The Pusher component is used to push a validated model to a deployment target during model training or re-training. Before the deployment, Pusher relies on one or more blessings from other validation components to decide whether to push the model or not.
- Evaluator blesses the model if the new trained model is "good enough" to be pushed to production.
- (Optional but recommended) InfraValidator blesses the model if the model is mechanically servable in a production environment.
A Pusher component consumes a trained model in SavedModel format, and produces the same SavedModel, along with versioning metadata.
Using the Pusher Component
A Pusher pipeline component is typically very easy to deploy and requires little customization, since all of the work is done by the Pusher TFX component. Typical code looks like this:
from tfx import components
...
pusher = components.Pusher(
model=trainer.outputs['model'],
model_blessing=evaluator.outputs['blessing'],
infra_blessing=infra_validator.outputs['blessing'],
push_destination=pusher_pb2.PushDestination(
filesystem=pusher_pb2.PushDestination.Filesystem(
base_directory=serving_model_dir)
)
)