[gym_jiminy/common] Train/eval mode API of InterfaceJiminyEnv is now …

…consistent with PyTorch.

python/gym_jiminy/common/gym_jiminy/common/bases/compositions.py

@@ -434,7 +434,7 @@ def __init__(self,
                  grace_period: float = 0.0,
                  *,
                  is_truncation: bool = False,
-                 is_training_only: bool = False) -> None:
+                 training_only: bool = False) -> None:
         """
         :param env: Base or wrapped jiminy environment.
         :param name: Desired name of the termination condition. This name will
@@ -451,16 +451,16 @@ def __init__(self,
                               terminated or truncated whenever the termination
                               condition is triggered.
                               Optional: False by default.
-        :param is_training_only: Whether the termination condition should be
-                                 completely by-passed if the environment is in
-                                 evaluation mode.
-                                 Optional: False by default.
+        :param training_only: Whether the termination condition should be
+                              completely by-passed if the environment is in
+                              evaluation mode.
+                              Optional: False by default.
         """
         self.env = env
         self._name = name
         self.grace_period = grace_period
         self.is_truncation = is_truncation
-        self.is_training_only = is_training_only
+        self.training_only = training_only
 
     @property
     def name(self) -> str:
@@ -505,7 +505,7 @@ def __call__(self, info: InfoType) -> Tuple[bool, bool]:
         """
         # Skip termination condition in eval mode or during grace period
         termination_info: InfoType = {}
-        if (self.is_training_only and not self.env.is_training) or (
+        if (self.training_only and not self.env.training) or (
                 self.env.stepper_state.t < self.grace_period):
             # Always continue
             is_terminated, is_truncated = False, False
@@ -555,7 +555,7 @@ def __init__(self,
                  grace_period: float = 0.0,
                  *,
                  is_truncation: bool = False,
-                 is_training_only: bool = False) -> None:
+                 training_only: bool = False) -> None:
         """
         :param env: Base or wrapped jiminy environment.
         :param name: Desired name of the termination condition. This name will
@@ -578,10 +578,10 @@ def __init__(self,
                               terminated or truncated whenever the termination
                               condition is triggered.
                               Optional: False by default.
-        :param is_training_only: Whether the termination condition should be
-                                 completely by-passed if the environment is in
-                                 evaluation mode.
-                                 Optional: False by default.
+        :param training_only: Whether the termination condition should be
+                              completely by-passed if the environment is in
+                              evaluation mode.
+                              Optional: False by default.
         """
         # Backup user argument(s)
         self.low = np.asarray(low) if isinstance(low, Sequence) else low
@@ -593,7 +593,7 @@ def __init__(self,
             name,
             grace_period,
             is_truncation=is_truncation,
-            is_training_only=is_training_only)
+            training_only=training_only)
 
         # Add quantity to the set of quantities managed by the environment
         self.env.quantities[self.name] = quantity

python/gym_jiminy/common/gym_jiminy/common/bases/interfaces.py

@@ -410,11 +410,13 @@ def has_terminated(self, info: InfoType) -> Tuple[bool, bool]:
         """
 
     @abstractmethod
-    def train(self) -> None:
-        """Sets the environment in training mode.
+    def train(self, mode: bool = True) -> None:
+        """Sets the environment in training or evaluation mode.
+
+        :param mode: Whether to set training (True) or evaluation mode (False).
+                     Optional: `True` by default.
         """
 
-    @abstractmethod
     def eval(self) -> None:
         """Sets the environment in evaluation mode.
 
@@ -423,6 +425,19 @@ def eval(self) -> None:
         time specifically. See documentations of a given environment for
         details about their behaviors in training and evaluation modes.
         """
+        self.train(False)
+
+    @property
+    @abstractmethod
+    def training(self) -> bool:
+        """Check whether the environment is in training or evaluation mode.
+        """
+
+    @training.setter
+    def training(self, mode: bool) -> None:
+        """Sets the environment in training or evaluation mode.
+        """
+        self.train(mode)
 
