Confidential ML Training 🏋️

Python client library for Training instance of the decentriq platform.

What is Confidential ML Training?

By using Confidential Computing, Confidential ML Training enables you to train machine learning models based on data that nobody ever can access; not you, not us, not the infrastructure provider, nobody. This removes the risk for data breaches or data misuse.

Installation

Prerequisites

You need to have Python >= 3.6.1 installed (check using python --version) and internet access.

Release

The library is published on PyPi so it can be easily installed by running:

pip install avato-training

Development

After cloning the repository you can use different installation methods:

Poetry

To install with poetry just run:

poetry install

Notes:

1. If you get an error during the installation try to delete the poetry.lock file

2. Poetry installs the library in its own virtualenv, if you want to use it in your global python installation disable virtualenvs in poetry

poetry config settings.virtualenvs.create false

Pip

To install with pip just run:

pip install .

Usage:

To use the library just import the avato and avato_training modules and start interacting with the platform!

Example-1: instance creation

from avato import Client
from avato_training import TRAINING_Instance

client = Client(
    api_token="SECRET_TOKEN",
    instance_types=[TRAINING_Instance]
)
instance = client.create_instance(
    "genesis", TRAINING_Instance.type, ["friend@decentriq.ch"]
)
client.get_instances()
instance.shutdown()
instance.delete()

Examples

Some examples showing how the library can be used are provided in the examples folder. For more examples check out the tests directory