XSum test set has different sample ordering from the one downloaded by huggingface datasets' load_dataset
Closed this issue · 1 comments
chris-opendata commented
I found that the XSum test set from this research project has different sample ordering from the one I downloaded using huggingface datasets' load_dataset despite the total number of samples being same. I am wondering if there are any rules that can be applied to align them in terms of index mapping.
Regards
Chris
chris-opendata commented
Found that sample ids in dataset. So, they can be used to match up.