     @property
     @abstractmethod
@@ -436,9 +451,3 @@ def unwrapped(self) -> "BaseJiminyEnv":
     def step_dt(self) -> float:
         """Get timestep of a single 'step'.
         """
-
-    @property
-    @abstractmethod
-    def is_training(self) -> bool:
-        """Check whether the environment is in 'train' or 'eval' mode.
-        """

python/gym_jiminy/common/gym_jiminy/common/bases/pipeline.py

@@ -222,7 +222,7 @@ def __getattr__(self, name: str) -> Any:
         try:
             # Make sure that no simulaton is running
             if (self.__getattribute__('is_simulation_running') and
-                    self.env.is_training and not hasattr(sys, 'ps1')):
+                    self.env.training and not hasattr(sys, 'ps1')):
                 # `hasattr(sys, 'ps1')` is used to detect whether the method
                 # was called from an interpreter or within a script. For
                 # details, see: https://stackoverflow.com/a/64523765/4820605
@@ -294,15 +294,12 @@ def unwrapped(self) -> "BaseJiminyEnv":
     def step_dt(self) -> float:
         return self.env.step_dt
 
-    @property
-    def is_training(self) -> bool:
-        return self.env.is_training
-
-    def train(self) -> None:
-        self.env.train()
+    @InterfaceJiminyEnv.training.getter  # type: ignore[attr-defined]
+    def training(self) -> bool:
+        return self.env.training
 
-    def eval(self) -> None:
-        self.env.eval()
+    def train(self, mode: bool = True) -> None:
+        self.env.train(mode)
 
     def update_pipeline(self, derived: Optional[InterfaceJiminyEnv]) -> None:
         if derived is None:

python/gym_jiminy/common/gym_jiminy/common/blocks/motor_safety_limit.py

@@ -219,7 +219,7 @@ def _setup(self) -> None:
 
         # Unbounded effort limits in evaluation mode.
         # Note that training/evaluation cannot be changed at this point.
-        self._enable_limit_soft = self.env.is_training
+        self._enable_limit_soft = self.env.training
 
     def compute_command(self,
                         action: np.ndarray,

python/gym_jiminy/common/gym_jiminy/common/blocks/proportional_derivative_controller.py

@@ -488,7 +488,7 @@ def _setup(self) -> None:
 
         # Enable deadband in evaluation mode only.
         # Note that training/evaluation cannot be changed at this point.
-        self._enable_deadband = not self.env.is_training
+        self._enable_deadband = not self.env.training
 
     @property
     def fieldnames(self) -> List[str]:

python/gym_jiminy/common/gym_jiminy/common/compositions/generic.py

@@ -176,7 +176,7 @@ def __init__(self,
                  post_fn: Optional[Callable[
                     [ArrayOrScalar], ArrayOrScalar]] = None,
                  is_truncation: bool = False,
-                 is_training_only: bool = False) -> None:
+                 training_only: bool = False) -> None:
         """
         :param env: Base or wrapped jiminy environment.
         :param name: Desired name of the termination condition. This name will
@@ -214,10 +214,10 @@ def __init__(self,
                               terminated or truncated whenever the termination
                               condition is triggered.
                               Optional: False by default.
-        :param is_training_only: Whether the termination condition should be
-                                 completely by-passed if the environment is in
-                                 evaluation mode.
-                                 Optional: False by default.
+        :param training_only: Whether the termination condition should be
+                              completely by-passed if the environment is in
+                              evaluation mode.
+                              Optional: False by default.
         """
         # pylint: disable=unnecessary-lambda-assignment
 
@@ -256,7 +256,7 @@ def __init__(self,
                          high,
                          grace_period,
                          is_truncation=is_truncation,
-                         is_training_only=is_training_only)
+                         training_only=training_only)
 
     def _compute_drift_error(self,
                              left: np.ndarray,
@@ -298,7 +298,7 @@ def __init__(self,
                  *,
                  op: Callable[[np.ndarray, np.ndarray], np.ndarray] = sub,
                  is_truncation: bool = False,
-                 is_training_only: bool = False) -> None:
+                 training_only: bool = False) -> None:
         """
         :param env: Base or wrapped jiminy environment.
         :param name: Desired name of the termination condition. This name will
@@ -335,10 +335,10 @@ def __init__(self,
                               terminated or truncated whenever the termination
                               condition is triggered.
                               Optional: False by default.
-        :param is_training_only: Whether the termination condition should be
-                                 completely by-passed if the environment is in
-                                 evaluation mode.
-                                 Optional: False by default.
+        :param training_only: Whether the termination condition should be
+                              completely by-passed if the environment is in
+                              evaluation mode.
+                              Optional: False by default.
         """
         # pylint: disable=unnecessary-lambda-assignment
 
