Updated gymnasium to be equivalent to gym v26.2 (#36)

.github/workflows/build.yml

@@ -1,6 +1,9 @@
 name: build
 on: [pull_request, push]
 
+permissions:
+  contents: read # to fetch code (actions/checkout)
+
 jobs:
   build:
     runs-on: ubuntu-latest

.github/workflows/pre-commit.yml

@@ -5,6 +5,10 @@ on:
   pull_request:
   push:
     branches: [master]
+
+permissions:
+  contents: read # to fetch code (actions/checkout)
+
 jobs:
   pre-commit:
     runs-on: ubuntu-latest

gymnasium/envs/box2d/bipedal_walker.py

@@ -608,6 +608,14 @@ class BipedalWalker(gym.Env, EzPickle):
         return np.array(state, dtype=np.float32), reward, terminated, False, {}
 
     def render(self):
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+
         try:
             import pygame
             from pygame import gfxdraw

gymnasium/envs/box2d/car_racing.py

@@ -569,7 +569,15 @@ class CarRacing(gym.Env, EzPickle):
         return self.state, step_reward, terminated, truncated, {}
 
     def render(self):
-        return self._render(self.render_mode)
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+        else:
+            return self._render(self.render_mode)
 
     def _render(self, mode: str):
         assert mode in self.metadata["render_modes"]

gymnasium/envs/box2d/lunar_lander.py

@@ -602,6 +602,14 @@ class LunarLander(gym.Env, EzPickle):
         return np.array(state, dtype=np.float32), reward, terminated, False, {}
 
     def render(self):
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+
         try:
             import pygame
             from pygame import gfxdraw

gymnasium/envs/classic_control/acrobot.py

@@ -4,7 +4,9 @@ from typing import Optional
 import numpy as np
 from numpy import cos, pi, sin
 
+import gymnasium as gym
 from gymnasium import Env, spaces
+from gymnasium.envs.classic_control import utils
 from gymnasium.error import DependencyNotInstalled
 
 __copyright__ = "Copyright 2013, RLPy http://acl.mit.edu/RLPy"
@@ -20,7 +22,6 @@ __author__ = "Christoph Dann <cdann@cdann.de>"
 
 # SOURCE:
 # https://github.com/rlpy/rlpy/blob/master/rlpy/Domains/Acrobot.py
-from gymnasium.envs.classic_control import utils
 
 
 class AcrobotEnv(Env):
@@ -280,6 +281,14 @@ class AcrobotEnv(Env):
         return dtheta1, dtheta2, ddtheta1, ddtheta2, 0.0
 
     def render(self):
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+
         try:
             import pygame
             from pygame import gfxdraw

gymnasium/envs/classic_control/cartpole.py

@@ -209,6 +209,14 @@ class CartPoleEnv(gym.Env[np.ndarray, Union[int, np.ndarray]]):
         return np.array(self.state, dtype=np.float32), {}
 
     def render(self):
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+
         try:
             import pygame
             from pygame import gfxdraw

gymnasium/envs/classic_control/continuous_mountain_car.py

@@ -193,6 +193,14 @@ class Continuous_MountainCarEnv(gym.Env):
         return np.sin(3 * xs) * 0.45 + 0.55
 
     def render(self):
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+
         try:
             import pygame
             from pygame import gfxdraw

gymnasium/envs/classic_control/mountain_car.py

@@ -170,6 +170,14 @@ class MountainCarEnv(gym.Env):
         return np.sin(3 * xs) * 0.45 + 0.55
 
     def render(self):
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+
         try:
             import pygame
             from pygame import gfxdraw

gymnasium/envs/classic_control/pendulum.py

@@ -167,6 +167,14 @@ class PendulumEnv(gym.Env):
         return np.array([np.cos(theta), np.sin(theta), thetadot], dtype=np.float32)
 
     def render(self):
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+
         try:
             import pygame
             from pygame import gfxdraw

gymnasium/envs/toy_text/blackjack.py

@@ -191,6 +191,14 @@ class BlackjackEnv(gym.Env):
         return self._get_obs(), {}
 
     def render(self):
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+
         try:
             import pygame
         except ImportError:

