Deep Vector Quantization. Turn images into sequences of numbers that AI models can learn from.

github.com/karpathy/deep-vector-quantization

Vaya's read on this project

Problem, audience, market, and the verdict — sign in to see it.

Updates

March 2021
  • bunch of changes and hygiene. everything still needs much tuning
February 2021
  • refactoring the skeleton to create a modular loss function, an L2 los…
  • refactoring the api to make it more modular and symmetric, args clean…
  • sequester loss-related stuff to its own file to reduce clutter
  • add openai dall-e models as options for training
  • ramp/decay alpha, beta, tau as openai describes in dall-e roouughly. …
  • make model properly own its args
  • just putting some lipstick on a pig
  • reshuffling the code around a bit as a step 1 of conforming a bit bet…
  • refactoring code structure into more modular form in preparation for …
  • Merge branch 'main' of github.com:karpathy/deep-vector-quantization i…
  • apply commitment loss to to the other term, ty @CDitzel
  • Update README.md
  • president's day h4xzoring on vqvaes