Hi all,
I have a newbie question about clear-ml data. I have four data sources that get combined to train a model. I have put each of these datasets into clear ml so that I can track their versions, and then create the fifth 'combined' dataset using the ids from the others. It works great.

My question is the correct method to update that fifth dataset if one of the other datasets changes. Say for example I create a new version of dataset1 , what is the correct method for creating the updated version of combined_dataset ?

• Use the ids from each of the datasets the same way I did the first time ( Dataset.create(..., parent_datasets = [dataset1.id, dataset2.id, ...] ). In which case, will this actually be a second version of combined_dataset?
• Do the same as above, but include the previous version of the combined dataset id as a parent as well?
• Just do Dataset.create(..., parent_datasets = [combined_dataset.id]) and assume that clearml will take care of the rest?