models

DDM(*, sigma_min=0.05, sigma_max=0.5, is_periodic=True, mlp_network, key, n_sample_steps=100, n_epochs=100, batch_size=128, wandb_log=False, gamma_energy_regulariztion=1e-05, fourier_features=1, warmup_steps=50, box_size=1.0, symmetric=False) dataclass

Bases: LinearPriorSchedule, UniformPrior, ExponetialVarianceNoiseSchedule, DefaultDataClass

Energy-based denoising diffusion model for periodic data on [0, 1].

DrivenDDM(*, sigma_min=0.05, sigma_max=0.5, is_periodic=True, mlp_network, key, n_sample_steps=100, n_epochs=100, batch_size=128, wandb_log=False, gamma_energy_regulariztion=1e-05, fourier_features=1, warmup_steps=50, box_size=1.0, symmetric=False, pbc_bins=20, diffusion=lambda x: 1.0) dataclass

Bases: LinearForceSchedule, DDM

EB-based denoising diffusion model for driven periodic data on [0, 1].