ModelBox
An extensible AI Metadata and Model Operations Service
AI Metadata Store
Log, Search and Compare metadata from experiments, models and checkpoints. Support for logging metrics and events from various services involved in training and deployment of models.
Model Registry
Track and store trained models or upload them to a built in artifact server. Automatically apply recipes to transform models to representations understood by inference platforms and runtimes.
Continuous Model Evaluation
Measure performance metrics such as throughput and accuracy of models. Write deployment checks based on metrics to automatically tag models to promote to production.