Datasets:
Indexing problem in the "mrpc".
For the train and validation split in the "mrpc" dataset, the "idx" column does not increase linearly. For instance, in the train dataset index jumps from 8 to 10. I am not sure anyone uses the "idx" column, but for my research, it is essential, and I use it.
Please note that this behavior is expected: each sample is either in the "train" or in the "validation" split.
Therefore, the sample with idx = 9 does not appear in the "train" split (https://huggingface.co/datasets/glue/viewer/mrpc/train) because it is included in the "validation" split (https://huggingface.co/datasets/glue/viewer/mrpc/validation): the first sample in the "validation" split has idx=9.
The same applies for the rest of samples in the "validation" split (with idx: 18, 25, 32,...).
I got this, but this is not the case for the rest of the datasets. They do not follow the same convention.
I agree, that is a matter of convention and each dataset chooses their own.
Maybe we could explicitly explain it in the dataset card.