Updated to Gymnasium v0.26.2

.gitignore

@@ -106,3 +106,13 @@ venv.bak/
 
 # mypy
 .mypy_cache/
+
+# tests
+test_result.txt
+test_results.txt
+
+# VS Code
+.vscode
+
+# MacOS
+.DS_Store

README.md

@@ -73,7 +73,7 @@ This simulator was created as part of work done at [Mila](https://mila.quebec/).
 
 Requirements:
 - Python 3.5+
-- OpenAI Gym
+- Gymnasium
 - NumPy
 - Pyglet (OpenGL 3D graphics)
 - GPU for 3D graphics acceleration (optional)

gym_miniworld/envs/__init__.py

@@ -1,6 +1,6 @@
 import inspect
 
-import gym
+import gymnasium as gym
 
 from gym_miniworld.envs.collecthealth import CollectHealth
 from gym_miniworld.envs.fourrooms import FourRooms

gym_miniworld/envs/collecthealth.py

@@ -43,7 +43,7 @@ def _gen_world(self):
         self.health = 100
 
     def step(self, action):
-        obs, reward, done, info = super().step(action)
+        obs, reward, termination, truncation, info = super().step(action)
 
         self.health -= 2
 
@@ -62,9 +62,9 @@ def step(self, action):
             reward = 2
         else:
             reward = -100
-            done = True
+            termination = True
 
         # Pass current health value in info dict
         info["health"] = self.health
 
-        return obs, reward, done, info
+        return obs, reward, termination, truncation, info

gym_miniworld/envs/fourrooms.py

@@ -1,4 +1,4 @@
-from gym import spaces
+from gymnasium import spaces
 
 from gym_miniworld.entity import Box
 from gym_miniworld.miniworld import MiniWorldEnv
@@ -37,10 +37,10 @@ def _gen_world(self):
         self.place_agent()
 
     def step(self, action):
-        obs, reward, done, info = super().step(action)
+        obs, reward, termination, truncation, info = super().step(action)
 
         if self.near(self.box):
             reward += self._reward()
-            done = True
+            termination = True
 
-        return obs, reward, done, info
+        return obs, reward, termination, truncation, info

gym_miniworld/envs/hallway.py

@@ -1,6 +1,6 @@
 import math
 
-from gym import spaces
+from gymnasium import spaces
 
 from gym_miniworld.entity import Box
 from gym_miniworld.miniworld import MiniWorldEnv
@@ -34,10 +34,10 @@ def _gen_world(self):
         )
 
     def step(self, action):
-        obs, reward, done, info = super().step(action)
+        obs, reward, termination, truncation, info = super().step(action)
 
         if self.near(self.box):
             reward += self._reward()
-            done = True
+            termination = True
 
-        return obs, reward, done, info
+        return obs, reward, termination, truncation, info

gym_miniworld/envs/maze.py

@@ -1,4 +1,4 @@
-from gym import spaces
+from gymnasium import spaces
 
 from gym_miniworld.entity import Box
 from gym_miniworld.miniworld import MiniWorldEnv
@@ -102,13 +102,13 @@ def visit(i, j):
         self.place_agent()
 
     def step(self, action):
-        obs, reward, done, info = super().step(action)
+        obs, reward, termination, truncation, info = super().step(action)
 
         if self.near(self.box):
             reward += self._reward()
-            done = True
+            termination = True
 
-        return obs, reward, done, info
+        return obs, reward, termination, truncation, info
 
 
 class MazeS2(Maze):

gym_miniworld/envs/oneroom.py

@@ -1,4 +1,4 @@
-from gym import spaces
+from gymnasium import spaces
 
 from gym_miniworld.entity import Box
 from gym_miniworld.miniworld import MiniWorldEnv
@@ -27,13 +27,13 @@ def _gen_world(self):
         self.place_agent()
 
     def step(self, action):
-        obs, reward, done, info = super().step(action)
+        obs, reward, termination, truncation, info = super().step(action)
 
         if self.near(self.box):
             reward += self._reward()
-            done = True
+            termination = True
 
-        return obs, reward, done, info
+        return obs, reward, termination, truncation, info
 
 
 class OneRoomS6(OneRoom):

gym_miniworld/envs/pickupobjs.py

@@ -1,4 +1,4 @@
-from gym import spaces
+from gymnasium import spaces
 
 from gym_miniworld.entity import Ball, Box, Key
 from gym_miniworld.miniworld import MiniWorldEnv
@@ -49,7 +49,7 @@ def _gen_world(self):
         self.num_picked_up = 0
 
     def step(self, action):
-        obs, reward, done, info = super().step(action)
+        obs, reward, termination, truncation, info = super().step(action)
 
         if self.agent.carrying:
             self.entities.remove(self.agent.carrying)
@@ -58,6 +58,6 @@ def step(self, action):
             reward = 1
 
             if self.num_picked_up == self.num_objs:
-                done = True
+                termination = True
 
-        return obs, reward, done, info
+        return obs, reward, termination, truncation, info

gym_miniworld/envs/putnext.py

@@ -31,11 +31,11 @@ def _gen_world(self):
         self.place_agent()
 
     def step(self, action):
-        obs, reward, done, info = super().step(action)
+        obs, reward, termination, truncation, info = super().step(action)
 
         if not self.agent.carrying:
             if self.near(self.red_box, self.yellow_box):
                 reward += self._reward()
-                done = True
+                termination = True
 
-        return obs, reward, done, info
+        return obs, reward, termination, truncation, info

gym_miniworld/envs/remotebot.py

@@ -2,12 +2,11 @@
 
 import math
 
-import gym
+import gymnasium as gym
 import numpy
 import numpy as np
 import pyglet
-from gym import spaces
-from gym.utils import seeding
+from gymnasium import spaces
 
