pipelines/samples/contrib/seldon

Seldon Samples

Seldon is a model serving solution that supports multiple deployment strategies and provides out of the box observability. These examples are an overview of how seldon can be used with pipelines:

• iris_storagebucket.py is the simplest case and best place to start. It shows how seldon can serve a packaged model from a storage bucket URI.
• mabdeploy_seldon.py shows how components can be assembled into inference graphs such as multi-armed bandits.
• mnist_tf.py shows how seldon can be used with custom-built serving images where a model is baked into the docker image.
• mnist_tf_volume.py shows a model being stored in a volume and served using a serving image.

See the seldon docs for other uses such as autoscaling, canaries and tracing