Release v2.0.0 (#1571)

* RUF012: Explicit ClassVar

* Prepare v2.0.0

* Update docs/misc/changelog.rst

---------

Co-authored-by: Quentin Gallouédec <[email protected]>

docs/conf.py

@@ -13,8 +13,7 @@
 #
 import os
 import sys
-from typing import Dict, List
-from unittest.mock import MagicMock
+from typing import Dict
 
 # We CANNOT enable 'sphinxcontrib.spelling' because ReadTheDocs.org does not support
 # PyEnchant.
@@ -36,22 +35,6 @@
 # source code directory, relative to this file, for sphinx-autobuild
 sys.path.insert(0, os.path.abspath(".."))
 
-
-class Mock(MagicMock):
-    __subclasses__ = []  # type: ignore
-
-    @classmethod
-    def __getattr__(cls, name):
-        return MagicMock()
-
-
-# Mock modules that requires C modules
-# Note: because of that we cannot test examples using CI
-# 'torch', 'torch.nn', 'torch.nn.functional',
-# DO not mock modules for now, we will need to do that for read the docs later
-MOCK_MODULES: List[str] = []
-sys.modules.update((mod_name, Mock()) for mod_name in MOCK_MODULES)
-
 # Read version from file
 version_file = os.path.join(os.path.dirname(__file__), "../stable_baselines3", "version.txt")
 with open(version_file) as file_handler:

docs/misc/changelog.rst

@@ -3,7 +3,7 @@
 Changelog
 ==========
 
-Release 2.0.0a14 (WIP)
+Release 2.0.0 (2023-06-22)
 --------------------------
 
 **Gymnasium support**
@@ -26,13 +26,20 @@ Breaking Changes:
 
 New Features:
 ^^^^^^^^^^^^^
-
+- Added Gymnasium support (Gym 0.21 and 0.26 are supported via the ``shimmy`` package)
 
 `SB3-Contrib`_
 ^^^^^^^^^^^^^^
+- Fixed QRDQN update interval for multi envs
+
 
 `RL Zoo`_
 ^^^^^^^^^
+- Gym 0.26+ patches to continue working with pybullet and TimeLimit wrapper
+- Renamed `CarRacing-v1` to `CarRacing-v2` in hyperparameters
+- Huggingface push to hub now accepts a `--n-timesteps` argument to adjust the length of the video
+- Fixed `record_video` steps (before it was stepping in a closed env)
+- Dropped Gym 0.21 support
 
 Bug Fixes:
 ^^^^^^^^^^

pyproject.toml

@@ -5,13 +5,16 @@ line-length = 127
 target-version = "py37"
 # See https://beta.ruff.rs/docs/rules/
 select = ["E", "F", "B", "UP", "C90", "RUF"]
-# Ignore explicit stacklevel`
-ignore = ["B028"]
+# B028: Ignore explicit stacklevel`
+# RUF013: Too many false positives (implicit optional)
+ignore = ["B028", "RUF013"]
 
 [tool.ruff.per-file-ignores]
 # Default implementation in abstract methods
 "./stable_baselines3/common/callbacks.py"= ["B027"]
 "./stable_baselines3/common/noise.py"= ["B027"]
+# ClassVar, implicit optional check not needed for tests
+"./tests/*.py"= ["RUF012", "RUF013"]
 
 
 [tool.ruff.mccabe]

stable_baselines3/a2c/a2c.py

@@ -1,4 +1,4 @@
-from typing import Any, Dict, Optional, Type, TypeVar, Union
+from typing import Any, ClassVar, Dict, Optional, Type, TypeVar, Union
 
 import torch as th
 from gymnasium import spaces
@@ -54,7 +54,7 @@ class A2C(OnPolicyAlgorithm):
     :param _init_setup_model: Whether or not to build the network at the creation of the instance
     """
 
-    policy_aliases: Dict[str, Type[BasePolicy]] = {
+    policy_aliases: ClassVar[Dict[str, Type[BasePolicy]]] = {
         "MlpPolicy": ActorCriticPolicy,
         "CnnPolicy": ActorCriticCnnPolicy,
         "MultiInputPolicy": MultiInputActorCriticPolicy,

stable_baselines3/common/base_class.py

@@ -6,7 +6,7 @@
 import warnings
 from abc import ABC, abstractmethod
 from collections import deque
-from typing import Any, Dict, Iterable, List, Optional, Tuple, Type, TypeVar, Union
+from typing import Any, ClassVar, Dict, Iterable, List, Optional, Tuple, Type, TypeVar, Union
 
 import gymnasium as gym
 import numpy as np
@@ -94,7 +94,7 @@ class BaseAlgorithm(ABC):
     """
 
