tamnd commited on
Commit
768215b
·
verified ·
1 Parent(s): cb61f94

Publish shard CC-MAIN-2026-08/00095

Browse files
Files changed (3) hide show
  1. README.md +10 -10
  2. data/CC-MAIN-2026-08/00095.parquet +3 -0
  3. stats.csv +2 -1
README.md CHANGED
@@ -32,7 +32,7 @@ configs:
32
 
33
  Open Index is a large-scale web text dataset built from [Common Crawl](https://commoncrawl.org). Every page goes through a pipeline that extracts the main content from raw HTML, converts it to clean Markdown using [trafilatura](https://github.com/adbar/trafilatura), and packages the result into Parquet files with full WARC metadata preserved.
34
 
35
- The dataset currently includes crawl **CC-MAIN-2026-08** with **15,280,111 documents across 790 shards**. We plan to add more snapshots over time.
36
 
37
  Open Index is released under the **Open Data Commons Attribution License (ODC-By) v1.0**, the same license used by Common Crawl.
38
 
@@ -164,28 +164,28 @@ Empty conversions (pages where trafilatura could not extract meaningful content)
164
 
165
  ### Compression Ratios
166
 
167
- Numbers below are actual measurements summed across all 790 files of CC-MAIN-2026-08 (15,280,111 pages total), projected to the full crawl of 100,000 WARC files.
168
 
169
- | Stage | 790 files (measured) | 100,000 files (projected) | Reduction |
170
  |---|---|---|---|
171
- | Raw WARC (.warc.gz, downloaded) | ~640.3 GB | ~83 TB | — |
172
  | HTML extracted (uncompressed) | 1.8 TB | ~295 TB | — |
173
  | Packed markdown WARC (.md.warc.gz) | ~32.4 GB | ~3.7 TB | **-98.3%** vs HTML |
174
  | Final Parquet (Zstd level 19) | 21.9 GB | ~2.9 TB | **-32.4%** vs packed WARC |
175
 
176
- The big win is the HTML → Markdown step: trafilatura strips all tags, scripts, styles, navigation, and ads, keeping only the main content. This cuts 1.8 TB of uncompressed HTML down to 68.9 GB of markdown — a **98.3% reduction** — before any file-level compression is applied. Parquet with Zstd level 19 then compresses the markdown a further 68.2%.
177
 
178
- End to end: ~640.3 GB of raw gzipped WARCs becomes **21.9 GB of Parquet** — a **96.6% total reduction** — containing 15,280,111 clean markdown documents.
179
 
180
  ### Processing Times
181
 
182
- Pipeline timings across 790 shards of CC-MAIN-2026-08:
183
 
184
  ```
185
  Download (raw WARC) ████████████████████████ total 45h 7m 40s avg 3m 25s
186
- Convert (HTML → MD) █░░░░░░░░░░░░░░░░░░░░░░░ total 1h 56m 12s avg 8s
187
- Export (Parquet) █████░░░░░░░░░░░░░░░░░░░ total 9h 36m 35s avg 43s
188
- Publish (HuggingFace) ███░░░░░░░░░░░░░░░░░░░░░ total 5h 53m 34s avg 26s
189
  ```
190
 
191
  ### Dataset Charts
 
32
 
33
  Open Index is a large-scale web text dataset built from [Common Crawl](https://commoncrawl.org). Every page goes through a pipeline that extracts the main content from raw HTML, converts it to clean Markdown using [trafilatura](https://github.com/adbar/trafilatura), and packages the result into Parquet files with full WARC metadata preserved.
34
 
35
+ The dataset currently includes crawl **CC-MAIN-2026-08** with **15,299,505 documents across 791 shards**. We plan to add more snapshots over time.
36
 
37
  Open Index is released under the **Open Data Commons Attribution License (ODC-By) v1.0**, the same license used by Common Crawl.
38
 
 
164
 
165
  ### Compression Ratios
166
 
167
+ Numbers below are actual measurements summed across all 791 files of CC-MAIN-2026-08 (15,299,505 pages total), projected to the full crawl of 100,000 WARC files.
168
 
169
+ | Stage | 791 files (measured) | 100,000 files (projected) | Reduction |
170
  |---|---|---|---|
171
+ | Raw WARC (.warc.gz, downloaded) | ~641.1 GB | ~83 TB | — |
172
  | HTML extracted (uncompressed) | 1.8 TB | ~295 TB | — |
173
  | Packed markdown WARC (.md.warc.gz) | ~32.4 GB | ~3.7 TB | **-98.3%** vs HTML |
174
  | Final Parquet (Zstd level 19) | 21.9 GB | ~2.9 TB | **-32.4%** vs packed WARC |
175
 
176
+ The big win is the HTML → Markdown step: trafilatura strips all tags, scripts, styles, navigation, and ads, keeping only the main content. This cuts 1.8 TB of uncompressed HTML down to 69.0 GB of markdown — a **98.3% reduction** — before any file-level compression is applied. Parquet with Zstd level 19 then compresses the markdown a further 68.2%.
177
 
178
+ End to end: ~641.1 GB of raw gzipped WARCs becomes **21.9 GB of Parquet** — a **96.6% total reduction** — containing 15,299,505 clean markdown documents.
179
 
