floralab 0.1.0

Python CLI for interacting with Flower-AI stacks on SLURM via florago

FloraLab

FloraLab is a Python CLI tool for managing Flower-AI federated learning deployments on SLURM clusters.

Features

• 🚀 Automated deployment of Flower-AI stacks on SLURM
• 🔧 Seamless integration with florago backend
• 📦 Bundled binaries - no manual setup required
• 🌐 SSH tunnel management for secure communication
• ⚡ Simple 3-command workflow: init, run, stop

Installation

pip install floralab

Or with uv:

uv pip install floralab

Quick Start

1. Initialize Configuration

floralab-cli init your-slurm-login-node.edu

This adds Flower-AI configuration to your pyproject.toml.

2. Run Federated Learning Job

floralab-cli run --nodes 3

This will:

• Copy florago to the SLURM cluster
• Initialize the environment
• Start the API server
• Spin up a Flower stack (1 server + 3 clients)
• Run your federated learning job with flwr run

3. Stop the Stack

floralab-cli stop

Tears down the Flower stack and cancels the SLURM job.

Commands

floralab-cli init <login-node>

Initialize floralab configuration in your project's pyproject.toml.

Arguments:

• login-node - Hostname of your SLURM login node

Options:

• --dir, -d - Project directory (default: current directory)

floralab-cli run

Run a federated learning job on the SLURM cluster.

Options:

• --nodes, -n - Number of client nodes (default: 2)
• --partition, -p - SLURM partition
• --memory, -m - Memory per node (e.g., "4G")
• --time, -t - Time limit (e.g., "01:00:00")
• --dir, -d - Project directory (default: current directory)
• --ssh-port - Local port for SSH tunnel (default: 8080)

floralab-cli stop

Stop the running Flower stack.

Options:

• --dir, -d - Project directory (default: current directory)
• --ssh-port - Local port for SSH tunnel (default: 8080)

Configuration

FloraLab adds the following to your pyproject.toml:

[tool.floralab]
login-node = "your-slurm-login-node.edu"

[tool.flwr.federations.floralab]
address = "127.0.0.1:9093"
insecure = true
root-certificates = null

The address field is automatically updated when you run floralab-cli run to point to the actual Flower server.

Requirements

• Python 3.12+
• SSH access to a SLURM cluster
• Flower AI installed (flwr command available)

License

MIT