rosinality/vq-vae-2-pytorch

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1803 stars · 281 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A PyTorch implementation of VQ-VAE-2 for generating diverse high-fidelity images, including training scripts for the VQ-VAE and PixelSNAIL prior stages. It supports hierarchical top/bottom priors at 256px resolution and distributed training across multiple GPUs.

Use cases

• train a VQ-VAE-2 model on my own image dataset
• generate high-fidelity images with a vector quantized VAE in PyTorch
• reproduce the VQ-VAE-2 paper results on FFHQ
• train a PixelSNAIL prior over discrete image codes
• extract quantized codes from images into an LMDB store
• learn how VQ-VAE-2 works with example code

When to choose

• you want a reference PyTorch implementation of VQ-VAE-2 for research or learning
• you need to train a two-stage generative model (VQ-VAE + autoregressive prior) on 256px images
• you want pretrained FFHQ checkpoints to start from

When to avoid

• you need a production-ready or actively maintained image generation library
• you want modern diffusion-based image generation instead of VQ-VAE
• you need a polished API rather than research scripts

Facets

library · maturity maintenance

machine-learning deep-learning image-processing deep-learning machine-learning image-processing python vq-vae pytorch pixelsnail image-generation generative-models research-code gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="rosinality/vq-vae-2-pytorch")