operatorai/modelstore

What it solves

It eliminates the boilerplate code required to save, version, and manage machine learning models across different storage backends. It provides a standardized way to handle model artifacts without needing a dedicated tracking server.

How it works

modelstore acts as an open-source model registry that interfaces with various storage types (such as AWS S3, GCP Buckets, Azure Blob Storage, MinIO, or local filesystems). It allows users to upload models from a wide range of ML libraries, automatically versions them upon upload, and enables loading them directly back into memory or downloading them via a command-line tool.

Who it’s for

Machine learning engineers and data scientists who need a simple, serverless way to version and store their models across multiple cloud providers or local environments.

Highlights

• Serverless Registry: No tracking server required; works directly with cloud buckets or local disks.
• Multi-Cloud Support: Compatible with AWS, GCP, and Azure.
• Broad Library Support: Works with PyTorch, TensorFlow, Scikit-Learn, XGBoost, Hugging Face Transformers, and many others.
• Lifecycle Management: Supports organizing models by domains and managing their states.