Extract values from a 1-D numpy array for a batch of indices. :param arr: the 1-D numpy array. :param timesteps: a tensor of indices int

Load the parameters of the actor-critic model. Args: state_dict (dict): State dictionary of the model. strict (bool):

update an entire section of ui0

modify UI

Reconstruct motion from features

Convert Euler angles to quaternion. Input: euler (..., 3) → roll, pitch, yaw Output: quaternion (..., 4) in xyzw order

Apply input action on the environment. The extra information is a dictionary. It includes metrics such as the episode reward, episode length,

Normalize features

Convert quaternion (xyzw) to Euler angles (roll, pitch, yaw) using XYZ convention. Input: q (..., 4) with xyzw format Output: (..., 3) →

Helper function to create Float32MultiArray with proper dimension info Args: data: numpy array or list of data

输入: motion_dict (root_trans_offset, root_rot, dof, contact_mask) 输出: FK后的全局位姿信息（dict，含global_translation, global_rotation等） 支

Apply the model to get p(x_{t-1} | x_t), as well as a prediction of the initial x, x_0. :param model: the model, which takes

Diffuse the dataset for a given number of diffusion steps. In other words, sample from q(x_t | x_0). :param x_start: the in

Save the model and training information.

Generate next motion segment using DAR model. Args: vae: VAE model for encoding/decoding denoiser: Denoiser model for di

Write the sequence of args, with no separators, to the console and output files (if you've configured an output file).

Convert rotation matrix to 6D rotation representation. Args: rotmat (B x T x 3 x 3): Batch of rotation matrices with temporal dimensi

Take the mean over all non-batch dimensions.

Pick the Last frame's absolute pose from the feature

WXYZ Convert rotations given as quaternions to rotation matrices. Args: quaternions: quaternions with real part first,

Helper function to create Float32MultiArray with proper dimension info

Extract data from Float32MultiArray message

每步训练后调用

每步训练前调用

Compute the mean and variance of the diffusion posterior: q(x_{t-1} | x_t, x_0)

The inverse of the rotation

Reset all environment instances. Returns: Tuple containing the observations and extras.

Importance-sample timesteps for a batch. :param batch_size: the number of timesteps. :param device: the torch device to save

clear all times