Currently, evaluating a Moodle learning analytics model must be done from the CLI for all but the smallest sites. However, many hosting situations do not allow access to the CLI. This means models cannot be evaluated on many sites before use, which is against recommended practice.
The evaluation process should not be run from the web browser because it takes an extended period of time, and when the web session ends, the evaluation process halts.
Like the model training process, the model evaluation process could be managed using the Scheduled Tasks feature. Marking a model for evaluation could identify it to the scheduled evaluation task when it runs. The task can run for a set period of time, then exit, and run again as scheduled. (The evaluation process can be run in multiple sessions this way-- it saves its work as it goes.)