jellyml-lightning 1.0.0

A plugin to jelly-fill lightning checkpoints

JellyMl

JellyML is an open-source tool (python API and command line) for effortlessly embedding a snapshot of your code | into a checkpoint of a pytorch model. Learn more at jellyml.com

Structure of the jellyml repository

(Note that the jellyml repository is a monorepo. If you are reading this from the python package source code, go to github.com/mmulet/jellyml to see the whole repository)

• jellyml is the source for the python package
• jellyml-lightning is the source for pytorch lightning plugin
• client is the source for the website
• dev_server is the source for the development server of the website

Build

jellyml

1. Make a venv

python3 -m venv venv
# activate the venv ( depends on your shell and OS)
# see https://docs.python.org/3/library/venv.html
# bash
source venv/bin/activate

2. Install build

pip install build

3. Build the package

cd jellyml
python -m build
pip install dist/jellyml-0.0.1-py3-none-any.whl

jellyml-lightning

1. Follow the directions for building and installing jellyml. jellyml is a dependency of jellyml-lightning.
2. Build the package

cd jellyml-lightning
python -m build
pip install dist/jellyml-lightning-0.0.1-py3-none-any.whl

Website

Build the website

cd client
npm install .
cd ../dev_server
npm install .
npm run build

Dev the website

cd client
npm install .
cd ../dev_server
npm install .
npm run build

Tests

Located in the source files in src/jellyml. Have the prefix test_. Run them as a module

cd src;
python3 -m jellyml.test_all