MLOps tools for research and development.
Project description
MLOpus
A collection of MLOps tools for AI/ML/DS research and development.
Main features:
- Agnostic experiment tracking and model registry:
- Compatible with any "MLflow-like" provider through plugins.
- Search entities in MongoDB Query Language with predicate push-down to the MLflow provider.
- Local cache for artifacts and entity metadata.
- Offline mode to work with local cache only.
- Support for nested tags/params/metrics and JSON-encoded tags/params for non-scalar types.
- Not dependent on env vars, global vars or a single global active run.
Check the tutorials for a friendly walkthrough of (almost) everything you can do with MLOpus.
A minimal API reference is also available here.
Installation
Recommended software:
- Rclone CLI (required for artifact transfer from/to cloud storage)
Optional extras:
mlflow
: Enables support for the default MLflow plugin, which handles communication with open-source MLflow servers.search
: Enables searching entities with MongoDB query syntax
Using pip:
pip install mlopus[mlflow,search]
Using Poetry:
poetry add mlopus --extras "mlflow,search"
Using UV:
uv add mlopus --extra mlflow --extra search
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
mlopus-0.1.1.tar.gz
(43.4 kB
view hashes)