gymnasium/envs/toy_text/cliffwalking.py

@@ -5,6 +5,7 @@ from typing import Optional
 
 import numpy as np
 
+import gymnasium as gym
 from gymnasium import Env, spaces
 from gymnasium.envs.toy_text.utils import categorical_sample
 from gymnasium.error import DependencyNotInstalled
@@ -163,6 +164,14 @@ class CliffWalkingEnv(Env):
         return int(self.s), {"prob": 1}
 
     def render(self):
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+
         if self.render_mode == "ansi":
             return self._render_text()
         else:

gymnasium/envs/toy_text/frozen_lake.py

@@ -5,6 +5,7 @@ from typing import List, Optional
 
 import numpy as np
 
+import gymnasium as gym
 from gymnasium import Env, spaces, utils
 from gymnasium.envs.toy_text.utils import categorical_sample
 from gymnasium.error import DependencyNotInstalled
@@ -268,6 +269,14 @@ class FrozenLakeEnv(Env):
         return int(self.s), {"prob": 1}
 
     def render(self):
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+
         if self.render_mode == "ansi":
             return self._render_text()
         else:  # self.render_mode in {"human", "rgb_array"}:

gymnasium/envs/toy_text/taxi.py

@@ -5,6 +5,7 @@ from typing import Optional
 
 import numpy as np
 
+import gymnasium as gym
 from gymnasium import Env, spaces, utils
 from gymnasium.envs.toy_text.utils import categorical_sample
 from gymnasium.error import DependencyNotInstalled
@@ -279,7 +280,14 @@ class TaxiEnv(Env):
         return int(self.s), {"prob": 1.0, "action_mask": self.action_mask(self.s)}
 
     def render(self):
-        if self.render_mode == "ansi":
+        if self.render_mode is None:
+            gym.logger.warn(
+                "You are calling render method without specifying any render mode. "
+                "You can specify the render_mode at initialization, "
+                f'e.g. gym("{self.spec.id}", render_mode="rgb_array")'
+            )
+            return
+        elif self.render_mode == "ansi":
             return self._render_text()
         else:  # self.render_mode in {"human", "rgb_array"}:
             return self._render_gui(self.render_mode)

gymnasium/spaces/graph.py

@@ -14,8 +14,8 @@ class GraphInstance(NamedTuple):
     """A Graph space instance.
 
     * nodes (np.ndarray): an (n x ...) sized array representing the features for n nodes, (...) must adhere to the shape of the node space.
-    * edges (Optional[np.ndarray]): an (m x ...) sized array representing the features for m nodes, (...) must adhere to the shape of the edge space.
+    * edges (Optional[np.ndarray]): an (m x ...) sized array representing the features for m edges, (...) must adhere to the shape of the edge space.
-    * edge_links (Optional[np.ndarray]): an (m x 2) sized array of ints representing the two nodes that each edge connects.
+    * edge_links (Optional[np.ndarray]): an (m x 2) sized array of ints representing the indices of the two nodes that each edge connects.
     """
 
     nodes: np.ndarray

gymnasium/vector/async_vector_env.py

@@ -566,9 +566,10 @@ def _worker(index, env_fn, pipe, parent_pipe, shared_memory, error_queue):
                     info,
                 ) = env.step(data)
                 if terminated or truncated:
-                    old_observation = observation
+                    old_observation, old_info = observation, info
                     observation, info = env.reset()
                     info["final_observation"] = old_observation
+                    info["final_info"] = old_info
                 pipe.send(((observation, reward, terminated, truncated, info), True))
             elif command == "seed":
                 env.seed(data)
@@ -636,10 +637,10 @@ def _worker_shared_memory(index, env_fn, pipe, parent_pipe, shared_memory, error
                     info,
                 ) = env.step(data)
                 if terminated or truncated:
-                    old_observation = observation
+                    old_observation, old_info = observation, info
                     observation, info = env.reset()
                     info["final_observation"] = old_observation
-
+                    info["final_info"] = old_info
                 write_to_shared_memory(
                     observation_space, index, observation, shared_memory
                 )

gymnasium/vector/sync_vector_env.py

@@ -150,9 +150,10 @@ class SyncVectorEnv(VectorEnv):
             ) = env.step(action)
 
             if self._terminateds[i] or self._truncateds[i]:
-                old_observation = observation
+                old_observation, old_info = observation, info
                 observation, info = env.reset()
                 info["final_observation"] = old_observation
+                info["final_info"] = old_info
             observations.append(observation)
             infos = self._add_info(infos, info, i)
         self.observations = concatenate(

gymnasium/version.py

@@ -1 +1 @@
-VERSION = "0.26.1"
+VERSION = "0.26.2"

gymnasium/wrappers/atari_preprocessing.py

@@ -98,7 +98,6 @@ class AtariPreprocessing(gym.Wrapper):
                 np.empty(env.observation_space.shape, dtype=np.uint8),
             ]
 
