Rescale an OBJ mesh by a given scale factor. Args: input_path (str): Path to input OBJ file scale_factor (float): Scale fact

Reset all envs and return observations.

Reset the environment.

Reset environment. Use this to reset custom member variables. Always keep ``seed`` in the signature and forward it to ``super().reset

Reset the env; reset the checker unless ``skip_checker_reset`` is set. ``seed`` is forwarded to ``super().reset``. Leaves that set ``

Reset the checker, then the environment. The checker is reset *before* ``super().reset()`` — super().reset() may invoke reset-call

Called after switching the controller.

Capture the per-env starting position on reset (called from curriculum).

Reset episode tracking variables

Reshape selected dimensions for all tensors in nested dictionary or list or tuple to a target dimension. Args: x (dict or list o

Per-joint (lo, hi) clip range tensor, or None if unbounded.

Match a list of query regular expressions against a list of strings and return the matched indices and names. When a list of query regular expres

Resolve the EXACT URDF instance dir RoboTwin's create_sapien_urdf_obj uses. This must match RoboTwin 1:1, because modelid (random per episode) se

Controller-free rollout on robosuite's own ``MjSim``. ``ctrl_builder`` maps the model's actuator count ``nu`` to a concrete :data:`~tools.rob

Locate a local RoboTwin clone root (the dir that contains ``assets/``), or None. ``ROBOTWIN_ASSETS`` may point at either the repo root or its ``a

ManiSkill RollBall-v1 ``compute_dense_reward`` (max 30). Stateful: pass ``reached_status`` in. Returns ``(reward, reached_status)``; ``reached_st

Roll out trained policy and return mean ep_r.

Step MetaSim's `maniskill.pick_cube_v1` task and grab rgb frames.

Convert lerobot Rotation to numpy quaternion (w, x, y, z).

Convert a 3x3 rotation matrix to a normalized XYZW quaternion.

Run native-only benchmark teleop packets through one simulation session.

Run one task, write artifacts, return a summary dict.

Run trajectory replay with randomization.

Runs the simulation loop.

Runs the simulation loop.

For each image, randomly sample @num_crops crops of size (@crop_height, @crop_width), from @images. Args: images (torch.Tensor):

Samples t and r from a log-normal distribution.

Save all collected episodes to file

Save model parameters and training configuration to disk.

Save current poses of all objects and robots to a Python file from observation Args: obs: TensorState observation from task scena

Save current poses of all objects and robots to a Python file

Save current poses of all objects and robots to a Python file.

Set the entire sim state (e.g. a benchmark demo's initial frame).

This function disables printing when not in master process

Creates live visualizers for manager terms.

Concat frame-by-frame; pad shorter side to the same length.

Render both rollouts on the same MJCF and concat into a side-by-side mp4.

original dataset grasp_label files have redundant data, We can significantly save the memory cost