I remember in the previous jupyter mode, we can directly promote the saved files to a Floydhub dataset. I don't know the difference in backend engineering between "using floyd data in workspace" and "create data from output in a job." Maybe the latter is faster for transforming files to Floydhub dataset in your server?
The create data from output feature is faster because it requires fewer steps. More in detail, once the output is already finalized in the backend (when the Job is done), it's really fast to create a copy of this as a dataset rather than compress, upload and wait for the data finalization.
A similar (fast) experience can also be achieved with the Modify Data, here are the steps to follow:
- Use the Modify Data on a dataset
- Download / Create the dataset from this workspace
- Stop the workspace and you will find the data as a new dataset version
Note: You can still use the create data from output for Job execution.
Let me know if it is useful for you.