 # Try importing ZMQ
 from pyglet.gl import (
@@ -23,6 +22,9 @@
 
 from gym_miniworld.miniworld import MiniWorldEnv
 
+# from gym.utils import seeding
+
+
 try:
     import zmq
 except ImportError:
@@ -70,6 +72,7 @@ def __init__(
         serverPort=SERVER_PORT,
         obs_width=80,
         obs_height=60,
+        render_mode=None,
     ):
         assert zmq is not None, "Please install zmq (pip3 install zmq)"
 
@@ -94,6 +97,7 @@ def __init__(
 
         # For rendering
         self.window = None
+        self.render_mode = render_mode
 
         # We continually stream in images and then just take the latest one.
         self.latest_img = None
@@ -113,7 +117,6 @@ def __init__(
         self.socket.connect(addr_str)
 
         # Initialize the state
-        self.seed()
         self.reset()
         print("Connected")
 
@@ -128,7 +131,7 @@ def _recv_frame(self):
 
         self.img = img
 
-    def reset(self):
+    def reset(self, seed, options):
         # Step count since episode start
         self.step_count = 0
 
@@ -143,11 +146,7 @@ def reset(self):
         # Receive a camera image from the server
         self._recv_frame()
 
-        return self.img
-
-    def seed(self, seed=None):
-        self.np_random, _ = seeding.np_random(seed)
-        return [seed]
+        return self.img, {}
 
     def step(self, action):
         # Send the action to the server
@@ -162,17 +161,26 @@ def step(self, action):
 
         # We don't care about rewards or episodes since we're not training
         reward = 0
-        done = False
-
-        return self.img, reward, done, {}
+        termination = False
+        truncation = False
+
+        return self.img, reward, termination, truncation, {}
+
+    def render(self, close=False):
+        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
 
-    def render(self, mode="human", close=False):
         if close:
             if self.window:
                 self.window.close()
             return
 
-        if mode == "rgb_array":
+        if self.render_mode == "rgb_array":
             return self.img
 
         if self.window is None:
@@ -209,6 +217,6 @@ def render(self, mode="human", close=False):
 
         # If we are not running the Pyglet event loop,
         # we have to manually flip the buffers and dispatch events
-        if mode == "human":
+        if self.render_mode == "human":
             self.window.flip()
             self.window.dispatch_events()

gym_miniworld/envs/roomobjs.py

@@ -39,5 +39,5 @@ def _gen_world(self):
         self.place_agent()
 
     def step(self, action):
-        obs, reward, done, info = super().step(action)
-        return obs, reward, done, info
+        obs, reward, termination, truncation, info = super().step(action)
+        return obs, reward, termination, truncation, info

gym_miniworld/envs/sidewalk.py

@@ -1,7 +1,7 @@
 import math
 
 import numpy as np
-from gym import spaces
+from gymnasium import spaces
 
 from gym_miniworld.entity import Box, MeshEnt
 from gym_miniworld.miniworld import MiniWorldEnv
@@ -65,15 +65,15 @@ def _gen_world(self):
         self.place_agent(room=sidewalk, min_z=0, max_z=1.5)
 
     def step(self, action):
-        obs, reward, done, info = super().step(action)
+        obs, reward, termination, truncation, info = super().step(action)
 
         # Walking into the street ends the episode
         if self.street.point_inside(self.agent.pos):
             reward = 0
-            done = True
+            termination = True
 
         if self.near(self.box):
             reward += self._reward()
-            done = True
+            termination = True
 
-        return obs, reward, done, info
+        return obs, reward, termination, truncation, info

gym_miniworld/envs/sign.py

@@ -1,9 +1,9 @@
 import math
 from typing import Optional, Tuple, Union
 
-import gym
-from gym.core import ObsType
-from gym.spaces import Dict, Discrete
+import gymnasium as gym
+from gymnasium.core import ObsType
+from gymnasium.spaces import Dict, Discrete
 
 from gym_miniworld.entity import COLOR_NAMES, Box, Key, MeshEnt, TextFrame
 from gym_miniworld.miniworld import MiniWorldEnv
@@ -126,14 +126,15 @@ def _gen_world(self):
         self.place_agent(min_x=4, max_x=5, min_z=4, max_z=6)
 
     def step(self, action):
-        obs, reward, done, info = super().step(action)
+        obs, reward, termination, truncation, info = super().step(action)
+
         if action == self.actions.move_forward + 1:  # custom end episode action
-            done = True
+            termination = True
 
         for obj_index, object_pair in enumerate(self._objects):
             for color_index, obj in enumerate(object_pair):
                 if self.near(obj):
-                    done = True
+                    termination = True
                     reward = (
                         float(
                             color_index == self._color_index and obj_index == self._goal
@@ -143,20 +144,13 @@ def step(self, action):
                     )
 
         state = {"obs": obs, "goal": self._goal}
-        return state, reward, done, info
+        return state, reward, termination, truncation, info
 
     def reset(
         self,
         *,
         seed: Optional[int] = None,
-        return_info: bool = False,
         options: Optional[dict] = None,
     ) -> Union[ObsType, Tuple[ObsType, dict]]:
-        if return_info:
-            obs, info = super().reset(
-                seed=seed, return_info=return_info, options=options
-            )
-            return {"obs": obs, "goal": self._goal}, info
-        else:
-            obs = super().reset(seed=seed, return_info=return_info, options=options)
-            return {"obs": obs, "goal": self._goal}
+        obs, info = super().reset(seed=seed, options=options)
+        return {"obs": obs, "goal": self._goal}, info