constants
Bases: StrEnum
Enum for latent-related feature keys used in variational models.
Bases: StrEnum
Enum for different Mixture of Experts (MoE) routing strategies.
Bases: StrEnum
Initialization strategies for GMM mixture components.
Bases: StrEnum
What the diffusion model predicts.
Bases: StrEnum
Beta schedule for diffusion models.
Bases: StrEnum
Variance type for DDPM scheduler.
Bases: StrEnum
ODE solver types for flow matching.
Bases: StrEnum
Algorithm type for denoising-based generative models.
Bases: StrEnum
Sampling strategy for mixture-of-experts inference.
Bases: StrEnum
Types of Diffusion Transformer architectures.
Bases: StrEnum
Timestep conditioning strategy for denoising decoders.
Bases: StrEnum
How decoder action heads map to the action space.
Bases: StrEnum
Keys the algorithm injects into the decoder feature dict as inputs.
Bases: StrEnum
Keys for decoder outputs.