stable-baselines3 2.2.1__tar.gz → 2.3.0__tar.gz

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: stable_baselines3
-Version: 2.2.1
+Version: 2.3.0
 Summary: Pytorch version of Stable Baselines, implementations of reinforcement learning algorithms.
 Home-page: https://github.com/DLR-RM/stable-baselines3
 Author: Antonin Raffin
@@ -34,8 +34,8 @@ Requires-Dist: pytest-cov; extra == "tests"
 Requires-Dist: pytest-env; extra == "tests"
 Requires-Dist: pytest-xdist; extra == "tests"
 Requires-Dist: mypy; extra == "tests"
-Requires-Dist: ruff>=0.0.288; extra == "tests"
-Requires-Dist: black<24,>=23.9.1; extra == "tests"
+Requires-Dist: ruff>=0.3.1; extra == "tests"
+Requires-Dist: black<25,>=24.2.0; extra == "tests"
 Provides-Extra: docs
 Requires-Dist: sphinx<8,>=5; extra == "docs"
 Requires-Dist: sphinx-autobuild; extra == "docs"
@@ -99,7 +99,7 @@ import gymnasium
 
 from stable_baselines3 import PPO
 
-env = gymnasium.make("CartPole-v1")
+env = gymnasium.make("CartPole-v1", render_mode="human")
 
 model = PPO("MlpPolicy", env, verbose=1)
 model.learn(total_timesteps=10_000)

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/README.md

@@ -127,7 +127,7 @@ import gymnasium as gym
 
 from stable_baselines3 import PPO
 
-env = gym.make("CartPole-v1")
+env = gym.make("CartPole-v1", render_mode="human")
 
 model = PPO("MlpPolicy", env, verbose=1)
 model.learn(total_timesteps=10_000)

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/pyproject.toml

@@ -3,13 +3,15 @@
 line-length = 127
 # Assume Python 3.8
 target-version = "py38"
+
+[tool.ruff.lint]
 # See https://beta.ruff.rs/docs/rules/
 select = ["E", "F", "B", "UP", "C90", "RUF"]
 # B028: Ignore explicit stacklevel`
 # RUF013: Too many false positives (implicit optional)
 ignore = ["B028", "RUF013"]
 
-[tool.ruff.per-file-ignores]
+[tool.ruff.lint.per-file-ignores]
 # Default implementation in abstract methods
 "./stable_baselines3/common/callbacks.py"= ["B027"]
 "./stable_baselines3/common/noise.py"= ["B027"]
@@ -17,7 +19,7 @@ ignore = ["B028", "RUF013"]
 "./tests/*.py"= ["RUF012", "RUF013"]
 
 
-[tool.ruff.mccabe]
+[tool.ruff.lint.mccabe]
 # Unlike Flake8, default to a complexity level of 10.
 max-complexity = 15
 

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/setup.py

@@ -43,7 +43,7 @@ import gymnasium
 
 from stable_baselines3 import PPO
 
-env = gymnasium.make("CartPole-v1")
+env = gymnasium.make("CartPole-v1", render_mode="human")
 
 model = PPO("MlpPolicy", env, verbose=1)
 model.learn(total_timesteps=10_000)
@@ -120,9 +120,9 @@ setup(
             # Type check
             "mypy",
             # Lint code and sort imports (flake8 and isort replacement)
-            "ruff>=0.0.288",
+            "ruff>=0.3.1",
             # Reformat
-            "black>=23.9.1,<24",
+            "black>=24.2.0,<25",
         ],
         "docs": [
             "sphinx>=5,<8",

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/common/base_class.py

@@ -523,7 +523,10 @@ class BaseAlgorithm(ABC):
 
         :param total_timesteps: The total number of samples (env steps) to train on
         :param callback: callback(s) called at every step with state of the algorithm.
-        :param log_interval: The number of episodes before logging.
+        :param log_interval: for on-policy algos (e.g., PPO, A2C, ...) this is the number of
+            training iterations (i.e., log_interval * n_steps * n_envs timesteps) before logging;
+            for off-policy algos (e.g., TD3, SAC, ...) this is the number of episodes before
+            logging.
         :param tb_log_name: the name of the run for TensorBoard logging
         :param reset_num_timesteps: whether or not to reset the current timestep number (used in logging)
         :param progress_bar: Display a progress bar using tqdm and rich.

