| Sample from the Pile (`EleutherAI/pythia_pile_idxmaps`) used to compute the memorisation profiles. | |
| The unique sequence identified (`uid`) is simply the order of the sequence in the Pile. The dataset is already tokenized. | |
| Each file contains the same `uid` (for simplicity taken from `pietrolesci/pile-deduped-subset`), but since different seeds are used for data shuffling, the data are different. |