Datasets:

togethercomputer
/

RedPajama-Data-1T

text string	meta string	red_pajama_subset string
\section{Introduction} Let $G$ be a simple undirected graph with the \textit{vertex set} $V(G)$ and the \textit{edge set} $E(G)$. A vertex with degree one is called a \textit{pendant vertex}. The distance between the vertices $u$ and $v$ in graph $G$ is denoted by $d_G(u,v)$. A cycle $C$ is called \textit{chordless} if...	{'timestamp': '2016-07-19T02:04:55', 'yymm': '1607', 'arxiv_id': '1607.04768', 'language': 'en', 'url': 'https://arxiv.org/abs/1607.04768'}	arxiv
\section{Principle of nano strain-amplifier} \begin{figure*}[t!] \centering \includegraphics[width=5.4in]{Fig1} \vspace{-0.5em} \caption{Schematic sketches of nanowire strain sensors. (a)(b) Conventional non-released and released NW structure; (c)(d) The proposed nano strain-amplifier and its simplified...	{'timestamp': '2016-07-18T02:07:38', 'yymm': '1607', 'arxiv_id': '1607.04531', 'language': 'en', 'url': 'https://arxiv.org/abs/1607.04531'}	arxiv
\section{Introduction}\label{intro} Gas has a fundamental role in shaping the evolution of galaxies, through its accretion on to massive haloes, cooling and subsequent fuelling of star formation, to the triggering of extreme luminous activity around super massive black holes. Determining how the physical state of gas ...	{'timestamp': '2016-08-09T02:04:29', 'yymm': '1607', 'arxiv_id': '1607.04828', 'language': 'en', 'url': 'https://arxiv.org/abs/1607.04828'}	arxiv
\section{Introduction} Given $\rho>0$, we consider the problem \begin{equation}\label{eq:main_prob_U} \begin{cases} -\Delta U + \lambda U = \|U\|^{p-1}U & \text{in }\Omega,\smallskip\\ \int_\Omega U^2\,dx = \rho, \quad U=0 & \text{on }\partial\Omega, \end{cases} \end{equation} where $\Omega\subset{\mathbb{R}}^N$ is a ...	{'timestamp': '2016-07-18T02:07:27', 'yymm': '1607', 'arxiv_id': '1607.04520', 'language': 'en', 'url': 'https://arxiv.org/abs/1607.04520'}	arxiv
"\\section{Introduction}\n\\label{sec:intro}\nDespite the immense popularity and availability of onl(...TRUNCATED)	"{'timestamp': '2016-07-20T02:04:30', 'yymm': '1607', 'arxiv_id': '1607.04648', 'language': 'en', 'u(...TRUNCATED)	arxiv
"\\section{Introduction}\n\nThe wave-particle duality is an alternative statement of the complementa(...TRUNCATED)	"{'timestamp': '2017-01-26T02:07:48', 'yymm': '1607', 'arxiv_id': '1607.04617', 'language': 'en', 'u(...TRUNCATED)	arxiv
"\\section{#1}\\setcounter{equation}{0}}\n\\newcommand{\\subsect}[1]{\\subsection{#1}}\n\\renewcomma(...TRUNCATED)	"{'timestamp': '2016-07-19T02:06:15', 'yymm': '1607', 'arxiv_id': '1607.04824', 'language': 'en', 'u(...TRUNCATED)	arxiv
"\\section{Introduction}\n\\label{sec:introduction} \nIn recent years there has been a resurgen(...TRUNCATED)	"{'timestamp': '2016-12-01T02:04:39', 'yymm': '1607', 'arxiv_id': '1607.04724', 'language': 'en', 'u(...TRUNCATED)	arxiv
"\\section{Introduction} \\label{sec:intro}\n\nThe production of hadrons and jets at a future Electr(...TRUNCATED)	"{'timestamp': '2016-10-28T02:00:49', 'yymm': '1607', 'arxiv_id': '1607.04921', 'language': 'en', 'u(...TRUNCATED)	arxiv
"\\section{Introduction}\n\\label{sec:intro}\n\n\nCompact stars have a large number of pulsation mod(...TRUNCATED)	"{'timestamp': '2016-07-19T02:02:47', 'yymm': '1607', 'arxiv_id': '1607.04707', 'language': 'en', 'u(...TRUNCATED)	arxiv

End of preview.