     # Policy aliases (see _get_policy_from_name())
-    policy_aliases: Dict[str, Type[BasePolicy]] = {}
+    policy_aliases: ClassVar[Dict[str, Type[BasePolicy]]] = {}
     policy: BasePolicy
     observation_space: spaces.Space
     action_space: spaces.Space

stable_baselines3/common/vec_env/base_vec_env.py

@@ -54,8 +54,6 @@ class VecEnv(ABC):
     :param action_space: Action space
     """
 
-    metadata = {"render_modes": ["human", "rgb_array"]}
-
     def __init__(
         self,
         num_envs: int,
@@ -69,6 +67,7 @@ def __init__(
         self.reset_infos: List[Dict[str, Any]] = [{} for _ in range(num_envs)]
         # seeds to be used in the next call to env.reset()
         self._seeds: List[Optional[int]] = [None for _ in range(num_envs)]
+
         try:
             render_modes = self.get_attr("render_mode")
         except AttributeError:
@@ -80,6 +79,16 @@ def __init__(
         ), "render_mode mode should be the same for all environments"
         self.render_mode = render_modes[0]
 
+        render_modes = []
+        if self.render_mode is not None:
+            if self.render_mode == "rgb_array":
+                # SB3 uses OpenCV for the "human" mode
+                render_modes = ["human", "rgb_array"]
+            else:
+                render_modes = [self.render_mode]
+
+        self.metadata = {"render_modes": render_modes}
+
     def _reset_seeds(self) -> None:
         """
         Reset the seeds that are going to be used at the next reset.

stable_baselines3/dqn/dqn.py

@@ -1,5 +1,5 @@
 import warnings
-from typing import Any, Dict, List, Optional, Tuple, Type, TypeVar, Union
+from typing import Any, ClassVar, Dict, List, Optional, Tuple, Type, TypeVar, Union
 
 import numpy as np
 import torch as th
@@ -62,7 +62,7 @@ class DQN(OffPolicyAlgorithm):
     :param _init_setup_model: Whether or not to build the network at the creation of the instance
     """
 
-    policy_aliases: Dict[str, Type[BasePolicy]] = {
+    policy_aliases: ClassVar[Dict[str, Type[BasePolicy]]] = {
         "MlpPolicy": MlpPolicy,
         "CnnPolicy": CnnPolicy,
         "MultiInputPolicy": MultiInputPolicy,

stable_baselines3/ppo/ppo.py

@@ -1,5 +1,5 @@
 import warnings
-from typing import Any, Dict, Optional, Type, TypeVar, Union
+from typing import Any, ClassVar, Dict, Optional, Type, TypeVar, Union
 
 import numpy as np
 import torch as th
@@ -68,7 +68,7 @@ class PPO(OnPolicyAlgorithm):
     :param _init_setup_model: Whether or not to build the network at the creation of the instance
     """
 
-    policy_aliases: Dict[str, Type[BasePolicy]] = {
+    policy_aliases: ClassVar[Dict[str, Type[BasePolicy]]] = {
         "MlpPolicy": ActorCriticPolicy,
         "CnnPolicy": ActorCriticCnnPolicy,
         "MultiInputPolicy": MultiInputActorCriticPolicy,

stable_baselines3/sac/sac.py

@@ -1,4 +1,4 @@
-from typing import Any, Dict, List, Optional, Tuple, Type, TypeVar, Union
+from typing import Any, ClassVar, Dict, List, Optional, Tuple, Type, TypeVar, Union
 
 import numpy as np
 import torch as th
@@ -77,7 +77,7 @@ class SAC(OffPolicyAlgorithm):
     :param _init_setup_model: Whether or not to build the network at the creation of the instance
     """
 
-    policy_aliases: Dict[str, Type[BasePolicy]] = {
+    policy_aliases: ClassVar[Dict[str, Type[BasePolicy]]] = {
         "MlpPolicy": MlpPolicy,
         "CnnPolicy": CnnPolicy,
         "MultiInputPolicy": MultiInputPolicy,

stable_baselines3/td3/td3.py

@@ -1,4 +1,4 @@
-from typing import Any, Dict, List, Optional, Tuple, Type, TypeVar, Union
+from typing import Any, ClassVar, Dict, List, Optional, Tuple, Type, TypeVar, Union
 
 import numpy as np
 import torch as th
@@ -65,7 +65,7 @@ class TD3(OffPolicyAlgorithm):
     :param _init_setup_model: Whether or not to build the network at the creation of the instance
     """
 
-    policy_aliases: Dict[str, Type[BasePolicy]] = {
+    policy_aliases: ClassVar[Dict[str, Type[BasePolicy]]] = {
         "MlpPolicy": MlpPolicy,
         "CnnPolicy": CnnPolicy,
         "MultiInputPolicy": MultiInputPolicy,

stable_baselines3/version.txt

@@ -1 +1 @@
-2.0.0a14
+2.0.0