metadata
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.
{
"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
}