Datasets:

Modalities:
Tabular
Text
Formats:
parquet
Languages:
English
ArXiv:
License:
teagrjohnson commited on
Commit
229da07
·
verified ·
1 Parent(s): 0a8daf5

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -3
README.md CHANGED
@@ -102,9 +102,7 @@ feature.
102
 
103
  Provenance fields: `dolma_id` (for rehydration), `source` (Dolma sub-corpus), and
104
  `topic` (WebOrganizer topic, Common Crawl passages only).
105
-
106
- > Column names above are illustrative — adjust to match the released parquet files.
107
-
108
  ## How it was built
109
 
110
  1. **Extraction** — ~17M 3-sentence passages sampled from ~5M unique Dolma v1.7 documents, allocated proportionally across sources.
 
102
 
103
  Provenance fields: `dolma_id` (for rehydration), `source` (Dolma sub-corpus), and
104
  `topic` (WebOrganizer topic, Common Crawl passages only).
105
+
 
 
106
  ## How it was built
107
 
108
  1. **Extraction** — ~17M 3-sentence passages sampled from ~5M unique Dolma v1.7 documents, allocated proportionally across sources.