180
  ### Processing Times
181
 
182
+ Pipeline timings across 791 shards of CC-MAIN-2026-08:
183
 
184
  ```
185
  Download (raw WARC) ████████████████████████ total 45h 7m 40s avg 3m 25s
186
+ Convert (HTML → MD) █░░░░░░░░░░░░░░░░░░░░░░░ total 2h 2m 36s avg 9s
187
+ Export (Parquet) █████░░░░░░░░░░░░░░░░░░░ total 9h 38m 29s avg 43s
188
+ Publish (HuggingFace) ███░░░░░░░░░░░░░░░░░░░░░ total 5h 53m 52s avg 26s
189
  ```
190
 
191
  ### Dataset Charts
data/CC-MAIN-2026-08/00095.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a8581483de6eaf9d6b544126257d7793907a860c6c0d72f9b7b325fae3eaf51
3
+ size 30624565
stats.csv CHANGED
@@ -95,6 +95,7 @@ CC-MAIN-2026-08,91,19374,2578827857,94619739,29492428,2026-03-13T15:09:05Z,0,173
95
  CC-MAIN-2026-08,92,19520,2560251413,93812697,29804536,2026-03-13T15:14:30Z,343,0,53,0
96
  CC-MAIN-2026-08,93,19565,2625905695,97305588,30010782,2026-03-13T15:22:02Z,271,0,86,42
97
  CC-MAIN-2026-08,94,19407,2565026021,94786307,30258982,2026-03-13T17:40:03Z,0,339,94,30
 
98
  CC-MAIN-2026-08,98,19449,2589117299,98909003,31451722,2026-03-13T15:43:46Z,345,0,33,0
99
  CC-MAIN-2026-08,99,19479,2575024379,92544874,29917329,2026-03-13T15:48:21Z,227,0,28,0
100
  CC-MAIN-2026-08,100,19468,2560323629,93642547,30026373,2026-03-13T15:51:46Z,201,0,28,0
@@ -463,7 +464,7 @@ CC-MAIN-2026-08,1004,19467,2596884222,95469795,30330109,2026-03-13T11:50:10Z,471
463
  CC-MAIN-2026-08,1005,19447,2562067418,92075537,29388448,2026-03-13T11:56:02Z,313,0,38,0
464
  CC-MAIN-2026-08,1006,19558,2586529847,99212752,30981992,2026-03-13T11:59:05Z,145,0,36,0
465
  CC-MAIN-2026-08,1007,19353,2570554464,93567082,28716493,2026-03-13T12:03:21Z,222,0,32,0
466
- CC-MAIN-2026-08,1008,19481,2530476391,96707137,31064703,2026-03-13T17:43:34Z,515,0,89,0
467
  CC-MAIN-2026-08,1101,19721,2641197765,95719568,30240239,2026-03-13T09:59:33Z,564,0,296,0
468
  CC-MAIN-2026-08,1102,19562,2608129691,99280759,31748974,2026-03-13T10:28:07Z,1518,0,191,0
469
  CC-MAIN-2026-08,1103,19270,2517580674,89253461,28361470,2026-03-13T10:55:39Z,1380,0,268,0
 
95
  CC-MAIN-2026-08,92,19520,2560251413,93812697,29804536,2026-03-13T15:14:30Z,343,0,53,0
96
  CC-MAIN-2026-08,93,19565,2625905695,97305588,30010782,2026-03-13T15:22:02Z,271,0,86,42
97
  CC-MAIN-2026-08,94,19407,2565026021,94786307,30258982,2026-03-13T17:40:03Z,0,339,94,30
98
+ CC-MAIN-2026-08,95,19394,2589724753,96639975,30624565,2026-03-13T17:48:32Z,0,384,114,0
99
  CC-MAIN-2026-08,98,19449,2589117299,98909003,31451722,2026-03-13T15:43:46Z,345,0,33,0
100
  CC-MAIN-2026-08,99,19479,2575024379,92544874,29917329,2026-03-13T15:48:21Z,227,0,28,0
101
  CC-MAIN-2026-08,100,19468,2560323629,93642547,30026373,2026-03-13T15:51:46Z,201,0,28,0
 
464
  CC-MAIN-2026-08,1005,19447,2562067418,92075537,29388448,2026-03-13T11:56:02Z,313,0,38,0
465
  CC-MAIN-2026-08,1006,19558,2586529847,99212752,30981992,2026-03-13T11:59:05Z,145,0,36,0
466
  CC-MAIN-2026-08,1007,19353,2570554464,93567082,28716493,2026-03-13T12:03:21Z,222,0,32,0
467
+ CC-MAIN-2026-08,1008,19481,2530476391,96707137,31064703,2026-03-13T17:43:34Z,515,0,89,18
468
  CC-MAIN-2026-08,1101,19721,2641197765,95719568,30240239,2026-03-13T09:59:33Z,564,0,296,0
469
  CC-MAIN-2026-08,1102,19562,2608129691,99280759,31748974,2026-03-13T10:28:07Z,1518,0,191,0
470
  CC-MAIN-2026-08,1103,19270,2517580674,89253461,28361470,2026-03-13T10:55:39Z,1380,0,268,0