---
license: mit
datasets:
- PleIAs/common_corpus
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*.parquet
---

# Common Corpus Sample (Diverse & Deduplicated)

## Statistics
*Updated: 2026-01-20T04:19:16.000340*

- **Total Tokens:** 1.4622 Billion / 2 Billion Target
- **Rows Scanned:** 40,992,263
- **Rows Kept:** 2,100,000
- **Duplicates Removed:** 97,952
- **Short Texts Removed (<20):** 209,210
- **Runtime:** 8.18 hours

## Language Distribution
Target per language: ~200M tokens.

| Language | Documents | Tokens | % of Target |
| :--- | :--- | :--- | :--- |
| English | 174,775 | 220,000,073 | 100.0% |
| PHP | 308,775 | 220,000,126 | 100.0% |
| Java | 312,008 | 220,001,140 | 100.0% |
| R | 42,684 | 30,978,684 | 14.1% |
| C# | 366,838 | 220,002,167 | 100.0% |
| Python | 277,185 | 220,000,497 | 100.0% |
| Shell | 198,846 | 86,408,839 | 39.3% |
| JavaScript | 376,094 | 220,000,054 | 100.0% |
| Turkish | 31,676 | 16,161,256 | 7.3% |
| Azerbaijani | 11,119 | 8,611,861 | 3.9% |