I've been using the get_value function from df_cache.py in the main branch and noticed that self._trans_mean isn't initialized prior to its call. Upon investigating further by switching to tag v1.1.0, I discovered that this function was later removed in subsequent updates.
def transform_data(self, **kwargs) -> None:
if "shift_mean_to" in kwargs:
This standardizes the scene to be relative to the agent being predicted.
self._transf_mean = kwargs["shift_mean_to"]
if "rotate_by" in kwargs:
# This rotates the scene so that the predicted agent's current
# heading aligns with the x-axis.
agent_heading: float = kwargs["rotate_by"]
self._transf_rotmat: np.ndarray = np.array(
[
[np.cos(agent_heading), -np.sin(agent_heading)],
[np.sin(agent_heading), np.cos(agent_heading)],
]
)
I've been using the get_value function from df_cache.py in the main branch and noticed that self._trans_mean isn't initialized prior to its call. Upon investigating further by switching to tag v1.1.0, I discovered that this function was later removed in subsequent updates.
def transform_data(self, **kwargs) -> None: if "shift_mean_to" in kwargs:
This standardizes the scene to be relative to the agent being predicted.