@chongshenng/pathmnist-fedbn

0
4
flwr new @chongshenng/pathmnist-fedbn

PathMNIST FedBN with Flower

This Flower / PyTorch app trains a compact CNN on PathMNIST with FedBN. It follows the Flower PyTorch quickstart structure and uses the Message API (ServerApp, ClientApp, Message, ArrayRecord, and MetricRecord).

The app does not write model artifacts, metric artifacts, or other local training outputs. Running it only executes the federated training and evaluation workflow.

Project layout

.
├── pathmnist_fedbn
│   ├── __init__.py
│   ├── client_app.py
│   ├── server_app.py
│   └── task.py
├── pyproject.toml
└── README.md

Install

pip install -e .

With uv:

uv sync --python 3.12

Run

flwr run . --stream

Override run settings with --run-config:

flwr run . \
  --run-config "num-server-rounds=1 train-samples-per-client=8 val-samples-per-client=8" \
  --stream