Hugging Face's logo

Datasets:
LLM360
/
TxT360

License:
Dataset card Viewer Files Community
3
victormiller commited on
Commit
db497b1
•
1 Parent(s): 8e76605

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -1
README.md CHANGED
@@ -1,7 +1,7 @@
1
  ---
2
  license: odc-by
3
  ---
4
- # TxT360: a globally deduplicated dataset for LLM pretraining
5
  <center><img src="llm360_logo(1).png" alt="k2 eval table" /></center>
6
 
7
  ## We introduce TxT360 (Trillion eXtracted Text) the first dataset to globally deduplicate 99 CommonCrawl snapshots and 14 commonly used non-web data sources (e.g. FreeLaw, PG-19, etc.) providing pretraining teams with a recipe to easily adjust data weighting and train the most performant models.
 
1
  ---
2
  license: odc-by
3
  ---
4
+ # TxT360: A Top-Quality LLM Pre-training Dataset Requires the Perfect Blend
5
  <center><img src="llm360_logo(1).png" alt="k2 eval table" /></center>
6
 
7
  ## We introduce TxT360 (Trillion eXtracted Text) the first dataset to globally deduplicate 99 CommonCrawl snapshots and 14 commonly used non-web data sources (e.g. FreeLaw, PG-19, etc.) providing pretraining teams with a recipe to easily adjust data weighting and train the most performant models.