I'm not following why the ground_truth_key in AzureML's text generation pipeline component is a required argument. I understand it makes sense in summarization, translation, question_answering scenarios, but for text generation, which is what I'm using it for, just the input field should suffice.
What am I missing?