Update README with links, citation
Browse files
README.md
CHANGED
@@ -3,14 +3,13 @@ task_categories:
|
|
3 |
- text-generation
|
4 |
language:
|
5 |
- en
|
6 |
-
pretty_name: SlimPajama
|
7 |
-
license: apache-2.0
|
8 |
---
|
9 |
## Getting Started
|
10 |
|
11 |
-
|
12 |
|
13 |
-
You can download the dataset using
|
14 |
```python
|
15 |
from datasets import load_dataset
|
16 |
ds = load_dataset("cerebras/SlimPajama-627B")
|
@@ -18,39 +17,32 @@ ds = load_dataset("cerebras/SlimPajama-627B")
|
|
18 |
|
19 |
## Background
|
20 |
|
21 |
-
|
22 |
-
|
23 |
-
Applying [MinHashLSH](http://infolab.stanford.edu/~ullman/mmds/book0n.pdf) deduplication to Trillion token datasets like RedPajama was not possible with off-the-shelf open-source code. We made several optimizations to existing solutions to produce infrastructure that can perform MinHashLSH deduplication on Trillion token datasets in a distributed, multi-threaded and memory efficient fashion. Today we are open-sourcing this infrastructure to enable the community to develop higher quality, deduplicated datasets in the future.
|
24 |
-
|
25 |
-
|
26 |
-
### Our observations of the original data set
|
27 |
-
|
28 |
-
1. RedPajama contains a portion of partially downloaded files.
|
29 |
-
2. Some (~2%) of the examples contain empty text. They were downloaded correctly, but do not have useful content that a model can be trained on.
|
30 |
-
3. There are many (~50%) duplicates in the data. The RedPajama team deduplicated some sources (Books, GitHub, Commoncrawl), but did not deduplicate all sources.
|
31 |
|
|
|
32 |
|
33 |
### Our contributions
|
34 |
|
35 |
-
1. SlimPajama 627B – the largest deduplicated, multi-corpora, open dataset for LLM training. We release it under the Apache 2.0 license.
|
36 |
-
2. Releasing validation and test sets,
|
37 |
-
3. Library of methods to replicate or pre-process from scratch other datasets. To the best of our knowledge these are the first open
|
38 |
-
|
39 |
|
40 |
-
The full set of scripts to recreate the dataset from the original RedPajama dataset
|
41 |
|
42 |
## Dataset Summary
|
43 |
|
|
|
|
|
44 |
#### Comparison of dataset features
|
45 |
-
|
|
46 |
-
| --------------- |
|
47 |
-
| SlimPajama | 627B
|
48 |
-
| RedPajama | 1.21T
|
49 |
-
| RefinedWeb-600B | 600B
|
50 |
-
| RefinedWeb-5T | 5T
|
51 |
-
| LLaMA | 1.4T
|
52 |
-
| MPT | 1T
|
53 |
-
| MassiveText | 1.4T
|
54 |
|
55 |
|
56 |
#### Document low-length filter rates
|
@@ -66,18 +58,18 @@ The full set of scripts to recreate the dataset from the original RedPajama data
|
|
66 |
| StackExchange | 0.32% |
|
67 |
| Total | 1.86% |
|
68 |
|
69 |
-
####
|
70 |
|
71 |
-
| Data source |
|
72 |
-
| ------------- | ----------------------
|
73 |
-
| Commoncrawl | 63.76%
|
74 |
-
| C4 | 6.85%
|
75 |
-
| GitHub | 46.16%
|
76 |
-
| Books | 2.01%
|
77 |
-
| ArXiv | 0.06%
|
78 |
-
| Wikipedia | 2.24%
|
79 |
-
| StackExchange | 0.20%
|
80 |
-
| Total | 49.60%
|
81 |
|
82 |
#### Data source proportions for SlimPajama and RedPajama
|
83 |
|
@@ -110,7 +102,7 @@ The dataset consists of jsonl files, with structure as follows:
|
|
110 |
|
111 |
### Dataset Creation
|
112 |
|
113 |
-
SlimPajama was created by cleaning and deduplicating the [RedPajama dataset from Together Computer](https://github.com/togethercomputer/redpajama-data) via MinHashLSH. RedPajama is an open-source reproduction of the [
|
114 |
|
115 |
|
116 |
### Source Data
|
@@ -121,12 +113,13 @@ The data sources composing RedPajama are explained in [its model card](https://h
|
|
121 |
To cite SlimPajama, please use:
|
122 |
|
123 |
```
|
124 |
-
@
|
125 |
-
author = {
|
126 |
-
title = {SlimPajama: A 627B token cleaned and deduplicated version of RedPajama},
|
127 |
month = June,
|
128 |
year = 2023,
|
129 |
-
|
|
|
130 |
}
|
131 |
```
|
132 |
|
|
|
3 |
- text-generation
|
4 |
language:
|
5 |
- en
|
6 |
+
pretty_name: SlimPajama-627B
|
|
|
7 |
---
|
8 |
## Getting Started
|
9 |
|
10 |
+
The dataset consists of 59166 jsonl files. It is a cleaned and deduplicated version of [Together Computer's RedPajama](https://github.com/togethercomputer/redpajama-data). Check out our [blog post](https://www.cerebras.net/blog/slimpajama-a-627b-token-cleaned-and-deduplicated-version-of-redpajama) explaining our methods.
