Possibility to load multiple config for a single dataset

[jplu]

Hi,

It would be very nice to allow the possibility to load multiple config name for a single dataset. For example to train a multilingual model on two or more languages from the same dataset. Hence, to have a command line that looks like:

tner-train-search -m "bert-base-multilingual-cased" -c "output" -d "tner/multinerd" --dataset-name ['en','de'] 

This command will merge the en and de datasets to train the model on these both languages.

Do you think this would be feasible?

Thanks!