Inference logger#

Creation#

hsml.inference_logger.InferenceLogger(kafka_topic=DEFAULT, mode="ALL", **kwargs)

Configuration of an inference logger for a predictor.

Arguments

kafka_topic hsml.kafka_topic.KafkaTopic | dict | hopsworks_common.constants.Default | None: Kafka topic to send the inference logs to. By default, a new Kafka topic is configured.
mode str | None: Inference logging mode. (e.g., NONE, ALL, PREDICTIONS, or MODEL_INPUTS). By default, ALL inference logs are sent.

Returns

InferenceLogger. Configuration of an inference logger.

Inference loggers can be accessed from the predictor metadata objects.

predictor.inference_logger

Predictors can be found in the deployment metadata objects (see Predictor Reference). To retrieve a deployment, see the Deployment Reference.

Kafka topic to send the inference logs to.

Inference logging mode ("NONE", "ALL", "PREDICTIONS", or "MODEL_INPUTS").

InferenceLogger.describe()

Print a description of the inference logger

InferenceLogger.to_dict()