vdt/sematic

An open-source ML pipeline development toolkit

Prototype-to-production ML in days not weeks

Hi 👋

Sematic is an open-source development toolkit to help Data Scientists and Machine Learning (ML) Engineers prototype and productionize ML pipelines in days not weeks.

Sematic is based on experience working at leading tech companies.

Find our docs at docs.sematic.dev, and join us on Discord.

Sematic helps you

• Develop and run ML pipelines using native Python functions, no new DSL to learn
• Monitor, visualize, and track all inputs and outputs of all pipeline steps in a slick UI
• Collaborate with your team to keep the discussion close to the pipeline as opposed to scattered elsewhere
• Execute your pipelines locally or in your cloud
• [soon] Clone/re-run your pipelines with different inputs/configs
• [soon] Schedule your pipelines to keep your models fresh and relevant

Sematic is an alternative to other ML pipeline tools, such as KubeFlow Pipelines.

Installation

Instal Sematic with

$ pip install sematic

Usage

Start the app locally with

$ sematic start

Then run an example pipeline with

$ sematic run examples/mnist/pytorch

Create a new boilerplate project

$ sematic new my_new_project

Or from an existing example:

$ sematic new my_new_project --from examples/mnist/pytorch

Then run it with

$ python3 -m my_new_project

See our docs at docs.sematic.dev, and join us on Discord.

Contribute

See our Contributor guide at docs.sematic.dev.