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/common/distributions.py

@@ -113,7 +113,7 @@ def sum_independent_dims(tensor: th.Tensor) -> th.Tensor:
     so we can sum components of the ``log_prob`` or the entropy.
 
     :param tensor: shape: (n_batch, n_actions) or (n_batch,)
-    :return: shape: (n_batch,)
+    :return: shape: (n_batch,) for (n_batch, n_actions) input, scalar for (n_batch,) input
     """
     if len(tensor.shape) > 1:
         tensor = tensor.sum(dim=1)

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/common/env_checker.py

@@ -17,13 +17,37 @@ def _is_numpy_array_space(space: spaces.Space) -> bool:
     return not isinstance(space, (spaces.Dict, spaces.Tuple))
 
 
+def _starts_at_zero(space: Union[spaces.Discrete, spaces.MultiDiscrete]) -> bool:
+    """
+    Return False if a (Multi)Discrete space has a non-zero start.
+    """
+    return np.allclose(space.start, np.zeros_like(space.start))
+
+
+def _check_non_zero_start(space: spaces.Space, space_type: str = "observation", key: str = "") -> None:
+    """
+    :param space: Observation or action space
+    :param space_type: information about whether it is an observation or action space
+        (for the warning message)
+    :param key: When the observation space comes from a Dict space, we pass the
+        corresponding key to have more precise warning messages. Defaults to "".
+    """
+    if isinstance(space, (spaces.Discrete, spaces.MultiDiscrete)) and not _starts_at_zero(space):
+        maybe_key = f"(key='{key}')" if key else ""
+        warnings.warn(
+            f"{type(space).__name__} {space_type} space {maybe_key} with a non-zero start (start={space.start}) "
+            "is not supported by Stable-Baselines3. "
+            f"You can use a wrapper or update your {space_type} space."
+        )
+
+
 def _check_image_input(observation_space: spaces.Box, key: str = "") -> None:
     """
     Check that the input will be compatible with Stable-Baselines
     when the observation is apparently an image.
 
     :param observation_space: Observation space
-    :key: When the observation space comes from a Dict space, we pass the
+    :param key: When the observation space comes from a Dict space, we pass the
         corresponding key to have more precise warning messages. Defaults to "".
     """
     if observation_space.dtype != np.uint8:
@@ -63,11 +87,7 @@ def _check_unsupported_spaces(env: gym.Env, observation_space: spaces.Space, act
         for key, space in observation_space.spaces.items():
             if isinstance(space, spaces.Dict):
                 nested_dict = True
-            if isinstance(space, spaces.Discrete) and space.start != 0:
-                warnings.warn(
-                    f"Discrete observation space (key '{key}') with a non-zero start is not supported by Stable-Baselines3. "
-                    "You can use a wrapper or update your observation space."
-                )
+            _check_non_zero_start(space, "observation", key)
 
         if nested_dict:
             warnings.warn(
@@ -87,11 +107,7 @@ def _check_unsupported_spaces(env: gym.Env, observation_space: spaces.Space, act
             "which is supported by SB3."
         )
 
-    if isinstance(observation_space, spaces.Discrete) and observation_space.start != 0:
-        warnings.warn(
-            "Discrete observation space with a non-zero start is not supported by Stable-Baselines3. "
-            "You can use a wrapper or update your observation space."
-        )
+    _check_non_zero_start(observation_space, "observation")
 
     if isinstance(observation_space, spaces.Sequence):
         warnings.warn(
@@ -100,11 +116,7 @@ def _check_unsupported_spaces(env: gym.Env, observation_space: spaces.Space, act
             "Note: The checks for returned values are skipped."
         )
 
