Skip to content

Latest commit

 

History

History
11 lines (8 loc) · 607 Bytes

README.md

File metadata and controls

11 lines (8 loc) · 607 Bytes

mlx-bitnet-mingpt

This repo contains example code to demonstrate the following:

  1. A port of Karpathy's minGPT in MLX.
  2. An extension of mlx-mingpt to use BitLinear instead of nn.Linear to demonstrate training language models using 2-bits as reported in https://arxiv.org/abs/2402.17764

Reports

Take a peek at the ipynb pdfs under demo-results folder