Is there a leak here

[LujianYao]

hello,
In src/utils/data_generation.py file
I find the default parameters of split_name='all'. I think it regards all data as train data. And I try to replace it as a specific split, but the result is very poor. Did I do something wrong or is there a leak here?

[Dshdarkness]

I received the same inquiry. According to the RISE data paper, the split name denotes a separate split method for ONE data, which means that different splited data may contain the same video clips.