Klever Model Registry

English | 中文

Klever Model Registry is a Cloud Native ML model registry. Use Klever Model Registry in order to:

Manage your ML models
Version and deliver your ML models with the existing infrastructures
Keep track of ML models' hyperparameters and so on to help decision makers
Convert models between different formats ( e.g. TensorFlow SavedModel, ONNX )
Serve the model
Get the standalone executable program to deploy ML model inference services on edge devices/servers ( Coming Soon! )

Klever Model Registry's features:

Deploy with Docker and Kubernetes
Keep non-Invasive for your business
Manage ML models like Docker ( With the help of kleveross/ormb )
Reuse Harbor to store models, without any new infrastructure
Support model serving for models managed by MLflow
Extract models signatures for:
- SavedModel
- ONNX
- GraphDef
- NetDef
- Keras H5
- CaffeModel
- TorchScript
- MXNetParams
- PMML
Convert models from:
- MXNetParams to ONNX
- Keras H5 to SavedModel
- CaffeModel to NetDef
- NetDef to ONNX

See our official documentations for more information。

UI MockUp

Installation

Build the image

Clone:

$ git clone https://github.com/kleveross/klever-model-registry
$ cd klever-model-registry

Get the dependencies:

$ go mod tidy

Build:

$ make docker-build

Install

Please have a look at docs/installation.md.

If you want to trial quickly, you can run installation script as follow.

$ wget https://raw.githubusercontent.com/kleveross/klever-model-registry/master/scripts/installation/install-k8s-1.14.sh
$ bash install-k8s-1.14.sh <master-ip>

$ wget https://raw.githubusercontent.com/kleveross/klever-model-registry/master/scripts/installation/install-k8s-1.19.sh
$ bash install-k8s-1.19.sh <master-ip>

Community

klever-model-registry project is part of Klever, a Cloud Native Machine Learning platform.