This repository contains the code for the project "Federated Learning from Scratch" for the course "Continual Learning" at the University of Pisa
.
├── README.md
├── histories
│ ├── history_5e10c.pkl
│ ├── history_5e10c_fashion.pkl
│ ├── history_5e128c.pkl
│ ├── history_5e20c.pkl
│ ├── history_5e5c.pkl
│ └── history_niid.pkl
├── models
│ ├── efficientnet_b0.pth
│ ├── efficientnet_cifar10.pth
│ ├── mnist_cnn_model.pth
│ ├── model5e10c_fashion.pth
│ ├── model_niid.pth
│ └── resnet18_cifar10.pth
├── requirements.txt
└── src
├── federated.py
├── finetuning.py
├── models.py
├── report.ipynb
└── utils.py
To run the code, you need to first install the required packages. You can do this by running the following command:
pip install -r requirements.txt
After that, you can run the code by running the following command:
python src/fedmnist.py [--help]
The script will train a federated model on the MNIST dataset. You can specify the number of clients, the number of epochs, the batch size, the learning rate, the optimizer, the model, and the dataset.
usage: fedmnist.py [-h] [--n_clients N_CLIENTS] [--local_epochs LOCAL_EPOCHS] [--lr LR] [--n_samples N_SAMPLES] [--device DEVICE] [--weight_decay WEIGHT_DECAY] [--iid IID] [--rounds ROUNDS] [--jobs JOBS]
FedMNIST
options:
-h, --help show this help message and exit
--n_clients N_CLIENTS
number of clients to initialize
--local_epochs LOCAL_EPOCHS
number of local epochs
--lr LR learning rate
--n_samples N_SAMPLES
number of samples per client
--device DEVICE device to train on
--weight_decay WEIGHT_DECAY
weight decay for optimizer
--iid IID whether to sample clients iid or not
--rounds ROUNDS number of rounds to train for
--jobs JOBS number of jobs to run in parallelResults of the experiments are reported in the norebook report.ipynb.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent