Language

unilab.tasks.motion_tracking.common.tensor_state_store.DeviceStateStore¶

class unilab.tasks.motion_tracking.common.tensor_state_store.DeviceStateStore[source]¶

Bases: object

Read qpos/qvel, scalar sensors, and tracked body state on one device.

Parameters:

backend (SimBackend)
device (device)
num_envs (int)
joint_qpos_ids (Tensor)
joint_qvel_ids (Tensor)
body_names (tuple[str, ...])
body_ids (Tensor)

Methods

`__init__`(*, backend, device, num_envs, ...)
`apply_reset`(reset_mask, qpos, qvel)	Apply a fixed-width reset through the public mask boundary.
`qviews`()	Return the current backend qpos/qvel views after a full read.
`read`([rows])
`refresh_named_sensors`()	Cross the backend's named-sensor projection boundary.
`step_tensor`(ctrl, nsteps)	Step through the negotiated packed plan when one is available.
`validate_finite`()

__init__(*, backend, device, num_envs, joint_qpos_ids, joint_qvel_ids, body_names, body_ids)[source]¶

Parameters:

backend (SimBackend)
device (device)
num_envs (int)
joint_qpos_ids (Tensor)
joint_qvel_ids (Tensor)
body_names (tuple[str, ...])
body_ids (Tensor)

qviews()[source]¶

Return the current backend qpos/qvel views after a full read.

Return type:: tuple[Tensor, Tensor]

read(rows=None)[source]¶

Parameters:: rows (Tensor | None)
Return type:: None

step_tensor(ctrl, nsteps)[source]¶

Step through the negotiated packed plan when one is available.

Parameters:

ctrl (Tensor)
nsteps (int)

Return type:

dict | None

refresh_named_sensors()[source]¶

Cross the backend’s named-sensor projection boundary.

DEVICE_RESIDENT adapters may recompute derived projections when a named sensor view is requested. Unlike read(), this method performs no row validation, state-view negotiation, or task-buffer copy.

Return type:: None

apply_reset(reset_mask, qpos, qvel)[source]¶

Apply a fixed-width reset through the public mask boundary.

Parameters:

reset_mask (Tensor)
qpos (Tensor)
qvel (Tensor)

Return type:

dict | None

validate_finite()[source]¶

Return type:: None