Spinup Gym to Gymnasium Migration with Package version updates

docs/user/installation.rst

@@ -83,7 +83,11 @@ Installing Spinning Up
     cd spinningup
     pip install -e .
 
-.. admonition:: You Should Know
+.. admonition:: 
+    *Note* I had to use `pip wheel --no-deps -w dist .` to move past this step. A reference was deprecated and I believe the omittion of
+    re-downloading of dependencies helped. I will need to look into this more, getting it up and running was the priority.
+
+    You Should Know
 
     Spinning Up defaults to installing everything in Gym **except** the MuJoCo environments. In case you run into any trouble with the Gym installation, check out the `Gym`_ github page for help. If you want the MuJoCo environments, see the optional installation section below.
 

setup.py

@@ -18,16 +18,16 @@
         'gym[atari,box2d,classic_control]~=0.15.3',
         'ipython',
         'joblib',
-        'matplotlib==3.1.1',
+        'matplotlib==3.10.0',
         'mpi4py',
         'numpy',
         'pandas',
         'pytest',
         'psutil',
         'scipy',
-        'seaborn==0.8.1',
-        'tensorflow>=1.8.0,<2.0',
-        'torch==1.3.1',
+        'seaborn==0.13.2',
+        'tensorflow==2.18.0',
+        'torch==2.6.0',
         'tqdm'
     ],
     description="Teaching tools for introducing people to deep RL.",

spinup/algos/pytorch/ddpg/ddpg.py

@@ -2,7 +2,7 @@
 import numpy as np
 import torch
 from torch.optim import Adam
-import gym
+import gymnasium
 import time
 import spinup.algos.pytorch.ddpg.core as core
 from spinup.utils.logx import EpochLogger

spinup/algos/pytorch/ppo/core.py

@@ -1,6 +1,6 @@
 import numpy as np
 import scipy.signal
-from gym.spaces import Box, Discrete
+from gymnasium.spaces import Box, Discrete
 
 import torch
 import torch.nn as nn

spinup/algos/pytorch/ppo/ppo.py

@@ -1,7 +1,7 @@
 import numpy as np
 import torch
 from torch.optim import Adam
-import gym
+import gymnasium
 import time
 import spinup.algos.pytorch.ppo.core as core
 from spinup.utils.logx import EpochLogger

spinup/algos/pytorch/sac/sac.py

@@ -3,7 +3,7 @@
 import numpy as np
 import torch
 from torch.optim import Adam
-import gym
+import gymnasium
 import time
 import spinup.algos.pytorch.sac.core as core
 from spinup.utils.logx import EpochLogger

spinup/algos/pytorch/td3/td3.py

@@ -3,7 +3,7 @@
 import numpy as np
 import torch
 from torch.optim import Adam
-import gym
+import gymnasium
 import time
 import spinup.algos.pytorch.td3.core as core
 from spinup.utils.logx import EpochLogger

spinup/algos/pytorch/vpg/core.py

@@ -1,6 +1,6 @@
 import numpy as np
 import scipy.signal
-from gym.spaces import Box, Discrete
+from gymnasium.spaces import Box, Discrete
 
 import torch
 import torch.nn as nn

spinup/algos/pytorch/vpg/vpg.py

@@ -1,7 +1,7 @@
 import numpy as np
 import torch
 from torch.optim import Adam
-import gym
+import gymnasium
 import time
 import spinup.algos.pytorch.vpg.core as core
 from spinup.utils.logx import EpochLogger

spinup/algos/tf1/ddpg/ddpg.py

@@ -1,6 +1,6 @@
 import numpy as np
 import tensorflow as tf
-import gym
+import gymnasium
 import time
 from spinup.algos.tf1.ddpg import core
 from spinup.algos.tf1.ddpg.core import get_vars

spinup/algos/tf1/ppo/core.py

@@ -1,7 +1,7 @@
 import numpy as np
 import tensorflow as tf
 import scipy.signal
-from gym.spaces import Box, Discrete
+from gymnasium.spaces import Box, Discrete
 
 EPS = 1e-8
 

spinup/algos/tf1/ppo/ppo.py

@@ -1,6 +1,6 @@
 import numpy as np
 import tensorflow as tf
-import gym
+import gymnasium as gym
 import time
 import spinup.algos.tf1.ppo.core as core
 from spinup.utils.logx import EpochLogger

spinup/algos/tf1/sac/sac.py

@@ -1,6 +1,6 @@
 import numpy as np
 import tensorflow as tf
-import gym
+import gymnasium
 import time
 from spinup.algos.tf1.sac import core
 from spinup.algos.tf1.sac.core import get_vars

spinup/algos/tf1/td3/td3.py

@@ -1,6 +1,6 @@
 import numpy as np
 import tensorflow as tf
-import gym
+import gymnasium
 import time
 from spinup.algos.tf1.td3 import core
 from spinup.algos.tf1.td3.core import get_vars

spinup/algos/tf1/trpo/core.py

@@ -1,7 +1,7 @@
 import numpy as np
 import tensorflow as tf
 import scipy.signal
-from gym.spaces import Box, Discrete
+from gymnasium.spaces import Box, Discrete
 
 EPS = 1e-8
 

spinup/algos/tf1/trpo/trpo.py

@@ -1,6 +1,6 @@
 import numpy as np
 import tensorflow as tf
-import gym
+import gymnasium
 import time
 import spinup.algos.tf1.trpo.core as core
 from spinup.utils.logx import EpochLogger

spinup/algos/tf1/vpg/core.py

@@ -1,7 +1,7 @@
 import numpy as np
 import tensorflow as tf
 import scipy.signal
-from gym.spaces import Box, Discrete
+from gymnasium.spaces import Box, Discrete
 
 EPS = 1e-8
 

spinup/algos/tf1/vpg/vpg.py

@@ -1,6 +1,6 @@
 import numpy as np
 import tensorflow as tf
-import gym
+import gymnasium
 import time
 import spinup.algos.tf1.vpg.core as core
 from spinup.utils.logx import EpochLogger

spinup/run.py

@@ -3,7 +3,7 @@
 from spinup.utils.run_utils import ExperimentGrid
 from spinup.utils.serialization_utils import convert_json
 import argparse
-import gym
+import gymnasium as gym
 import json
 import os, subprocess, sys
 import os.path as osp
@@ -153,7 +153,7 @@ def process(arg):
 
     # Special handling for environment: make sure that env_name is a real,
     # registered gym environment.
-    valid_envs = [e.id for e in list(gym.envs.registry.all())]
+    valid_envs = [id for e in list(gym.envs.registry.keys())]
     assert 'env_name' in arg_dict, \
         friendly_err("You did not give a value for --env_name! Add one and try again.")
     for env_name in arg_dict['env_name']:
@@ -245,4 +245,4 @@ def process(arg):
         # Assume that the user plans to execute an algorithm. Run custom
         # parsing on the arguments and build a grid search to execute.
         args = sys.argv[2:]
-        parse_and_execute_grid_search(cmd, args)
+        parse_and_execute_grid_search(cmd, args)

spinup/utils/mpi_tf.py

@@ -26,7 +26,7 @@ def sync_all_params():
     return sync_params(tf.global_variables())
 
 
-class MpiAdamOptimizer(tf.train.AdamOptimizer):
+class MpiAdamOptimizer(tf.compat.v1.train.AdamOptimizer):
     """
     Adam optimizer that averages gradients across MPI processes.