Dynamic dispatch table for GLFW functions required by Simulate. This allows us to use GLFW without introducing a link-time dependency on the library,

Normalize mean and variance of values based on empirical values.

Observations for policy group.

On-policy runner for training and evaluation.

Configuration for a replay motions scene.

Action specifications for the MDP.

Command specifications for the MDP.

Curriculum terms for the MDP.

r"""Discounted average of rewards. The discounted average is defined as: .. math:: \bar{R}_t = \gamma \bar{R}_{t-1} + r_t Args

Reward normalization from Pathak's large scale study on PPO. Reward normalization. Since the reward function is non-stationary, it is useful to n

Configuration for events.

Forward kinematics calculator for humanoid robots. Maintains full compatibility with original interface including extend joints.

Utilities for training and sampling diffusion models. Ported directly from here, and then adapted over time to further experimentation.

State management for continuous motion generation.

Configuration for the terrain scene with a legged robot.

Summary writer for Neptune.

Observation specifications for the MDP.

Observation specifications for the MDP.

A module to convert 1D signal tensors into patches using torch operations.

Absolute pos embedding, learned.

This is a more standard version of the position embedding, very similar to the one used by the Attention is all you need paper, generalized t

Observation specifications for the MDP.

Observation specifications for the MDP.

Observation specifications for the MDP.

Implementation of Random Network Distillation (RND) [1] References: .. [1] Burda, Yuri, et al. "Exploration by random network distillatio

Configuration for a replay motions scene.

Configuration for a replay motions scene.

Reward terms for the MDP.

Dumps key/value pairs into TensorBoard's numeric format.

Termination terms for the MDP.

A module to convert 1D patches back into signal tensors.

Summary writer for Weights and Biases.

Abstract base class for training managers. Defines common interfaces and functionality for training management, including EMA support, h

Training manager for DAR (Diffusion AutoRegressive) model. Manages both VAE and denoiser models, with support for full DDPM sampling and

Ideal PD actuator with delayed command application. This class extends the :class:`IdealPDActuator` class by adding a delay to the actuator comma

Configuration for a delayed PD actuator.

Distillation algorithm for training a student model to mimic a teacher model.