In the CUSTOM tier for Google Cloud Machine Learning Engine (now called AI Platform), you have the flexibility to specify the number of workers and parameter servers for your training jobs. This allows for customization of the infrastructure to match the scale of your machine learning workloads.

• Masters: The master node is automatically managed by Google Cloud in the custom tier and does not require user specification.
• Workers: These are responsible for executing the training tasks.
• Parameter Servers: They manage and store the parameters of the machine learning model during the training process.

Thus, the correct configuration in the CUSTOM tier involves specifying the count of workers and parameter servers, excluding masters. For more details, visit Google Cloud's documentation on job configuration parameters.