haoranxu/X-ALMA-13B-Pretrain

X-ALMA-13B-Pretrain is a 13 billion parameter multilingual pre-trained base model developed by Haoran Xu, expanding upon the ALMA-R architecture. It supports 50 languages through a plug-and-play design with language-specific modules and a specialized training recipe. This model is primarily designed for high-quality translation tasks and multilingual open-ended question answering.

Warm
Public
13B
FP8
4096
License: mit
Hugging Face

No reviews yet. Be the first to review!