TorchEBM API Reference¶

Welcome to the TorchEBM API reference documentation. This section provides detailed information about the classes and functions available in TorchEBM.

Package Structure¶

TorchEBM is organized into several modules:

Getting Started with the API¶

If you're new to TorchEBM, we recommend starting with the following classes:

BaseModel: Base class for all models
BaseSampler: Base class for all sampling algorithms
LangevinDynamics: Implementation of Langevin dynamics sampling

Core Components¶

Models¶

TorchEBM provides various built-in models:

Model	Description
`GaussianModel`	Multivariate Gaussian energy function
`DoubleWellModel`	Double well potential energy function
`RastriginModel`	Rastrigin function for testing optimization algorithms
`RosenbrockModel`	Rosenbrock function (banana function)
`AckleyModel`	Ackley function, a multimodal test function
`HarmonicModel`	Harmonic oscillator energy function

Samplers¶

Available sampling algorithms:

Sampler	Description
`LangevinDynamics`	Langevin dynamics sampling algorithm
`HamiltonianMonteCarlo`	Hamiltonian Monte Carlo sampling

BaseLoss Functions¶

TorchEBM implements several loss functions for training EBMs:

BaseLoss Function	Description
`ContrastiveDivergence`	Standard contrastive divergence (CD-k)
`PersistentContrastiveDivergence`	Persistent contrastive divergence
`ParallelTemperingCD`	Parallel tempering contrastive divergence

Module Details¶

For detailed information about each module, follow the links below: