A developer wants to bring a custom algorithm to Amazon SageMaker by packaging it in a Docker image. How should the container be prepared so SageMaker can start the training job correctly?

SageMaker expects the training program to be defined as an ENTRYPOINT named train within the Docker container. When SageMaker launches the container for a training job, it executes this ENTRYPOINT. Incorrect options: Editing bashprofile is not the standard or reliable method for SageMaker to execute a training program. SageMaker doesn't necessarily run an interactive shell that would source bashprofile. Using CMD alone is insufficient because CMD provides default arguments for an ENTRYPOINT or executes a command if no ENTRYPOINT is defined. SageMaker specifically looks for an ENTRYPOINT named train. Copying the program to /opt/ml/train only places the file; it doesn't instruct the container to execute it. SageMaker expects the execution command to be part of the container's definition.