Skip to content

scrambledpie/toyGPT

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

29 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

toyGPT

An implementation of a GPT model in plain jax/jax numpy trained on wine reviews

  • Self attention layer implemented from scratch in JAX
  • mixed precision support
  • multi-GPU support (data parrallel)

Kaggle wine reviews dataset

Setup

  • install JAX
  • download the dataset and unzip into datasets/

Usage

All entry points to run code are in scripts/

  • train.py uses full precision training on a single GPU

About

A small toy GPT model implemented in JAX from scratch

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages