Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/OpenBMB/MiniCPM
MiniCPM-2B: An end-side LLM outperforms Llama2-13B.
https://github.com/OpenBMB/MiniCPM
Last synced: about 2 months ago
JSON representation
MiniCPM-2B: An end-side LLM outperforms Llama2-13B.
- Host: GitHub
- URL: https://github.com/OpenBMB/MiniCPM
- Owner: OpenBMB
- License: apache-2.0
- Created: 2024-01-29T08:21:15.000Z (8 months ago)
- Default Branch: main
- Last Pushed: 2024-04-24T12:09:15.000Z (5 months ago)
- Last Synced: 2024-04-28T05:53:28.225Z (5 months ago)
- Language: Jupyter Notebook
- Homepage:
- Size: 60.9 MB
- Stars: 3,803
- Watchers: 51
- Forks: 262
- Open Issues: 27
-
Metadata Files:
- Readme: README-en.md
- License: LICENSE
Awesome Lists containing this project
- ai-game-devtools - MiniCPM-2B - side LLM outperforms Llama2-13B. | | | Tool | (Project List / <span id="tool">Tool (AI LLM)</span>)
- StarryDivineSky - OpenBMB/MiniCPM - 2B 仅有 24亿的非词嵌入参数量, 总计2.7B参数量。经过 SFT 后,在公开综合性评测集上,与 Mistral-7B相近(中文、数学、代码能力更优),整体性能超越 Llama2-13B、MPT-30B、Falcon-40B 等模型。经过 DPO 后,在当前最接近用户体感的评测集 MTBench上,也超越了 Llama2-70B-Chat、Vicuna-33B、Mistral-7B-Instruct-v0.1、Zephyr-7B-alpha 等众多代表性开源大模型。以 MiniCPM-2B 为基础构建端侧多模态大模型 MiniCPM-V,整体性能在同规模模型中实现最佳,超越基于 Phi-2 构建的现有多模态大模型,在部分评测集上达到与 9.6B Qwen-VL-Chat 相当甚至更好的性能。经过 Int4 量化后,可在手机上进行部署推理,流式输出速度略高于人类说话速度。也直接跑通了多模态大模型在手机上的部署。一张1080/2080可高效参数微调,一张3090/4090可全参数微调,一台机器可持续训练 MiniCPM,二次开发成本较低。 (文本生成、文本对话 / 大语言对话模型及数据)
- Awesome-Text2SQL - [code
- AiTreasureBox - OpenBMB/MiniCPM - 09-08_5339_3](https://img.shields.io/github/stars/OpenBMB/MiniCPM.svg)|MiniCPM-2B: An end-side LLM outperforms Llama2-13B.| (Repos)