modelstack 1.0.0

ModelStack - Lightweight Local Model Registry

ModelStack

ModelStack is a lightweight, local-first model registry designed for data scientists and machine learning engineers. It allows efficient storage, versioning, and retrieval of ML models along with their associated metadata, without requiring any cloud services or external infrastructure.

Features

• Local registry to manage models directly on your machine
• Command Line Interface (CLI) for intuitive interaction
• Version control for models
• Support for PyTorch, Joblib, and Pickle formats
• Metadata tracking (accuracy, metrics, timestamp, etc.)

Installation

To install ModelStack locally, run:

pip install modelstack

Usage

After installation, you can use the modelstack CLI.

To register a model in db

modelstack register "name_of_model metadata"

To list all models in db

modelstack list

To delete a model

modelstack delete "name_of_model"

To delete all

modelstack delete --all