SecretFlow is a unified framework for privacy-preserving knowledge intelligence and machine studying. To realize this aim, it offers:
- An summary system layer consists of plain gadgets and secret gadgets which encapsulate numerous cryptographic protocols.
- A tool stream layer modeling greater algorithms as system object stream and DAG.
- An algorithm layer to do knowledge evaluation and machine studying with horizontal or vertical partitioned knowledge.
- A workflow layer that seamlessly integrates knowledge processing, mannequin coaching, and hyperparameter tuning.
Â
Set up
For customers who wish to strive SecretFlow, you’ll be able to set up the present launch from pypi. Be aware that it requires python model == 3.8, you’ll be able to create a digital setting with conda if not happy.
pip set up -U secretflow
Strive you first SecretFlow program
>>> import secretflow as sf
>>> sf.init(['alice', 'bob', 'carol'], num_cpus=8, log_to_driver=True)
>>> dev = sf.PYU('alice')
>>> import numpy as np
>>> knowledge = dev(np.random.rand)(3, 4)
>>> knowledge
<secretflow.system.system.pyu.PYUObject object at 0x7fdec24a15b0>
Getting began
Deployment
Contribution information
For builders who wish to contribute to SecretFlow, you’ll be able to arrange an setting with the next instruction.
git clone https://github.com/secretflow/secretflow.git# elective
git lfs set up
conda create -n secretflow python=3.8
conda activate secretflow
pip set up -r dev-requirements.txt -r necessities.txt
Coding Type
We choose black as our code formatter. For numerous editor customers, please check with editor integration. Cross -S, --skip-string-normalization
to black to keep away from string quotes or prefixes normalization.
Disclaimer
Non-release variations of SecretFlow are prohibited to make use of in any manufacturing setting because of potential bugs, glitches, lack of performance, safety points or different issues.