Show the share of training tokens by source as a table
Browse files
README.md
CHANGED
|
@@ -39,7 +39,14 @@ The corpus indexes **170,236 documents and 468.1M tokens**. Philosophy and psych
|
|
| 39 |
|
| 40 |
## Sources and construction
|
| 41 |
|
| 42 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
|
| 44 |
Web selection uses term-density and source-specific quality filters. Documents are split by work. The pipeline removes exact duplicates and screens additional documents for overlap using sampled word 13-grams. Shared passages and reprinted material can still cross splits.
|
| 45 |
|
|
|
|
| 39 |
|
| 40 |
## Sources and construction
|
| 41 |
|
| 42 |
+
| Source | Training tokens |
|
| 43 |
+
| --- | ---: |
|
| 44 |
+
| [FineWeb-Edu](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu) web pages | 139.6M (31.0%) |
|
| 45 |
+
| [Essential-Web](https://huggingface.co/datasets/Research-EAI/essential-web-1t-sample-fdc-partitioned) web pages | 120.6M (26.8%) |
|
| 46 |
+
| Books from Project Gutenberg, Standard Ebooks, Wikisource and Classics in the History of Psychology | 68.8M (15.2%) |
|
| 47 |
+
| Stanford and Internet Encyclopedias of Philosophy, Notre Dame Philosophical Reviews | 58.5M (13.0%) |
|
| 48 |
+
| Wikipedia | 45.8M (10.2%) |
|
| 49 |
+
| PubMed Central reviews, open textbooks, Noba modules | 17.7M (3.9%) |
|
| 50 |
|
| 51 |
Web selection uses term-density and source-specific quality filters. Documents are split by work. The pipeline removes exact duplicates and screens additional documents for overlap using sampled word 13-grams. Shared passages and reprinted material can still cross splits.
|
| 52 |
|