Draw arrows for active impulse forces.

Detect which ``ActuatorCmd`` field an XML actuator expects as ctrl input. Uses strict predicates that match the invariants set by MuJoCo's ``mjs_

Detect NaN/Inf values in physics state (qpos, qvel, qacc, qacc_warmstart). Args: data: MuJoCo simulation data containing physics state.

Get the dimensionality of the joint in qvel.

Download and verify an asset if needed.

Evaluate a single run and return metrics with metadata.

Filter names based on regex patterns.

Format a duration in seconds as a human-readable time-ago string.

Get a fresh Go1 robot configuration instance. Returns a new EntityCfg instance each time to avoid mutation issues when the config is shared acros

Get the configuration of a specific event term by name.

Return ``(name, func)`` pairs for class-based terms with debug_vis.

Convert HSV to RGB color space. Args: hsv: HSV color representation. Returns: RGB color tuple with values in range [0, 1].

Penalize joint torques applied on the articulation using L2 squared kernel.

Create base cube lifting task configuration.

Create base velocity tracking task configuration.

Convert rotations given as Euler angles (intrinsic) in radians to rotation matrices. Args: euler_angles: Euler angles in radians. Sh

Print a top-level usage message and exit if argv is bare -h/--help. The ``train`` and ``play`` entrypoints use a two-stage tyro parse with ``add_

Return option field names that differ from MjSpec defaults.

Reset debug visuals when switching selected environment.

Request a scene refresh when switching environments with contacts enabled.

Called when the viewer pause state changes.

Converts poses from one coordinate frame to another. Transforms matrices representing point C in frame A to matrices representing the same po

Get the full name with entity prefix if applicable.

Get the dimensionality of the joint in qpos.

Performs spherical linear interpolation (SLERP) between two quaternions. This function does not support batch processing. Args: q1:

Queue environment debug visualizers for the current frame.

Called at the end of every ``env.step()`` with fresh observations. At this point ``obs_buf`` holds the new observation for every environment and

Called in ``env.step()`` before terminated environments are reset. **What is available:** - ``obs_buf`` contains the observation from the *e

Compute reflected inertia of a single-stage gearbox.

Convert RGB to HSV color space. Args: rgb: RGB color tuple with values in range [0, 1]. Returns: HSV color representation.

Run policy evaluation and compute metrics.

Set effort limits for specified environments. Args: env_ids: Environment indices to update. effort_limit: New effort limits. Shape: (

Set PD gains for specified environments. Args: env_ids: Environment indices to update. kp: New proportional gains. Shape: (num_envs,

Step the environment and optionally record video. Args: action: Action tensor. Returns: Tuple of (obs, reward, terminated, t

Advance one tick: drain actions, step physics, maybe render. Returns True when a render frame was produced, False otherwise.

Create RL runner configuration for Unitree G1 velocity task.

Create RL runner configuration for Unitree G1 tracking task.

Create RL runner configuration for Unitree Go1 velocity task.

Splits transformation matrices into positions and rotation matrices. Args: pose: Batch of pose matrices with last 2 dimensions of (4, 4).

Update the displayed environment index.

Recompute ``body_pos_relative_w`` and ``body_quat_relative_w``. Called after ``reset_to_frame`` so that termination checks that compare relat

Optional validation method to be overridden by subclasses.

Validate collision configuration parameters.

Validate that all variant specs share the same kinematic topology. Performs recursive body-tree, joint, and primitive-geom comparison; spec-level

Verify file integrity using SHA256.

Multi-cube task: depth + segmentation mask for goal conditioning.

Extract the yaw component of a quaternion. Args: quat: The orientation in (w, x, y, z). Shape is (..., 4) Returns: A quatern

Apply noise to input data.

Set item(s) in the tensor using standard indexing.

Draw a downward arrow above the selected environment.