---
license: cc-by-sa-4.0
language:
- ban
task_categories:
- text-generation
---

# Balinese continued-pretraining corpus (v0)

Cleaned monolingual Balinese for LLM continued-pretraining (FineWeb-2 ban_Latn + Balinese Wikipedia + story/text sets; normalize + quality-filter + exact-dedup). Part of [Open Indonesia Models](https://github.com/timothydillan/open-indonesia-models).

```json
{
  "docs": 29825,
  "stats": {
    "in": 33670,
    "empty": 4,
    "rejected": 3539,
    "dup": 302,
    "kept": 29825
  },
  "per_source": {
    "HuggingFaceFW/fineweb-2": 9233,
    "wikimedia/wikipedia": 20986,
    "pradanaadn/balinese-story-texts-extends": 3338,
    "1ndianajones/balinese_story_telling_dataset": 89,
    "1ndianajones/balinese_semarandana_dataset": 24
  },
  "approx_words": 10318531
}
```