Merlin release v0.8.0-alpha

Enable users to deploy a custom transformer for pre/post-processing alongside their models.

Merlin release v0.7.0

This is the first public release of Merlin, a platform for managing, deploying, and serving machine learning models.

Features:

• Deployment of standard models (Tensorflow, XGBoost, SKLearn, PyTorch), user-defined models (PyFunc), and batch prediction job
• Multi-environment deployment
• Traffic routing to the running model version using Istio
• UI for model and batch prediction job management and deployment
• Python SDK library to interacting with Merlin API
• CLI to provides a way to deploy and undeploy Merlin model
• Authentication using Google OAuth and authorization across resources using Keto
• Model logging in the Merlin UI
• ML model performance metrics monitoring with Grafana
• Alerting for a model endpoint using Prometheus
• Resources configuration including the number of minimum and maximum replicas, CPU, and memory

Merlin release v0.7.0-alpha

Make Batch Job monitoring URL configurable from env var (#9)