Distribution Base

Module: generative_models.core.distributions.base

Source: generative_models/core/distributions/base.py

Overview

Concrete shared base class for Artifex probability distributions.

This module owns the reusable runtime behavior for:

• sampling with nnx.Rngs
• log-probability evaluation
• entropy and KL caching
• numerical-stability helpers
• batching and chunked vectorization helpers

Classes

Distribution

class Distribution

Shared concrete distribution foundation for Artifex distribution modules.

Key Methods

sample

def sample()

Draw samples using the distribution's RNG-aware runtime.

log_prob

def log_prob()

Compute log probabilities with parameter validation and finite-value checks.

entropy

def entropy()

Compute entropy with optional cache reuse in non-JIT contexts.

kl_divergence

def kl_divergence()

Compute KL divergence against another Artifex distribution with optional cache reuse.