-    if isinstance(action_space, spaces.Discrete) and action_space.start != 0:
-        warnings.warn(
-            "Discrete action space with a non-zero start is not supported by Stable-Baselines3. "
-            "You can use a wrapper or update your action space."
-        )
+    _check_non_zero_start(action_space, "action")
 
     if not _is_numpy_array_space(action_space):
         warnings.warn(

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/common/on_policy_algorithm.py

@@ -92,6 +92,7 @@ class OnPolicyAlgorithm(BaseAlgorithm):
             use_sde=use_sde,
             sde_sample_freq=sde_sample_freq,
             support_multi_env=True,
+            monitor_wrapper=monitor_wrapper,
             seed=seed,
             stats_window_size=stats_window_size,
             tensorboard_log=tensorboard_log,
@@ -200,7 +201,7 @@ class OnPolicyAlgorithm(BaseAlgorithm):
             if not callback.on_step():
                 return False
 
-            self._update_info_buffer(infos)
+            self._update_info_buffer(infos, dones)
             n_steps += 1
 
             if isinstance(self.action_space, spaces.Discrete):
@@ -250,6 +251,28 @@ class OnPolicyAlgorithm(BaseAlgorithm):
         """
         raise NotImplementedError
 
+    def _dump_logs(self, iteration: int) -> None:
+        """
+        Write log.
+
+        :param iteration: Current logging iteration
+        """
+        assert self.ep_info_buffer is not None
+        assert self.ep_success_buffer is not None
+
+        time_elapsed = max((time.time_ns() - self.start_time) / 1e9, sys.float_info.epsilon)
+        fps = int((self.num_timesteps - self._num_timesteps_at_start) / time_elapsed)
+        self.logger.record("time/iterations", iteration, exclude="tensorboard")
+        if len(self.ep_info_buffer) > 0 and len(self.ep_info_buffer[0]) > 0:
+            self.logger.record("rollout/ep_rew_mean", safe_mean([ep_info["r"] for ep_info in self.ep_info_buffer]))
+            self.logger.record("rollout/ep_len_mean", safe_mean([ep_info["l"] for ep_info in self.ep_info_buffer]))
+        self.logger.record("time/fps", fps)
+        self.logger.record("time/time_elapsed", int(time_elapsed), exclude="tensorboard")
+        self.logger.record("time/total_timesteps", self.num_timesteps, exclude="tensorboard")
+        if len(self.ep_success_buffer) > 0:
+            self.logger.record("rollout/success_rate", safe_mean(self.ep_success_buffer))
+        self.logger.dump(step=self.num_timesteps)
+
     def learn(
         self: SelfOnPolicyAlgorithm,
         total_timesteps: int,
@@ -285,16 +308,7 @@ class OnPolicyAlgorithm(BaseAlgorithm):
             # Display training infos
             if log_interval is not None and iteration % log_interval == 0:
                 assert self.ep_info_buffer is not None
-                time_elapsed = max((time.time_ns() - self.start_time) / 1e9, sys.float_info.epsilon)
-                fps = int((self.num_timesteps - self._num_timesteps_at_start) / time_elapsed)
-                self.logger.record("time/iterations", iteration, exclude="tensorboard")
-                if len(self.ep_info_buffer) > 0 and len(self.ep_info_buffer[0]) > 0:
-                    self.logger.record("rollout/ep_rew_mean", safe_mean([ep_info["r"] for ep_info in self.ep_info_buffer]))
-                    self.logger.record("rollout/ep_len_mean", safe_mean([ep_info["l"] for ep_info in self.ep_info_buffer]))
-                self.logger.record("time/fps", fps)
-                self.logger.record("time/time_elapsed", int(time_elapsed), exclude="tensorboard")
-                self.logger.record("time/total_timesteps", self.num_timesteps, exclude="tensorboard")
-                self.logger.dump(step=self.num_timesteps)
+                self._dump_logs(iteration)
 
             self.train()
 

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/common/policies.py

@@ -173,7 +173,9 @@ class BaseModel(nn.Module):
         :return:
         """
         device = get_device(device)
-        saved_variables = th.load(path, map_location=device)
+        # Note(antonin): we cannot use `weights_only=True` here because we need to allow
+        # gymnasium imports for the policy to be loaded successfully
+        saved_variables = th.load(path, map_location=device, weights_only=False)
 
