Fix NonMatchingSplitsSizesError

✅ Solution

NonMatchingSplitsSizesError usually arises when the manually defined split proportions don't align with the actual number of examples discovered during dataset generation. Ensure that the sum of `split` argument values in `datasets.SplitGenerator` equals the true number of examples, or avoid specifying `split` entirely to let the library automatically infer the split sizes. If a dataset is sharded, verify that downloaded shards are complete and not corrupted, causing an incorrect example count.