secretnote 0.0.4

Notebook suite for SecretFlow

SecretNote

Notebook suite for SecretFlow.

This is currently a developer preview.

API may change without prior notice. No guarantee is made about the security, correctness, performance, or usefulness of this feature.

• SecretNote
  • Install
  • Start
  • Features
    • Using the profiler
  • FAQ
    • Environment and dependency versioning

Install

Please read our note about dependencies first.

We recommend installing in a clean Python environment.

python --version
# Python 3.8.18
python -m venv .venv
source .venv/bin/activate
python -m pip install 'secretflow==1.2.0b0' 'secretnote==0.1.0.dev0'

Start

In your terminal, run:

secretnote

Or:

python -m secretnote.server

Commonly-used Jupyter command line options are supported. For example, to change the port:

secretnote --ServerApp.port 8889

Features

Using the profiler

from secretnote.instrumentation.sdk import create_profiler

with create_profiler() as profiler:
    # SecretFlow code here
    weights = alice(...)(...)
    bias = bob(...)(...)
    result = spu(...)(...)

profiler.visualize()

FAQ

Environment and dependency versioning

For best results, install this project in a clean Python environment.

SecretFlow has many platform-specific dependencies. As this project is in early stage, we do not yet guarantee that this project will work in all environments, or even that installation will succeed.

Supported Python versions match those of SecretFlow, currently only Python 3.8.

To see versions of commonly used dependencies, run python -m secretnote.utils.version:

> python -m secretnote.utils.version
environment info
├── system
│   └── platform = macosx-11.0-arm64
├── python
│   ├── version = 3.8.18
│   ├── path = ...
│   └── implementation = cpython
└── packages
    ├── grpc = 1.56.2
    ├── heu = 0.4.4b0
    ├── ipykernel = 6.27.0
    ├── ipython = 8.12.3
    ├── jaxlib = 0.4.12
    ├── jupyter_server = 2.10.1
    ├── pydantic = 1.10.13
    ├── ray = 2.6.3
    ├── secretflow = 1.2.0b0
    ├── secretnote = 0.1.0-dev.0
    ├── spu = 0.6.0b0
    ├── tensorflow = unavailable
    └── torch = 2.1.1

Issues usually arise when the versions of grpc, jaxlib, ray, and/or secretflow are not compatible with each other. Additionally, you may encounter issues on macos-*-arm64 (Apple Silicon).