ -465,9 +490,15 @@ def get_player_reward(observation):
             assert self.observation_space[i].contains(obs), \
                 f"obs space error: obs: {obs}, obs_space: {self.observation_space[i]}"
 
-        return nobs, nreward, ndone, ninfo
+        truncated_term = False
+        # To turn this into a single agent task, you need to sum the nreward and the ndone
+        return nobs, nreward, ndone, truncated_term, ninfo
+
+    def test_make_gym_obs(self):
+        ''' Test wrapper to test the current observation in a public manner. '''
+        return self._make_gym_obs()
 
-    def reset(self):
+    def reset(self, *, seed=None, options=None):
         self.field = np.zeros(self.field_size, np.int32)
         self.spawn_players(self.max_player_level)
         player_levels = sorted([player.level for player in self.players])
@@ -479,8 +510,10 @@ def reset(self):
         self._game_over = False
         self._gen_valid_moves()
 
-        nobs, _, _, _ = self._make_gym_obs()
-        return nobs
+        nobs, _, _, _, _ = self._make_gym_obs()
+        # The new gym spec and gym utils require that
+        # the new observation and a dictionary with info is returned
+        return nobs, {}
 
     def step(self, actions):
         self.current_step += 1

lbforaging/foraging/rendering.py

@@ -29,6 +29,18 @@
     """
     )
 
+try:
+    from pyglet import gl
+except ImportError as e:
+    raise ImportError(
+        """
+    Cannot 'from pyglet import gl'
+    HINT: you can install pyglet directly via 'pip install pyglet'.
+    But if you really just want to install all Gym dependencies and not have to think about it,
+    'pip install -e .[all]' or 'pip install gym[all]' will do it.
+    """
+    )
+
 try:
     from pyglet.gl import *
 except ImportError as e:

setup.py

@@ -13,8 +13,9 @@
         "Programming Language :: Python :: 3.5",
         "Programming Language :: Python :: 3.6",
         "Programming Language :: Python :: 3.7",
+        "Programming Language :: Python :: 3.10",
     ],
-    install_requires=["numpy", "gym==0.21", "pyglet"],
+    install_requires=["numpy", "gym==0.26", "pyglet<2"],
     extras_require={"test": ["pytest"]},
     include_package_data=True,
 )