|
11 |
|
12 |
+
You can download the dataset using Hugging Face datasets:
|
13 |
```python
|
14 |
from datasets import load_dataset
|
15 |
ds = load_dataset("cerebras/SlimPajama-627B")
|
|
|
17 |
|
18 |
## Background
|
19 |
|
20 |
+
Today we are releasing SlimPajama – the largest extensively deduplicated, multi-corpora, open-source dataset for training large language models. SlimPajama was created by cleaning and deduplicating the 1.2T token RedPajama dataset from Together. By filtering out low quality data and duplicates, we were able to remove 49.6% of bytes, slimming down the dataset from 1210B to 627B tokens. We believe SlimPajama offers the highest quality and most compute efficient data to train on for runs up to 627B tokens. When upsampled, we expect SlimPajama to perform equal to or better than RedPajama-1T when training at trillion token scale.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
+
In addition to the data, we are also releasing the tools we built to create SlimPajama. Applying [MinHashLSH](http://infolab.stanford.edu/~ullman/mmds/book0n.pdf) deduplication to trillion token datasets like RedPajama was not possible with off-the-shelf open-source code. We made several improvements to existing solutions to produce an infrastructure that can perform MinHashLSH deduplication on trillion token datasets in a distributed, multi-threaded, and memory efficient fashion. Today we are open-sourcing this infrastructure to enable the community to easily create higher quality, extensively deduplicated datasets in the future.
|
23 |
|
24 |
### Our contributions
|
25 |
|
26 |
+
1. SlimPajama 627B – the largest extensively deduplicated, multi-corpora, open dataset for LLM training. We release it under the Apache 2.0 license.
|
27 |
+
2. Releasing validation and test sets, 500M tokens each, which has been decontaminated against the training data.
|
28 |
+
3. Library of methods to replicate or pre-process from scratch other datasets. To the best of our knowledge these are the first open-source tools to enable cleaning and MinHashLSH deduplication of text data at trillion token scale.
|
|
|
29 |
|
30 |
+
The full set of scripts to recreate the dataset from the original RedPajama dataset will be available on the Cerebras github. A deeper explanation of our cleaning and deduplication process can be found in the [SlimPajama blog post](https://www.cerebras.net/blog/slimpajama-a-627b-token-cleaned-and-deduplicated-version-of-redpajama).
|
31 |
|
32 |
## Dataset Summary
|
33 |
|
34 |
+
The [latest research](https://arxiv.org/abs/2306.01116) has shown that data quality is as important as data quantity. While training on more than one data epoch can be beneficial, this should be a choice rather than a side-effect of duplicates in the dataset. We decided to extensively deduplicate RedPajama to produce a dataset with higher information density. This means when using SlimPajama, you can achieve higher accuracy with the same compute budget when compared to other datasets.
|
35 |
+
|
36 |
#### Comparison of dataset features
|
37 |
+
| Data source | Tokens | Open Source | Curated Data Sources | Deduplication Level |
|
38 |
+
| --------------- | ------- | ----------- | -------------------- | ------------------- |
|
39 |
+
| SlimPajama | **627B**| Yes | Yes | **Extensive** |
|
40 |
+
| RedPajama | 1.21T | Yes | Yes | Partial |
|
41 |
+
| RefinedWeb-600B | 600B | Yes | No | **Extensive** |
|
42 |
+
| RefinedWeb-5T | **5T** | No | No | **Extensive** |
|
43 |
+
| LLaMA | 1.4T | No | Yes | Partial |
|
44 |
+
| MPT | 1T | No | Yes | Unknown |
|
45 |
+
| MassiveText | 1.4T | No | Yes | **Extensive** |
|
46 |
|
47 |
|
48 |
#### Document low-length filter rates
|
|
|
58 |
| StackExchange | 0.32% |
|
59 |
| Total | 1.86% |
|
60 |
|
61 |
+
#### Data source byte deduplication rates
|
62 |
|
63 |
+
| Data source | Byte deduplication rate |
|
64 |
+
| ------------- | ---------------------- |
|
65 |
+
| Commoncrawl | 63.76% |
|
66 |
+
| C4 | 6.85% |
|
67 |
+
| GitHub | 46.16% |
|
68 |
+
| Books | 2.01% |
|
69 |
+
| ArXiv | 0.06% |
|
70 |
+
| Wikipedia | 2.24% |
|
71 |
+
| StackExchange | 0.20% |
|
72 |
+
| Total | 49.60% |
|
73 |
|
74 |
#### Data source proportions for SlimPajama and RedPajama
|
75 |
|
|
|
102 |
|
103 |
### Dataset Creation
|
104 |
|
105 |
+
SlimPajama was created by cleaning and deduplicating the [RedPajama dataset from Together Computer](https://github.com/togethercomputer/redpajama-data) via MinHashLSH. RedPajama is an open-source reproduction of the [LLaMA](https://arxiv.org/abs/2302.13971) data collection methodology.
|
106 |
|
107 |
|
108 |
### Source Data
|
|
|
113 |
To cite SlimPajama, please use:
|
114 |
|
115 |
```
|
116 |
+
@misc{cerebras2023slimpajama,
|
117 |
+
author = {Soboleva, Daria and Al-Khateeb, Faisal and Myers, Robert, Steeves, Jacob R and Hestness, Joel and Dey, Nolan},
|
118 |
+
title = {{SlimPajama: A 627B token cleaned and deduplicated version of RedPajama}},
|
119 |
month = June,
|
120 |
year = 2023,
|
121 |
+
howpublished = {\url{https://www.cerebras.net/blog/slimpajama-a-627b-token-cleaned-and-deduplicated-version-of-redpajama}},
|
122 |
+
url = {https://huggingface.co/datasets/cerebras/SlimPajama-627B},
|
123 |
}
|
124 |
```
|
125 |
|