minhnguyent546/fineweb-edu-10BT-for-modern-gpt

Dataset Summary

This is a tokenized Fineweb-Edu (10BT subset) using GPT2 tokenizer (via tiktoken) for pre-training GPT2 model. Data is divided into shards (.npy files) for easier to load with PyTorch IterableDataset.

Each .npy file can be loaded with numpy.load('file_name.npy').

Split	# Documents	# Shards	# Tokens
train	9,575,380	99	9,853,755,380 (9.85B)
val	96,721	1	100,233,964 (0.1B)
Total	9,672,101	100	9,953,989,344 (9.95B)

Example of usage

uvx hf download minhnguyent546/fineweb-edu-10BT-for-gpt2 \
  --repo-type dataset \
  --local-dir fineweb_edu_10bt

import numpy as np

tokens = np.load("fineweb_edu_10bt/train/fineweb_edu_10bt_train_000000-of-000099.npy")

print(tokens.shape)
# (100000000,)

print(tokens.dtype)
# uint16

print(tokens[:20])
# [50256    12  3363    11   428   318   257   922   640   284  4618  6868
#  8701  9403   287   262  2323    13   921   743]