StarCoder 2 (2024年)（スターコーダー2）

概要

StarCoder 2 は2024年2月BigCode (HuggingFace + ServiceNow)発表・Industry-leading 3B/7B/15B open-source code LLM 2024年BigCode StarCoder 2 OpenRAIL open-source code LLM position確立。Industry-leading 17 programming languages + The Stack v2 dataset 4T tokens + Fill-in-the-Middle + 16K context + OpenRAIL-M license。

主な特徴・仕組み

• Industry-leading BigCode Project (HuggingFace + ServiceNow Research collaboration) 2024年2月 release
• Industry-leading StarCoder 2 open-source code LLM signature
• Industry-leading 3B/7B/15B model family signature
• Industry-leading 17 programming languages signature (vs StarCoder 1 86 languages, focused)
• Industry-leading The Stack v2 dataset 4T tokens signature
• Industry-leading Fill-in-the-Middle (FIM) capability signature
• Industry-leading 16K context window
• Industry-leading OpenRAIL-M Responsible AI License signature
• Industry-leading HuggingFace + ServiceNow Research collaboration
• Industry-leading governance + data transparency signature
• Industry-leading 2024年2月 ArXiv paper "StarCoder 2 and The Stack v2 The Next Generation"
• Industry-leading BigCode Project 2022 HuggingFace + ServiceNow 2 years

まとめ

StarCoder 2 = 2024年2月BigCode HuggingFace + ServiceNow発表のpremium OpenRAIL open-source code LLM。Industry-leading 3B/7B/15B model family + 17 programming languages focused vs StarCoder 1 86 + The Stack v2 dataset 4T tokens + Fill-in-the-Middle FIM capability + 16K context window + OpenRAIL-M Responsible AI License + HuggingFace + ServiceNow Research collaboration + governance + data transparency + 2024年2月 ArXiv paper "StarCoder 2 and The Stack v2 The Next Generation" + BigCode Project 2022 HuggingFace + ServiceNow 2 years position確立。