BramVanroy commited on
Commit
0bf91c9
·
verified ·
1 Parent(s): 1bd8439

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +61 -94
README.md CHANGED
@@ -23,97 +23,64 @@ language:
23
  - nl
24
  - es
25
  configs:
26
- - config_name: CC-MAIN-2024-51
27
- data_files: data/CC-MAIN-2024-51/**/*.parquet
28
- - config_name: CC-MAIN-2024-51-af
29
- data_files:
30
- - split: train
31
- path: data/CC-MAIN-2024-51/af/train-*
32
- - config_name: CC-MAIN-2024-51-de
33
- data_files: data/CC-MAIN-2024-51/de/*.parquet
34
- - config_name: CC-MAIN-2024-51-en
35
- data_files: data/CC-MAIN-2024-51/en/*.parquet
36
- - config_name: CC-MAIN-2024-51-es
37
- data_files: data/CC-MAIN-2024-51/es/*.parquet
38
- - config_name: CC-MAIN-2024-51-fr
39
- data_files: data/CC-MAIN-2024-51/fr/*.parquet
40
- - config_name: CC-MAIN-2024-51-fy
41
- data_files: data/CC-MAIN-2024-51/fy/*.parquet
42
- - config_name: CC-MAIN-2024-51-it
43
- data_files: data/CC-MAIN-2024-51/it/*.parquet
44
- - config_name: CC-MAIN-2024-51-nl
45
- data_files: data/CC-MAIN-2024-51/nl/*.parquet
46
- - config_name: af
47
- data_files: data/**/af/*.parquet
48
- - config_name: de
49
- data_files: data/**/de/*.parquet
50
  - config_name: default
51
  data_files: data/**/*.parquet
52
- - config_name: en
53
- data_files: data/**/en/*.parquet
54
- - config_name: es
55
- data_files: data/**/es/*.parquet
56
- - config_name: fr
57
- data_files: data/**/fr/*.parquet
58
- - config_name: fy
59
- data_files: data/**/fy/*.parquet
60
- - config_name: it
61
- data_files: data/**/it/*.parquet
62
- - config_name: nl
63
- data_files: data/**/nl/*.parquet
64
- dataset_info:
65
- config_name: CC-MAIN-2024-51-af
66
- features:
67
- - name: text
68
- dtype: string
69
- - name: id
70
- dtype: string
71
- - name: dump
72
- dtype: string
73
- - name: url
74
- dtype: string
75
- - name: date
76
- dtype: string
77
- - name: file_path
78
- dtype: string
79
- - name: license_abbr
80
- dtype: string
81
- - name: license_version
82
- dtype: string
83
- - name: license_location
84
- dtype: string
85
- - name: license_in_head
86
- dtype: bool
87
- - name: license_in_footer
88
- dtype: bool
89
- - name: potential_licenses
90
- struct:
91
- - name: abbr
92
- sequence: string
93
- - name: in_footer
94
- sequence: bool
95
- - name: in_head
96
- sequence: bool
97
- - name: location
98
- sequence: string
99
- - name: version
100
- sequence: string
101
- - name: license_parse_error
102
- dtype: bool
103
- - name: license_disagreement
104
- dtype: bool
105
- - name: language
106
- dtype: string
107
- - name: language_score
108
- dtype: float64
109
- - name: in_fw2
110
- dtype: bool
111
- splits:
112
- - name: train
113
- num_bytes: 75356511
114
- num_examples: 17733
115
- download_size: 41860751
116
- dataset_size: 75356511
117
  ---
118
 
119
  > **Raw CommonCrawl crawls, annotated with potential Creative Commons license information**
@@ -129,13 +96,13 @@ from datasets import load_dataset
129
  ds = load_dataset("BramVanroy/CommonCrawl-CreativeCommons")
130
 
131
  # Single dump, all languages
132
- ds = load_dataset("BramVanroy/CommonCrawl-CreativeCommons", "CC-MAIN-2024-51")
133
 
134
  # Single language, all dumps
135
  ds = load_dataset("BramVanroy/CommonCrawl-CreativeCommons", "nl")
136
 
137
  # Single language, single dump
138
- ds = load_dataset("BramVanroy/CommonCrawl-CreativeCommons", "CC-MAIN-2024-51-nl")
139
  ```
140
 
141
  ## Fields
@@ -175,7 +142,7 @@ Based on these criteria, the "best guessed" license is picked as the one in the
175
 
176
  ## Progress
177
 
178
- The attempt is to at least process all five RedPyjama crawls + `CC-MAIN-2024-51`.
179
 
180
  Done:
181
 
@@ -190,7 +157,7 @@ To do:
190
 
191
  - CC-MAIN-2022-05
192
  - CC-MAIN-2023-06
193
- - CC-MAIN-2024-51
194
  - CC-MAIN-2025-05
195
 
196
  ## Languages
@@ -234,7 +201,7 @@ The following languages are included.
234
 
235
  ## Acknowledgments
236
 
237
- - [TNO](https://www.tno.nl/nl/), who funded the work hours to accomplish this collection. They intend to use parts of this material for the [GPT-NL project](https://gpt-nl.nl/).
238
  - [Flemish Supercomputer Center](https://www.vscentrum.be/) for part of the compute under grant 2024-107
239
  - Guilherme Penedo ([@guipenedo](https://huggingface.co/guipenedo)) and the rest of the [FineWeb](https://huggingface.co/datasets/HuggingFaceFW/fineweb) and [datatrove](https://github.com/huggingface/datatrove) team for the help and insights
240
  - ML6 and specifically Robin Van Craenenbroek for their [Fondant Creative Commons](https://github.com/ml6team/fondant-usecase-filter-creative-commons/tree/add-fondant-usecase-cc-image-extraction) filter for image datasets. While my approach is different, their code did serve as inspiration.
 
23
  - nl
24
  - es
25
  configs:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
26
  - config_name: default
27
  data_files: data/**/*.parquet
28
+ # Languages
29
+ - config_name: afr
30
+ data_files: data/**/afr/*.parquet
31
+ - config_name: deu
32
+ data_files: data/**/deu/*.parquet
33
+ - config_name: eng
34
+ data_files: data/**/eng/*.parquet
35
+ - config_name: spa
36
+ data_files: data/**/spa/*.parquet
37
+ - config_name: fra
38
+ data_files: data/**/fra/*.parquet
39
+ - config_name: fry
40
+ data_files: data/**/fry/*.parquet
41
+ - config_name: ita
42
+ data_files: data/**/ita/*.parquet
43
+ - config_name: nld
44
+ data_files: data/**/nld/*.parquet
45
+ # Per-crawl
46
+ # CC-MAIN-2019-30
47
+ - config_name: CC-MAIN-2019-30
48
+ data_files: data/CC-MAIN-2019-30/**/*.parquet
49
+ - config_name: CC-MAIN-2019-30-afr
50
+ data_files: data/CC-MAIN-2019-30/afr/*.parquet
51
+ - config_name: CC-MAIN-2019-30-deu
52
+ data_files: data/CC-MAIN-2019-30/deu/*.parquet
53
+ - config_name: CC-MAIN-2019-30-eng
54
+ data_files: data/CC-MAIN-2019-30/eng/*.parquet
55
+ - config_name: CC-MAIN-2019-30-spa
56
+ data_files: data/CC-MAIN-2019-30/spa/*.parquet
57
+ - config_name: CC-MAIN-2019-30-fra
58
+ data_files: data/CC-MAIN-2019-30/fra/*.parquet
59
+ - config_name: CC-MAIN-2019-30-fry
60
+ data_files: data/CC-MAIN-2019-30/fry/*.parquet
61
+ - config_name: CC-MAIN-2019-30-ita
62
+ data_files: data/CC-MAIN-2019-30/ita/*.parquet
63
+ - config_name: CC-MAIN-2019-30-nld
64
+ data_files: data/CC-MAIN-2019-30/nld/*.parquet
65
+ # CC-MAIN-2020-05
66
+ - config_name: CC-MAIN-2020-05
67
+ data_files: data/CC-MAIN-2020-05/**/*.parquet
68
+ - config_name: CC-MAIN-2020-05-afr
69
+ data_files: data/CC-MAIN-2020-05/afr/*.parquet
70
+ - config_name: CC-MAIN-2020-05-deu
71
+ data_files: data/CC-MAIN-2020-05/deu/*.parquet
72
+ - config_name: CC-MAIN-2020-05-eng
73
+ data_files: data/CC-MAIN-2020-05/eng/*.parquet
74
+ - config_name: CC-MAIN-2020-05-spa
75
+ data_files: data/CC-MAIN-2020-05/spa/*.parquet
76
+ - config_name: CC-MAIN-2020-05-fra
77
+ data_files: data/CC-MAIN-2020-05/fra/*.parquet
78
+ - config_name: CC-MAIN-2020-05-fry
79
+ data_files: data/CC-MAIN-2020-05/fry/*.parquet
80
+ - config_name: CC-MAIN-2020-05-ita
81
+ data_files: data/CC-MAIN-2020-05/ita/*.parquet
82
+ - config_name: CC-MAIN-2020-05-nld
83
+ data_files: data/CC-MAIN-2020-05/nld/*.parquet
 
 
 
 
 
 
 
 
 
84
  ---
85
 
86
  > **Raw CommonCrawl crawls, annotated with potential Creative Commons license information**
 
96
  ds = load_dataset("BramVanroy/CommonCrawl-CreativeCommons")
97
 
98
  # Single dump, all languages
99
+ ds = load_dataset("BramVanroy/CommonCrawl-CreativeCommons", "CC-MAIN-2019-30")
100
 
101
  # Single language, all dumps
102
  ds = load_dataset("BramVanroy/CommonCrawl-CreativeCommons", "nl")
103
 
104
  # Single language, single dump
105
+ ds = load_dataset("BramVanroy/CommonCrawl-CreativeCommons", "CC-MAIN-2019-30-nl")
106
  ```
107
 
108
  ## Fields
 
142
 
143
  ## Progress
144
 
145
+ The attempt is to at least process all five RedPyjama crawls + `CC-MAIN-2019-30`.
146
 
147
  Done:
148
 
 
157
 
158
  - CC-MAIN-2022-05
159
  - CC-MAIN-2023-06
160
+ - CC-MAIN-2019-30
161
  - CC-MAIN-2025-05
162
 
163
  ## Languages
 
201
 
202
  ## Acknowledgments
203
 
204
+ - [TNO](https://www.tno.nl/nld/), who funded the work hours to accomplish this collection. They intend to use parts of this material for the [GPT-NL project](https://gpt-nl.nl/).
205
  - [Flemish Supercomputer Center](https://www.vscentrum.be/) for part of the compute under grant 2024-107
206
  - Guilherme Penedo ([@guipenedo](https://huggingface.co/guipenedo)) and the rest of the [FineWeb](https://huggingface.co/datasets/HuggingFaceFW/fineweb) and [datatrove](https://github.com/huggingface/datatrove) team for the help and insights
207
  - ML6 and specifically Robin Van Craenenbroek for their [Fondant Creative Commons](https://github.com/ml6team/fondant-usecase-filter-creative-commons/tree/add-fondant-usecase-cc-image-extraction) filter for image datasets. While my approach is different, their code did serve as inspiration.