-        self.ale = env.unwrapped.ale
         self.lives = 0
         self.game_over = False
 
@@ -112,6 +111,11 @@ class AtariPreprocessing(gym.Wrapper):
             low=_low, high=_high, shape=_shape, dtype=_obs_dtype
         )
 
+    @property
+    def ale(self):
+        """Make ale as a class property to avoid serialization error."""
+        return self.env.unwrapped.ale
+
     def step(self, action):
         """Applies the preprocessing for an :meth:`env.step`."""
         total_reward, terminated, truncated, info = 0.0, False, False, {}

setup.py

@@ -18,7 +18,7 @@ extras = {
     "box2d": ["box2d-py==2.3.5", "pygame==2.1.0", "swig==4.*"],
     "classic_control": ["pygame==2.1.0"],
     "mujoco_py": ["mujoco_py<2.2,>=2.1"],
-    "mujoco": ["mujoco==2.2.0", "imageio>=2.14.1"],
+    "mujoco": ["mujoco==2.2", "imageio>=2.14.1"],
     "toy_text": ["pygame==2.1.0"],
     "other": ["lz4>=3.1.0", "opencv-python>=3.0", "matplotlib>=3.0", "moviepy>=1.0.0"],
 }

tests/vector/test_vector_env.py

@@ -1,10 +1,13 @@
+from functools import partial
+
 import numpy as np
 import pytest
 
-from gymnasium.spaces import Tuple
+from gymnasium.spaces import Discrete, Tuple
 from gymnasium.vector.async_vector_env import AsyncVectorEnv
 from gymnasium.vector.sync_vector_env import SyncVectorEnv
 from gymnasium.vector.vector_env import VectorEnv
+from tests.testing_env import GenericTestEnv
 from tests.vector.utils import CustomSpace, make_env
 
 
@@ -58,3 +61,65 @@ def test_custom_space_vector_env():
 
     assert isinstance(env.single_action_space, CustomSpace)
     assert isinstance(env.action_space, Tuple)
+
+
+@pytest.mark.parametrize(
+    "vectoriser",
+    (
+        SyncVectorEnv,
+        partial(AsyncVectorEnv, shared_memory=True),
+        partial(AsyncVectorEnv, shared_memory=False),
+    ),
+    ids=["Sync", "Async with shared memory", "Async without shared memory"],
+)
+def test_final_obs_info(vectoriser):
+    """Tests that the vector environments correctly return the final observation and info."""
+
+    def reset_fn(self, seed=None, options=None):
+        return 0, {"reset": True}
+
+    def thunk():
+        return GenericTestEnv(
+            action_space=Discrete(4),
+            observation_space=Discrete(4),
+            reset_fn=reset_fn,
+            step_fn=lambda self, action: (
+                action if action < 3 else 0,
+                0,
+                action >= 3,
+                False,
+                {"action": action},
+            ),
+        )
+
+    env = vectoriser([thunk])
+    obs, info = env.reset()
+    assert obs == np.array([0]) and info == {
+        "reset": np.array([True]),
+        "_reset": np.array([True]),
+    }
+
+    obs, _, termination, _, info = env.step([1])
+    assert (
+        obs == np.array([1])
+        and termination == np.array([False])
+        and info == {"action": np.array([1]), "_action": np.array([True])}
+    )
+
+    obs, _, termination, _, info = env.step([2])
+    assert (
+        obs == np.array([2])
+        and termination == np.array([False])
+        and info == {"action": np.array([2]), "_action": np.array([True])}
+    )
+
+    obs, _, termination, _, info = env.step([3])
+    assert (
+        obs == np.array([0])
+        and termination == np.array([True])
+        and info["reset"] == np.array([True])
+    )
+    assert "final_observation" in info and "final_info" in info
+    assert info["final_observation"] == np.array([0]) and info["final_info"] == {
+        "action": 3
+    }