Skywork-13B

model
Skywork
Bilingual (Chinese/English) foundation model pre-trained on 3.2 trillion tokens. Released with SkyPile, a 150B-token open Chinese web corpus. Variants: Base, Chat, Math, MM.
Paper (arXiv)HuggingFace

Model Details

Architecture DENSE
Parameters 13B
Context window 4,096

Paper

arXiv: 2310.19341

open-weightmultilingual

Related