metadata
dataset_info:
features:
- name: text
dtype: string
splits:
- name: train
num_bytes: 101351288
num_examples: 20000
download_size: 42778874
dataset_size: 101351288
Dataset Card for "c4-code-10k"
10K elements of C4 and 10K elements of code parrot clean (Python code).
Note that these are the datasets used to train my interpretability-friendly models, but is not of the correct mixture. Those models were trained on 83% C4 and 17% Python Code (ish) by tokens. This dataset has 10K strings of each, and by tokens is about 22M of code and 5M of C4 (code is longer and harder to compress!)