How was the valid and test split created?
#6
by
Parallaxixs
- opened
Hi, thank you for sharing the cleaned Pile dataset.
One question that I want to ask is how do you create the valid and test split, is it sampled from the cleaned Pile train set or is it from the original test and valid split of the Pile dataset?
It's from the original test and val splits. Each split was cleaned separately.
When I try loading the test split with:
load_dataset(
"monology/pile-uncopyrighted", split="test", trust_remote_code=True
)
I hit the error: ValueError: Unknown split "test". Should be one of ['train'].
How can I fix this?
If I try without the split="test", then the only split I see is "train"