Skip to content

Minimal implementation of LiGO (Learning to Grow Pretrained Models for Efficient Transformer Training))

License

Notifications You must be signed in to change notification settings

thanhlexyz/ligo

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

22 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

LiGO Minimal

Minimal implementation of LiGO (Learning to Grow Pretrained Models for Efficient Transformer Training)

Installation

Create virtual environment and install dependencies

python3 -m venv venv
source venv/bin/activate
pip3 install -r requirements
cd core

Usage

  • Train a small FC model from scratch for MNIST
cd core/
python3 main.py --model=fc1 --initializer=scratch
  • Training deeper and wider FC model, with LiGO initialization from small model for MNIST
python3 main.py --model=fc2 --initializer=ligo --pretrain_model=fc1_scratch
  • The main implementation is in ligo.py

About

Minimal implementation of LiGO (Learning to Grow Pretrained Models for Efficient Transformer Training))

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •