A retail company has an ecommerce application that uses ML to make product recommendations.

Deploy both model variants to an Amazon SageMaker AI serverless inference endpoint. In the endpoint configuration, set the weight for each model variant to 0.5. In the ecommerce application, invoke the inference endpoint.

Deploy each model variant to its own Amazon SageMaker AI serverless inference endpoint. In the ecommerce application, invoke each model variant endpoint directly by using a round robin strategy. Log variant invocations in Amazon CloudWatch Logs.

Deploy both variants to an Amazon SageMaker AI real-time inference endpoint. In the endpoint configuration, set the weight for each model variant to 0.5. In the ecommerce application, invoke the inference endpoint.

Deploy both variants to an Amazon SageMaker AI real-time inference endpoint. In the ecommerce application, invoke each model variant directly by using a round robin strategy. Log variant invocations in Amazon CloudWatch Logs.