further compat changes for gymnasium update;

deprl/dep_controller.py

@@ -2,7 +2,7 @@
 import os
 from collections import deque
 
-import gym
+import gymnasium as gym
 import torch
 
 torch.set_default_dtype(torch.float32)

deprl/env_wrappers/wrappers.py

@@ -1,6 +1,6 @@
 from abc import ABC, abstractmethod
 
-import gym
+import gymnasium as gym
 import numpy as np
 
 import deprl  # noqa

deprl/vendor/tonic/environments/builders.py

@@ -3,7 +3,7 @@
 import os
 from types import SimpleNamespace
 
-import gym.wrappers
+from gymnasium import wrappers
 import numpy as np
 
 try:
@@ -55,7 +55,7 @@ def _builder(name, *args, **kwargs):
         environment.spec = SimpleNamespace(
             max_episode_steps=time_limit, id="ostrichrl-dmcontrol"
         )
-        return gym.wrappers.TimeLimit(environment, time_limit)
+        return wrappers.TimeLimit(environment, time_limit)
 
     return build_environment(_builder, *args, **kwargs, header=None)
 
@@ -97,7 +97,7 @@ def build_environment(
 
     # Remove the TimeLimit wrapper if needed.
     if not terminal_timeouts:
-        if type(environment) == gym.wrappers.TimeLimit:
+        if type(environment) == wrappers.TimeLimit:
             environment = environment.env
 
     # Add time as a feature if needed.

deprl/vendor/tonic/environments/wrappers.py

@@ -1,6 +1,6 @@
 """Environment wrappers."""
 
-import gym
+import gymnasium as gym
 import numpy as np
 
 

examples/example_load_baseline_myosuite.py

@@ -5,31 +5,32 @@
 
 import myosuite  # noqa
 from myosuite.utils import gym
+from deprl import env_wrappers
 
 import deprl
 
 env = gym.make("myoLegWalk-v0", reset_type="random")
+env = env_wrappers.GymWrapper(env)
 policy = deprl.load_baseline(env)
 
 env.seed(0)
 for ep in range(10):
     ep_steps = 0
     ep_tot_reward = 0
-    state, _ = env.reset()
+    state = env.reset()
 
     while True:
         # samples random action
         action = policy(state)
         # applies action and advances environment by one step
-        state, reward, terminated, truncated, info = env.step(action)
+        state, reward, done, info = env.step(action)
 
         ep_steps += 1
         ep_tot_reward += reward
         env.mj_render()
         time.sleep(0.01)
 
         # check if done
-        done = terminated or truncated
         if done or (ep_steps >= 1000):
             print(
                 f"Episode {ep} ending; steps={ep_steps}; reward={ep_tot_reward:0.3f};"

examples/example_only_dep_myosuite.py

@@ -1,7 +1,7 @@
 import time
 
-import gym
 import myosuite  # noqa
+from myosuite.utils import gym
 
 from deprl import env_wrappers
 from deprl.dep_controller import DEP

pyproject.toml

@@ -14,7 +14,7 @@ numpy = "^1.22.4"
 termcolor = "^2.2.0"
 pandas = "^2.0.1"
 gdown = "^5.1.0"
-gym = "0.13.0"
+gymnasium = "*"
 wandb = "^0.15.4"
 # torch = {version="2.1.0", source="pytorch-cpu"}
 torch = ">=2.1.0"

tests/test_files/test_settings.json

@@ -1,7 +1,7 @@
 
 {
  "tonic": {
-    "header": "import deprl, gym, myosuite",
+    "header": "import deprl, myosuite; from myosuite.utils import gym",
     "agent": "deprl.custom_agents.dep_factory(3, deprl.custom_mpo_torch.TunedMPO())(replay=deprl.replays.buffers.Buffer(return_steps=3, batch_size=256, steps_between_batches=1000, batch_iterations=30, steps_before_batches=2e5, size=int(1e4)))",
     "environment": "deprl.environments.Gym('myoLegWalk-v0', scaled_actions=False, reset_type='random')",
     "trainer": "deprl.custom_trainer.Trainer(steps=int(2e3), epoch_steps=int(1e3), save_steps=int(1e3))",

tests/test_files/test_settings_load.json

@@ -1,7 +1,7 @@
 
 {
  "tonic": {
-    "header": "import deprl, gym, myosuite",
+    "header": "import deprl, myosuite; from myosuite.utils import gym",
     "agent": "deprl.custom_agents.dep_factory(3, deprl.custom_mpo_torch.TunedMPO())(replay=deprl.replays.buffers.Buffer(return_steps=3, batch_size=256, steps_between_batches=1000, batch_iterations=30, steps_before_batches=2e5, size=int(1e4)))",
     "environment": "deprl.environments.Gym('myoLegWalk-v0', scaled_actions=False, reset_type='random')",
     "trainer": "deprl.custom_trainer.Trainer(steps=int(3e3), epoch_steps=int(1e3), save_steps=int(1e3))",

tests/test_files/test_settings_load_no_resume.json

@@ -1,7 +1,7 @@
 
 {
  "tonic": {
-    "header": "import deprl, gym, myosuite",
+    "header": "import deprl, myosuite; from myosuite.utils import gym",
     "agent": "deprl.custom_agents.dep_factory(3, deprl.custom_mpo_torch.TunedMPO())(replay=deprl.replays.buffers.Buffer(return_steps=3, batch_size=256, steps_between_batches=1000, batch_iterations=30, steps_before_batches=2e5, size=int(1e4)))",
     "environment": "deprl.environments.Gym('myoLegWalk-v0', scaled_actions=False, reset_type='random')",
     "trainer": "deprl.custom_trainer.Trainer(steps=int(3e3), epoch_steps=int(1e3), save_steps=int(1e3))",

tests/test_files/test_settings_load_resume.json

@@ -1,7 +1,7 @@
 
 {
  "tonic": {
-    "header": "import deprl, gym, myosuite",
+    "header": "import deprl, myosuite; from myosuite.utils import gym",
     "agent": "deprl.custom_agents.dep_factory(3, deprl.custom_mpo_torch.TunedMPO())(replay=deprl.replays.buffers.Buffer(return_steps=3, batch_size=256, steps_between_batches=1000, batch_iterations=30, steps_before_batches=2e5, size=int(1e4)))",
     "environment": "deprl.environments.Gym('myoLegWalk-v0', scaled_actions=False, reset_type='random')",
     "trainer": "deprl.custom_trainer.Trainer(steps=int(3e3), epoch_steps=int(1e3), save_steps=int(1e3))",