Skip to content

TrajectoryMarker ​

python
TrajectoryMarker(id: int, marker_type: TrajectoryMarkerKind, trigger_index: int | None, episode_result: int, truncated: bool, values: tuple[int, ...], snapshot: bytes, frame_stack: bytes | None = None) -> None

Python view of one trajectory marker block.

Source

Values ​

episode_result ​

python
episode_result: int

Source

frame_stack ​

python
frame_stack: bytes | None = None

The observation frame stack captured at this marker, when the recording was configured with frame_stacks: true. Restoring a snapshot without it leaves the policy blind to motion for the first frames - 1 decisions, so a restart source should always carry one.

Source

id ​

python
id: int

Source

marker_type ​

python
marker_type: TrajectoryMarkerKind

Source

snapshot ​

python
snapshot: bytes

Source

trigger_index ​

python
trigger_index: int | None

Source

truncated ​

python
truncated: bool

Source

values ​

python
values: tuple[int, ...]

Source