Getting Started

The dataset consists of 2084 jsonl files. You can download the dataset using HuggingFace:

from datasets import load_dataset
ds = load_dataset("togethercomputer/RedPajama-Data-1T")

Or you can directly download the files using the following command:

wget 'https://data.together.xyz/redpajama-data-1T/v1.0.0/urls.txt'
while read line; do
    dload_loc=${line#https://data.together.xyz/redpajama-data-1T/v1.0.0/}
    mkdir -p $(dirname $dload_loc)
    wget "$line" -O "$dload_loc"
done < urls.txt

After downloading the files, you can load the dataset from disk by setting the RED_PAJAMA_DATA_DIR environment variable to the directory containing the files:

import os
from datasets import load_dataset
os.environ["RED_PAJAMA_DATA_DIR"] = "/path/to/download"
ds = load_dataset("togethercomputer/RedPajama-Data-1T")

A smaller 1B-token sample of the dataset can be found here.

A full set of scripts to recreate the dataset from scratch can be found here.

Dataset Summary

RedPajama is a clean-room, fully open-source implementation of the LLaMa dataset.

Dataset	Token Count
Commoncrawl	878 Billion
C4	175 Billion
GitHub	59 Billion
ArXiv	28 Billion
Wikipedia	24 Billion
StackExchange	20 Billion
Total	1.2 Trillion

Languages

Primarily English, though the Wikipedia slice contains multiple languages.

Dataset Structure

The dataset structure is as follows:

{
    "text": ...,
    "meta": {"url": "...", "timestamp": "...", "source": "...", "language": "...", ...},
    "red_pajama_subset": "common_crawl" | "c4" | "github" | "arxiv" | "wikipedia" | "stackexchange"
}

Dataset Creation

This dataset was created to follow the LLaMa paper as closely as possible to try to reproduce its recipe.

Source Data

Commoncrawl

We download five dumps from Commoncrawl, and run the dumps through the official cc_net pipeline. We then deduplicate on the paragraph level, and filter out low quality text using a linear classifier trained to classify paragraphs as Wikipedia references or random Commoncrawl samples.

C4

C4 is downloaded from Huggingface. The only preprocessing step is to bring the data into our own format.

GitHub

The raw GitHub data is downloaded from Google BigQuery. We deduplicate on the file level and filter out low quality files and only keep projects that are distributed under the MIT, BSD, or Apache license.

Wikipedia

We use the Wikipedia dataset available on Huggingface, which is based on the Wikipedia dump from 2023-03-20 and contains text in 20 different languages. The dataset comes in preprocessed format, so that hyperlinks, comments and other formatting boilerplate has been removed.

Gutenberg and Books3

Defunct: The 'book' config is defunct and no longer accessible due to reported copyright infringement for the Book3 dataset contained in this config.

ArXiv

ArXiv data is downloaded from Amazon S3 in the arxiv requester pays bucket. We only keep latex source files and remove preambles, comments, macros and bibliographies.

Stackexchange

The Stack Exchange split of the dataset is download from the Internet Archive. Here we only keep the posts from the 28 largest sites, remove html tags, group the posts into question-answer pairs, and order answers by their score.

SHA256 Checksums

SHA256 checksums for the dataset files for each data source are available here:

https://data.together.xyz/redpajama-data-1T/v1.0.0/sha256/arxiv_SHA256SUMS.txt
https://data.together.xyz/redpajama-data-1T/v1.0.0/sha256/c4_SHA256SUMS.txt
https://data.together.xyz/redpajama-data-1T/v1.0.0/sha256/common_crawl_SHA256SUMS.txt
https://data.together.xyz/redpajama-data-1T/v1.0.0/sha256/github_SHA256SUMS.txt
https://data.together.xyz/redpajama-data-1T/v1.0.0/sha256/stackexchange_SHA256SUMS.txt
https://data.together.xyz/redpajama-data-1T/v1.0.0/sha256/wikipedia_SHA256SUMS.txt

To cite RedPajama, please use:

@software{together2023redpajama,
  author = {Together Computer},
  title = {RedPajama: An Open Source Recipe to Reproduce LLaMA training dataset},
  month = April,
  year = 2023,
  url = {https://github.com/togethercomputer/RedPajama-Data}
}