How to use multiple datasets for PI05 training

[sunmoon2018]

Currently, by setting a custom dataset storage path in the repo_id of TrainConfig, what if multiple sets of dataset paths are configured for pi05 training?

Look forward to your reply！