The machine learning dashboard includes a download link for training outputs. I think a user would likely expect that they could "download" a training run this way and then import it into another XNAT (or the same XNAT in a different project). However, even if the files downloaded were the appropriate ones for importing a model, the file structure definitely is not. That's because this is using the default XNAT resource API (like http://10.1.100.17/data/experiments/XNAT_E00033/resources/Results/files?format=zip). We would need a custom API to download the model in the appropriate file structure. As-is, when you import a model like this, you get what you see in the screenshot.