stable_pretraining.data.synthetic_data

Synthetic and simulated data generators.

This module provides various synthetic data generators including manifold datasets, noise generators, statistical models, and simulated environments for testing and experimentation purposes.

Functions

generate_perlin_noise_2d(shape, res[, ...])	Generate 2D Perlin noise.
perlin_noise_3d(x, y, z)	Generate 3D Perlin noise at given coordinates.
swiss_roll(N[, margin, sampler_time, ...])	Generate Swiss Roll dataset points.

Classes

Categorical(values, probabilities)	Categorical distribution for sampling discrete values with given probabilities.
ExponentialMixtureNoiseModel(rates, prior[, ...])	Exponential mixture noise model for data augmentation or sampling.
ExponentialNormalNoiseModel(rate, mean, std, ...)	Exponential-normal noise model combining exponential and normal distributions.
GMM([num_components, num_samples, dim])	Gaussian Mixture Model dataset for synthetic data generation.
MinariEpisodeDataset(dataset)	Dataset for Minari reinforcement learning data with episode-based access.
MinariStepsDataset(dataset[, num_steps, ...])	Dataset for Minari reinforcement learning data with step-based access.