FELT Flow - FELT Labs command line too

Running federated learning via python

Installation

1. Install Package

First you have to install this python package:

git clone https://github.com/FELT-Labs/python-flow.git
cd python-flow
pip install .

2. Environment variables

You can rename .env.example file in python-flow directory to .env and place correct values for following variables:

export WEB3_INFURA_PROJECT_ID="a0b1..."
export PRIVATE_KEY="0xa0b1..."

Alternatively you can export those variables directly in your terminal or place these exports into .bashrc file.

• WEB3_INFURA_PROJECT_ID can be obtained from infura.io and is used for accessing nodes on blockchain
• PRIVATE_KEY is private key of your blockchain account which will be used for signing the transaction (purchasing data and ordering training)

After that you should be ready for running FELT Flow.

Examples

Example command line for running FELT Flow:

felt-flow --launch_token <launch_token_from_feltlabs_app>

This token is obtained through our web application: app.feltlabs.ai.

Development

Install

pip install -e .
pip install -r requirements-dev.txt
pre-commit install

Test

pytest --cov=feltflow --cov-report=term

Versioning

Do following steps when updating to new version:

1. Update version number in setup.py
2. Run following commands (replace 0.0.0 with correct version number):

   git add -A
   git commit -m "New version 0.0.0"
   git tag v0.0.0
   git push origin v0.0.0

3. Create new release on Github based on tag

Potential issues

Error while running pip install .

There might be some issues with installing ocean-lib package. Mainly on mac M1 problems with installing wheel or coincurve Workaround is explained here https://github.com/oceanprotocol/ocean.py/blob/main/READMEs/install.md