Differences in input of deepar

[creamiracle]

Hey there, I just wanna ask a small question, "is there any differences between the input dimention for deepar?"
To make it more clear, I'll give an example here: Assume there are 10 time series, I have two kinds of way to build models:
1.make these 10 series A ListDataset, and use this ONE dataset as input and train deepar model, then use a for loop to predict each of them.
2.make 10 ListDatasets, use each of them to train model, then there are 10 models, and will do predict by itself.
The question here is, is there any differences between these two methods? If so, which one is much more better for deepar?
Thanks.
P.S. If you need more information, just let me know.