metadata
dataset_info:
features:
- name: input_ids
sequence: int64
- name: seq_idx
dtype: int64
splits:
- name: train
num_bytes: 234577200
num_examples: 14300
- name: validation
num_bytes: 32808000
num_examples: 2000
download_size: 58650299
dataset_size: 267385200
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
language:
- en
This dataset has been created as an artefact of the paper Causal Estimation of Memorisation Profiles (Lesci et al., 2024).
More info about this dataset in the related collection Memorisation-Profiles.
Sample from the Pile (EleutherAI/pile-deduped-pythia-preshuffled) used in the experiments. The unique sequence identified (seq_idx) is simply the order of the sequence in the Pile. The dataset is already tokenized.