Autoregressive Models API Reference

Complete API documentation for autoregressive models in Artifex.

Coming Soon

Autoregressive model implementations are planned for a future release. This documentation will be updated when the feature is available.

Overview

Autoregressive models will include:

• PixelCNN: Pixel-by-pixel image generation
• WaveNet: Audio waveform generation
• Transformer-based: GPT-style text generation
• MAR (Masked Autoregressive): Masked autoregressive models

Planned API

Base Class

from artifex.generative_models.models.autoregressive.base import AutoregressiveModel

model = AutoregressiveModel(
    config: AutoregressiveConfig,
    *,
    rngs: nnx.Rngs,
)

Configuration

from artifex.generative_models.core.configuration import AutoregressiveConfig

config = AutoregressiveConfig(
    name="autoregressive_model",
    sequence_length=256,
    vocab_size=10000,
    hidden_dim=512,
    num_layers=6,
)

Related Documentation

• Autoregressive Models Guide - Conceptual overview
• Autoregressive Concepts - Understanding autoregressive modeling

References

• van den Oord et al., "Pixel Recurrent Neural Networks" (2016)
• van den Oord et al., "WaveNet: A Generative Model for Raw Audio" (2016)
• Radford et al., "Language Models are Unsupervised Multitask Learners" (2019)