Building Socially-Equitable Public Models

This is the PyTorch implementation of our paper:

Building Socially-Equitable Public Models
Yejia Liu, Jianyi Yang, Pengfei Li, Tongxin Li, Shaolei Ren

The Forty-first International Conference on Machine Learning (ICML 2024)

[ arXiv ] [ GitHub ]

Installation

1. Create and activate a local conda environment:

conda create --name <env_name> python=3.7
conda activate <env_name>

3. Install PyTorch following the instructions on https://pytorch.org/ (we used PyTorch 1.5.1 in our experiments). For example,

conda install pytorch==1.5.1 torchvision==0.6.1 -c pytorch

4. Clone the repository and enter its root directory:

git clone https://github.com/Ren-Research/Socially-Equitable-Public-Models.git
cd Socially-Equitable-Public-Models

5. Install required packages and dependencies in requirement.txt:

conda install --file requirements.txt

6. Download data following this guide

Training and Evaluation

Run Trained Public Models for Paper Results

For reproducibility, we release our trained public models here.
To run these trained models, please use the script run_public_models.sh:

sh run_public_models.sh

Train Your Own Public Models

• For the Data Center Workload application, here are example training commands:

# Train an equitable model with different lambda, similar groups
python main_dc_workload.py --training --lr 0.05 --n_epochs 50 --batch_size 128 --diff_lambda --q_idx 1.5

# Train an equitable model with same lambda, similar groups
python main_dc_workload.py --training --lr 0.05 --n_epochs 50 --batch_size 128 --q_idx 1.1

# Train a baseline model with different lambda, different groups
python main_dc_workload.py --training --lr 0.05 --n_epochs 100 --batch_size 128 --diff_lambda --baseline --diff_group_dist

• For the EV Charging application, here are example training commands:

# Train an equitable model with similar groups
python -m pdb main_ev_charging.py --training --lr 1e-4 --n_epochs 50 --batch_size 128 --q_idx 30

# Train a baseline model with different groups
python -m pdb main_ev_charging.py --training --lr 1e-4 --n_epochs 100 --batch_size 128 --diff_group_dist

Visualization

We provide the code for our paper plots in the following Jupyter python notebooks:

• dc_plots
• ev_plots

Citation

@article{SociallyEquitablePUblicModel_ICML_2024,
  title = {Building Socially-Equitable Public Models},
  author = {Liu, Yejia and Yang, Jianyi and Li, Pengfei and Li, Tongxin and Ren, Shaolei},
  journal = {ICML},
  year = {2024}
}