AksaraLLM/aksara-pretrain-clean-v1
收藏Hugging Face2026-04-23 更新2026-04-26 收录
下载链接：
https://hf-mirror.com/datasets/AksaraLLM/aksara-pretrain-clean-v1
下载链接
链接失效反馈官方服务：
资源简介：
AksaraLLM Pretrain Clean v1是一个经过清理的印度尼西亚语预训练数据集，包含来自Wikipedia-id（2023年11月新鲜转储）、CulturaX和NusaX等多个来源的文本。数据集经过严格的去重、语言过滤和质量控制，确保数据的高质量和语言纯度。数据集主要用于文本生成和填充掩码任务，适用于自然语言处理研究和模型训练。
AksaraLLM Pretrain Clean v1 is a cleaned version of the Indonesian pretrain dataset, incorporating text from multiple sources including Wikipedia-id (fresh dump from Nov 2023), CulturaX, and NusaX. The dataset has undergone rigorous deduplication, language filtering, and quality control to ensure high data quality and language purity. It is primarily used for text-generation and fill-mask tasks, suitable for NLP research and model training.
提供机构：
AksaraLLM
