MolecularDiffusion.modules.models.noisemodel¶

Classes¶

Functions¶

`clip_noise_schedule`(alphas2[, clip_value])	For a noise schedule given by alpha^2, this clips alpha_t / alpha_t-1. This may help improve stability during
`cosine_schedule`(timesteps, nu_array[, s])	A noise schedule based on the cosine schedule from Nichol & Dhariwal 2021.
`plot_cosine_scheduler`(timesteps[, powers, s])
`polynomial_schedule`(timesteps[, powers, s])	A noise schedule based on a simple polynomial equation: 1 - x^power.

Module Contents¶

class MolecularDiffusion.modules.models.noisemodel.NoiseModel(timestep, noise_precision=1e-05, nu_arr=[0.5, 0.5, 0.5, 0.5], mapping=['pos', 'categorical', 'integer', 'extra'], device='cpu')¶

get_alpha_bar(t_normalized=None, t_int=None, key=None)¶

get_alpha_sigma_t(t_int)¶

get_sigma_bar(t_normalized=None, t_int=None, key=None)¶

T¶

inverse_mapping¶

mapping = ['pos', 'categorical', 'integer', 'extra']¶

nu_arr = [0.5, 0.5, 0.5, 0.5]¶

MolecularDiffusion.modules.models.noisemodel.clip_noise_schedule(alphas2, clip_value=0.001)¶: For a noise schedule given by alpha^2, this clips alpha_t / alpha_t-1. This may help improve stability during sampling.

MolecularDiffusion.modules.models.noisemodel.cosine_schedule(timesteps: int, nu_array: List[float], s: float = 0.0001)¶: A noise schedule based on the cosine schedule from Nichol & Dhariwal 2021.

MolecularDiffusion.modules.models.noisemodel.plot_cosine_scheduler(timesteps, powers=[2], s=0.0001)¶

MolecularDiffusion.modules.models.noisemodel.polynomial_schedule(timesteps: int, powers: list = [2], s: float = 0.0001)¶: A noise schedule based on a simple polynomial equation: 1 - x^power.