@@ -388,7 +388,7 @@ def min_norm(values: np.ndarray) -> float:
                          np.array(thr),
                          max(grace_period, horizon),
                          is_truncation=is_truncation,
-                         is_training_only=is_training_only)
+                         training_only=training_only)
 
     def _compute_min_distance(self,
                               left: np.ndarray,
@@ -483,7 +483,7 @@ def __init__(self,
                  velocity_max: float,
                  grace_period: float = 0.0,
                  *,
-                 is_training_only: bool = False) -> None:
+                 training_only: bool = False) -> None:
         """
         :param env: Base or wrapped jiminy environment.
         :param position_margin: Distance of actuated joints from their
@@ -496,10 +496,10 @@ def __init__(self,
                              of the episode, during which the latter is bound
                              to continue whatever happens.
                              Optional: 0.0 by default.
-        :param is_training_only: Whether the termination condition should be
-                                 completely by-passed if the environment is in
-                                 evaluation mode.
-                                 Optional: False by default.
+        :param training_only: Whether the termination condition should be
+                              completely by-passed if the environment is in
+                              evaluation mode.
+                              Optional: False by default.
         """
         # Backup user argument(s)
         self.position_margin = position_margin
@@ -511,7 +511,7 @@ def __init__(self,
             "termination_mechanical_safety",
             grace_period,
             is_truncation=False,
-            is_training_only=is_training_only)
+            training_only=training_only)
 
         # Add quantity to the set of quantities managed by the environment
         self.env.quantities["_".join((self.name, "position_delta"))] = (
@@ -577,7 +577,7 @@ def __init__(
             generator_mode: EnergyGenerationMode = EnergyGenerationMode.CHARGE,
             grace_period: float = 0.0,
             *,
-            is_training_only: bool = False) -> None:
+            training_only: bool = False) -> None:
         """
         :param env: Base or wrapped jiminy environment.
         :param max_power: Maximum average mechanical power consumption applied
@@ -589,10 +589,10 @@ def __init__(
                              Optional: 0.0 by default.
         :param horizon: Horizon over which values of the quantity will be
                         stacked before computing the average.
-        :param is_training_only: Whether the termination condition should be
-                                 completely by-passed if the environment is in
-                                 evaluation mode.
-                                 Optional: False by default.
+        :param training_only: Whether the termination condition should be
+                              completely by-passed if the environment is in
+                              evaluation mode.
+                              Optional: False by default.
         """
         # Backup user argument(s)
         self.max_power = max_power
@@ -610,7 +610,7 @@ def __init__(
             self.max_power,
             grace_period,
             is_truncation=False,
-            is_training_only=is_training_only)
+            training_only=training_only)
 
 
 class ShiftTrackingMotorPositionsTermination(ShiftTrackingQuantityTermination):
@@ -638,7 +638,7 @@ def __init__(self,
                  horizon: float,
                  grace_period: float = 0.0,
                  *,
-                 is_training_only: bool = False) -> None:
+                 training_only: bool = False) -> None:
         """
         :param env: Base or wrapped jiminy environment.
         :param thr: Maximum shift above which termination is triggered.
@@ -648,10 +648,10 @@ def __init__(self,
                              of the episode, during which the latter is bound
                              to continue whatever happens.
                              Optional: 0.0 by default.
-        :param is_training_only: Whether the termination condition should be
-                                 completely by-passed if the environment is in
-                                 evaluation mode.
-                                 Optional: False by default.
+        :param training_only: Whether the termination condition should be
+                              completely by-passed if the environment is in
+                              evaluation mode.
+                              Optional: False by default.
         """
         # Call base implementation
         super().__init__(
@@ -665,4 +665,4 @@ def __init__(self,
             horizon,
             grace_period,
             is_truncation=False,
-            is_training_only=is_training_only)
+            training_only=training_only)