         # Create policy object
         model = cls(**saved_variables["data"])

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/common/save_util.py

@@ -2,6 +2,7 @@
 Save util taken from stable_baselines
 used to serialize data (class parameters) of model classes
 """
+
 import base64
 import functools
 import io
@@ -446,7 +447,7 @@ def load_from_zip_file(
                     file_content.seek(0)
                     # Load the parameters with the right ``map_location``.
                     # Remove ".pth" ending with splitext
-                    th_object = th.load(file_content, map_location=device)
+                    th_object = th.load(file_content, map_location=device, weights_only=True)
                     # "tensors.pth" was renamed "pytorch_variables.pth" in v0.9.0, see PR #138
                     if file_path == "pytorch_variables.pth" or file_path == "tensors.pth":
                         # PyTorch variables (not state_dicts)

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/common/type_aliases.py

@@ -1,4 +1,5 @@
 """Common aliases for type hints"""
+
 from enum import Enum
 from typing import TYPE_CHECKING, Any, Callable, Dict, List, NamedTuple, Optional, Protocol, SupportsFloat, Tuple, Union
 

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/common/vec_env/util.py

@@ -1,6 +1,7 @@
 """
 Helpers for dealing with vectorized environments.
 """
+
 from collections import OrderedDict
 from typing import Any, Dict, List, Tuple
 

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/common/vec_env/vec_frame_stack.py

@@ -29,7 +29,12 @@ class VecFrameStack(VecEnvWrapper):
 
     def step_wait(
         self,
-    ) -> Tuple[Union[np.ndarray, Dict[str, np.ndarray]], np.ndarray, np.ndarray, List[Dict[str, Any]],]:
+    ) -> Tuple[
+        Union[np.ndarray, Dict[str, np.ndarray]],
+        np.ndarray,
+        np.ndarray,
+        List[Dict[str, Any]],
+    ]:
         observations, rewards, dones, infos = self.venv.step_wait()
         observations, infos = self.stacked_obs.update(observations, dones, infos)  # type: ignore[arg-type]
         return observations, rewards, dones, infos

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/ddpg/ddpg.py

@@ -60,11 +60,11 @@ class DDPG(TD3):
         learning_rate: Union[float, Schedule] = 1e-3,
         buffer_size: int = 1_000_000,  # 1e6
         learning_starts: int = 100,
-        batch_size: int = 100,
+        batch_size: int = 256,
         tau: float = 0.005,
         gamma: float = 0.99,
-        train_freq: Union[int, Tuple[int, str]] = (1, "episode"),
-        gradient_steps: int = -1,
+        train_freq: Union[int, Tuple[int, str]] = 1,
+        gradient_steps: int = 1,
         action_noise: Optional[ActionNoise] = None,
         replay_buffer_class: Optional[Type[ReplayBuffer]] = None,
         replay_buffer_kwargs: Optional[Dict[str, Any]] = None,

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/dqn/dqn.py

@@ -79,7 +79,7 @@ class DQN(OffPolicyAlgorithm):
         env: Union[GymEnv, str],
         learning_rate: Union[float, Schedule] = 1e-4,
         buffer_size: int = 1_000_000,  # 1e6
-        learning_starts: int = 50000,
+        learning_starts: int = 100,
         batch_size: int = 32,
         tau: float = 1.0,
         gamma: float = 0.99,

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3/td3/td3.py

@@ -83,11 +83,11 @@ class TD3(OffPolicyAlgorithm):
         learning_rate: Union[float, Schedule] = 1e-3,
         buffer_size: int = 1_000_000,  # 1e6
         learning_starts: int = 100,
-        batch_size: int = 100,
+        batch_size: int = 256,
         tau: float = 0.005,
         gamma: float = 0.99,
-        train_freq: Union[int, Tuple[int, str]] = (1, "episode"),
-        gradient_steps: int = -1,
+        train_freq: Union[int, Tuple[int, str]] = 1,
+        gradient_steps: int = 1,
         action_noise: Optional[ActionNoise] = None,
         replay_buffer_class: Optional[Type[ReplayBuffer]] = None,
         replay_buffer_kwargs: Optional[Dict[str, Any]] = None,

stable_baselines3-2.3.0/stable_baselines3/version.txt

@@ -0,0 +1 @@
+2.3.0

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
-Name: stable-baselines3
-Version: 2.2.1
+Name: stable_baselines3
+Version: 2.3.0
 Summary: Pytorch version of Stable Baselines, implementations of reinforcement learning algorithms.
 Home-page: https://github.com/DLR-RM/stable-baselines3
 Author: Antonin Raffin
@@ -34,8 +34,8 @@ Requires-Dist: pytest-cov; extra == "tests"
 Requires-Dist: pytest-env; extra == "tests"
 Requires-Dist: pytest-xdist; extra == "tests"
 Requires-Dist: mypy; extra == "tests"
-Requires-Dist: ruff>=0.0.288; extra == "tests"
-Requires-Dist: black<24,>=23.9.1; extra == "tests"
+Requires-Dist: ruff>=0.3.1; extra == "tests"
+Requires-Dist: black<25,>=24.2.0; extra == "tests"
 Provides-Extra: docs
 Requires-Dist: sphinx<8,>=5; extra == "docs"
 Requires-Dist: sphinx-autobuild; extra == "docs"
@@ -99,7 +99,7 @@ import gymnasium
 
 from stable_baselines3 import PPO
 
-env = gymnasium.make("CartPole-v1")
+env = gymnasium.make("CartPole-v1", render_mode="human")
 
 model = PPO("MlpPolicy", env, verbose=1)
 model.learn(total_timesteps=10_000)

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/stable_baselines3.egg-info/requires.txt

@@ -39,5 +39,5 @@ pytest-cov
 pytest-env
 pytest-xdist
 mypy
-ruff>=0.0.288
-black<24,>=23.9.1
+ruff>=0.3.1
+black<25,>=24.2.0

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/tests/test_envs.py

@@ -123,6 +123,8 @@ def test_high_dimension_action_space():
         spaces.Dict({"img": spaces.Box(low=0, high=255, shape=(32, 32, 3), dtype=np.uint8)}),
         # Non zero start index
         spaces.Discrete(3, start=-1),
+        # Non zero start index (MultiDiscrete)
+        spaces.MultiDiscrete([4, 4], start=[1, 0]),
         # Non zero start index inside a Dict
         spaces.Dict({"obs": spaces.Discrete(3, start=1)}),
     ],
@@ -164,6 +166,8 @@ def test_non_default_spaces(new_obs_space):
         spaces.Box(low=np.array([-1, -1, -1]), high=np.array([1, 1, 0.99]), dtype=np.float32),
         # Non zero start index
         spaces.Discrete(3, start=-1),
+        # Non zero start index (MultiDiscrete)
+        spaces.MultiDiscrete([4, 4], start=[1, 0]),
     ],
 )
 def test_non_default_action_spaces(new_action_space):
@@ -179,7 +183,7 @@ def test_non_default_action_spaces(new_action_space):
     env.action_space = new_action_space
 
     # Discrete action space
-    if isinstance(new_action_space, spaces.Discrete):
+    if isinstance(new_action_space, (spaces.Discrete, spaces.MultiDiscrete)):
         with pytest.warns(UserWarning):
             check_env(env)
         return

{stable_baselines3-2.2.1 → stable_baselines3-2.3.0}/tests/test_logger.py

@@ -14,7 +14,7 @@ from gymnasium import spaces
 from matplotlib import pyplot as plt
 from pandas.errors import EmptyDataError
 
-from stable_baselines3 import A2C, DQN
+from stable_baselines3 import A2C, DQN, PPO
 from stable_baselines3.common.env_checker import check_env
 from stable_baselines3.common.logger import (
     DEBUG,
@@ -33,6 +33,7 @@ from stable_baselines3.common.logger import (
     read_csv,
     read_json,
 )
+from stable_baselines3.common.monitor import Monitor
 
 KEY_VALUES = {
     "test": 1,
@@ -474,3 +475,92 @@ def test_human_output_format_custom_test_io(base_class):
 """
 
     assert printed == desired_printed
+
+
+class DummySuccessEnv(gym.Env):
+    """
+    Create a dummy success environment that returns wether True or False for info['is_success']
+    at the end of an episode according to its dummy successes list
+    """
+
+    def __init__(self, dummy_successes, ep_steps):
+        """Init the dummy success env
+
+        :param dummy_successes: list of size (n_logs_iterations, n_episodes_per_log) that specifies
+            the success value of log iteration i at episode j
+        :param ep_steps: number of steps per episode (to activate truncated)
+        """
+        self.n_steps = 0
+        self.log_id = 0
+        self.ep_id = 0
+
+        self.ep_steps = ep_steps
+
+        self.dummy_success = dummy_successes
+        self.num_logs = len(dummy_successes)
+        self.ep_per_log = len(dummy_successes[0])
+        self.steps_per_log = self.ep_per_log * self.ep_steps
+
+        self.action_space = spaces.Discrete(2)
+        self.observation_space = spaces.Discrete(2)
+
+    def reset(self, seed=None, options=None):
+        """
+        Reset the env and advance to the next episode_id to get the next dummy success
+        """
+        self.n_steps = 0
+
+        if self.ep_id == self.ep_per_log:
+            self.ep_id = 0
+            self.log_id = (self.log_id + 1) % self.num_logs
+
+        return self.observation_space.sample(), {}
+
+    def step(self, action):
+        """
+        Step and return a dummy success when an episode is truncated
+        """
+        self.n_steps += 1
+        truncated = self.n_steps >= self.ep_steps
+
+        info = {}
+        if truncated:
+            maybe_success = self.dummy_success[self.log_id][self.ep_id]
+            info["is_success"] = maybe_success
+            self.ep_id += 1
+        return self.observation_space.sample(), 0.0, False, truncated, info
+
+
+def test_rollout_success_rate_on_policy_algorithm(tmp_path):
+    """
+    Test if the rollout/success_rate information is correctly logged with on policy algorithms
+
+    To do so, create a dummy environment that takes as argument dummy successes (i.e when an episode)
+    is going to be successfull or not.
+    """
+
+    STATS_WINDOW_SIZE = 10
+    # Add dummy successes with 0.3, 0.5 and 0.8 success_rate of length STATS_WINDOW_SIZE
+    dummy_successes = [
+        [True] * 3 + [False] * 7,
+        [True] * 5 + [False] * 5,
+        [True] * 8 + [False] * 2,
+    ]
+    ep_steps = 64
+
+    # Monitor the env to track the success info
+    monitor_file = str(tmp_path / "monitor.csv")
+    env = Monitor(DummySuccessEnv(dummy_successes, ep_steps), filename=monitor_file, info_keywords=("is_success",))
+
+    # Equip the model of a custom logger to check the success_rate info
+    model = PPO("MlpPolicy", env=env, stats_window_size=STATS_WINDOW_SIZE, n_steps=env.steps_per_log, verbose=1)
+    logger = InMemoryLogger()
+    model.set_logger(logger)
+
+    # Make the model learn and check that the success rate corresponds to the ratio of dummy successes
+    model.learn(total_timesteps=env.ep_per_log * ep_steps, log_interval=1)
+    assert logger.name_to_value["rollout/success_rate"] == 0.3
+    model.learn(total_timesteps=env.ep_per_log * ep_steps, log_interval=1)
+    assert logger.name_to_value["rollout/success_rate"] == 0.5
+    model.learn(total_timesteps=env.ep_per_log * ep_steps, log_interval=1)
+    assert logger.name_to_value["rollout/success_rate"] == 0.8

stable_baselines3-2.2.1/stable_baselines3/version.txt

@@ -1 +0,0 @@
-2.2.1