You have a trained and registered machine learning model. You create a batch inference pipeline that uses this model to generate predictions from multiple data files. You plan to publish this batch inference pipeline as a service that can be scheduled to run nightly.

Which compute target should you select for the inference service?