Skip to content
This repository was archived by the owner on May 21, 2025. It is now read-only.

MUN-DML/QuickP

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Efficient Device Placement for Distributed DNN Training (ICC 2025)

DOI

Installation

Install Gurobi certificate.

This solution requires the solver Gurobi to be installed and have an active license that students can apply for free

Dependency

Details are listed inside this project

Run steps

First, clone this project and the submodule of it from repo using SSH:

git clone git@github.com:MUN-DML/QuickP.git

Then, step into the ICC2025 folder:

cd ICC2025

Configure the environment variable for Gurobi license

os.environ['GRB_LICENSE_FILE'] = 'your_path/gurobi.lic'

Then, enter QuickP.py and adjust input parameters:

  • number_of_device (integer): number of GPUs,
  • model (string): DNN model name: ALEXNET, VGG, FNET, BERT,
  • beta (integer): WCC expanding threshold,
  • alpha (integer): merging threshold for operator fusion. Find the line "alpha = get_proper_alpha(comp_graph, deviceTopo, if_visualize=False)" and replace the function with your self-defined value,

Finally, make sure your current working directory is ICC2025 and run QuickP.py

python QuickP.py

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Contributors

Languages