MixlyGames/

Orpheus_Zero-v1-dataset

	Dataset Card for Orpheus Zero Pretraining Corpus








	


	Dataset Details




	


	Dataset Description

This dataset is a pre-tokenized Russian/English text corpus assembled for pretraining Orpheus Zero, a ~278M parameter causal language model trained from scratch. The corpus combines game lore/wiki text, literature, encyclopedic content, and forum data, cleaned and deduplicated, then tokenized with a custom BPE tokenizer.

Curated by: MixlyGames Language(s):… See the full description on the dataset page: https://huggingface.co/datasets/MixlyGames/Orpheus_Zero-v1-dataset.