DART:{ 'transl': 3, 'poses_6d': 22 * 6, 'transl_delta': 3, 'global_orient_delta_6d': 6, 'joints': 22 * 3, 'joints_delta':

trans_local: (B, T, 3) delta_trans_local: (B, T, 3) height: (B, T, 1) sincos: (B, T, 1) delta_yaw: (B, T, 1) contact_T: (B, T

Computes the frame blend for the motion.

Computes the frame blend for the motion.

Compute text embeddings efficiently

Computes the velocities of the motion.

Computes the velocities of the motion.

Configure multi-gpu training.

Convert batch of motion data to motion features

Generate all primitives from a single motion segment with proper overlapping

初始化buffer存储轨迹数据

Interpolates the motion to the output fps.

Interpolates the motion to the output fps.

Load and prepare data efficiently

Load or compute mean/std for normalization

Loads the motion from the motion data dict.

Loads the motion from the csv file.

Load motion statistics

Load or compute text embeddings

Load VAE from checkpoint following DART's approach

Load or compute mean/std for normalization

Organize primitives by primitive index for batching

Parse MJCF file to extract skeleton structure

Get the prior KL term for the variational lower-bound, measured in bits-per-dim. This term can't be optimized, as it only de

Publish all time steps from the current buffer using cached message instance

Sample a batch of motions and generate all their primitives

Spherical linear interpolation between two quaternions.

Spherical linear interpolation between two quaternions.

Computes the derivative of a sequence of SO3 rotations. Args: rotations: shape (B, 4). dt: time step. Returns

Computes the derivative of a sequence of SO3 rotations. Args: rotations: shape (B, 4). dt: time step. Returns

Returns torch.sqrt(torch.max(0, x)) but with a zero subgradient where x is 0.

Returns torch.sqrt(torch.max(0, x)) but with a zero subgradient where x is 0.

更新buffer，从motion数据中填充buffer

Get a term for the variational lower-bound. The resulting units are bits (rather than nats, as one might expect). This allow

Convert rotations given as axis/angle to quaternions. # wxyz Args: axis_angle: Rotations given as a vector in axis angle form,

Create a beta schedule that discretizes the given alpha_t_bar function, which defines the cumulative product of (1-beta) over time from t = [

Broadcast model parameters to all GPUs.

Compute observation matching PrivPrivObservationsCfg structure

Compute the mean for the previous step, given a function cond_fn that computes the gradient of a conditional log probability with res

Compute the mean for the previous step, given a function cond_fn that computes the gradient of a conditional log probability with res

Create a zero motion NPZ file using robotmdar's get_zero_feature directly Args: out: Path to save the NPZ file num_times

Use DDIM to sample from the model and yield intermediate samples from each timestep of DDIM. Same usage as p_sample_loop_pro

Use DDIM to sample from the model and yield intermediate samples from each timestep of DDIM. Same usage as p_sample_loop_pro

Sample x_{t-1} from the model using DDIM. Same usage as p_sample().

Denormalize features

Compute the log-likelihood of a Gaussian distribution discretizing to a given image. :param x: the target images. It is assumed that thi

Display benchmark statistics.