文章主题:moss-moon-003, moss-moon-003-sft-plugin, moss-moon-003-pm
IT之家 4 月 21 日消息,复旦大学自然语言处理实验室开发的新版 MOSS 模型今日正式上线,成为国内首个插件增强的开源对话语言模型。
目前,MOSS 模型已上线开源,相关代码、数据、模型参数已在 Github 和 Hugging Face 等平台开放,供科研人员下载。
据介绍,MOSS 是一个支持中英双语和多种插件的开源对话语言模型,moss-moon 系列模型具有 160 亿参数,在 FP16 精度下可在单张 A100 / A800 或两张 3090 显卡运行,在 INT4/8 精度下可在单张 3090 显卡运行。MOSS 基座语言模型在约七千亿中英文以及代码单词上预训练得到,后续经过对话指令微调、插件增强学习和人类偏好训练具备多轮对话能力及使用多种插件的能力。
MOSS 来自复旦大学自然语言处理实验室的邱锡鹏教授团队,名字与《流浪地球》电影中的 AI 同名,已发布至公开平台(https://moss.fastnlp.top/),邀请公众参与内测。
IT之家查看 MOSS 的 GitHub 页面发现,该项目所含代码采用 Apache 2.0 协议,数据采用 CC BY-NC 4.0 协议,模型权重采用 GNU AGPL 3.0 协议。如需将该项目所含模型用于商业用途或公开部署,需要签署文件并发送至 robot@fudan.edu.cn 取得授权,商用情况仅用于记录,不会收取任何费用。
MOSS 用例: ▲ 解方程▲ 生成图片▲ 无害性测试模型🌟 Moss-Moon-003 Base: MOSS-003 Foundation Model 🌟🚀 This is the MOSS-003 Foundation Model, a high-quality bilingual corpus-based pre-trained model trained on massive amounts of data. The pre-training corpus contains approximately 700 billion words and involves around 6.67 x 10^22 floating-point operations.💡 The model was trained using self-supervised learning techniques to enhance its language understanding capabilities. This approach allows the model to learn from unlabelled data, which is a more efficient way of training compared to traditional supervised methods.🌐 The pre-trained model can be easily integrated into various applications and systems, making it a versatile tool for natural language processing tasks. It can be fine-tuned on specific domains or datasets to improve its performance in those areas.💡 Moss-Moon-003 Base is an excellent resource for researchers, developers, and anyone interested in exploring the potential of pre-trained models in natural language processing. With its powerful capabilities and ease of use, it’s a must-have tool for any NLP project.# Moss-Moon-003 Base- Pre-trained corpus: 700 billion words- Training data: Self-supervised learning techniques- Integration: Easy integration into various applications and systems- Fine-tuning: Improves performance in specific domains or datasets- Versatile tool for natural language processing tasks- Powerful capabilities- Ease of use- Must-have tool for NLP projects
moss-moon-003-sft: 基座模型在约 110 万多轮对话数据上微调得到,具有指令遵循能力、多轮对话能力、规避有害请求能力。
🌟 Moss Moon 003 SFT Plugin: 🌟🚀 The base model was fine-tuned on over 110,000 rounds of dialogue data and approximately 30,000 plugin enhancements. It also has the ability to use search engines, generate text from images, perform calculations, and solve equations.💡 Moss Moon 003 SFT is built upon the foundation of Moss Moon 003, adding four additional features: search engines, image-to-text generation, calculator, and equation-solving capabilities.
🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀🚀 Moss Moon 003 PM 🚀
🚀 Moss Moon 003 – The Ultimate Model 🚀🚀 Moss Moon 003 is the result of a meticulous training process on the powerful Moss Moon 003-SFT foundation. This advanced model has been fine-tuned to provide unparalleled accuracy, security, and stability in its responses. 🌟🚀 We are excited to announce that this groundbreaking model will soon be available for public use. Stay tuned for updates on its release! 📈#MossMoon003 #Accuracy #Security #Stability #Release #PublicUse #TrainingProcess #Foundation #FineTuning #UnparalleledAccuracy #Security #Stability #RepliesQuality #Excitement #Announcement #Update #PublicUse #StayTuned #Release #PublicUse #UpcomingNews
🚀 Moss Moon 003 Plugin: A Stronger Intent Understanding and Plugin Usage Capable Model Trained on Moss Moon 003 SFT Plugin’s Foundation 🌟🚀 Moss Moon 003 Plugin is a powerful tool that has been trained using the Moss Moon 003 SFT Plugin as its foundation. This model has undergone extensive training to enhance its ability to understand user intent and make more informed decisions when it comes to plugin usage.💡 The Moss Moon 003 PM, which stands for Preference Model, plays a crucial role in this process. It helps the model learn from user behavior and preferences, allowing it to adapt to different situations and provide better results.🚀 The team behind Moss Moon 003 Plugin is committed to making their work accessible to everyone. They plan to release this updated version soon, so stay tuned for more updates!💡 If you’re interested in learning more about Moss Moon 003 Plugin or how it works, feel free to reach out to the team via email at mossmoon003@example.com.# Moss Moon 003 Plugin- A powerful tool trained on Moss Moon 003 SFT Plugin’s foundation.- Enhanced intent understanding and plugin usage capabilities through extensive training with Moss Moon 003 PM.- Soon to be released, stay tuned for updates!- Contact the team via email at mossmoon003@example.com.
数据# Moss-002-SFT-Data: A Comprehensive Analysis of User Engagement, Loyalty, and Safety in Multi-Round Dialogue DataAs a seasoned article writing expert, I understand the importance of presenting information in an engaging and informative manner. In this case, we have the Moss-002-SFT-Data, which provides insights into user engagement, loyalty, and safety in multi-round dialogue data.The Moss-002-SFT-Data is a comprehensive analysis that covers three key aspects: usefulness, loyalty, and safety. It includes over 570,000 English conversations generated by text-davinci-003 and over 590,000 Chinese conversations.This data provides valuable insights into how users interact with each other in multi-round dialogue scenarios. By analyzing the data, we can gain a better understanding of user behavior and preferences, which can be used to improve the overall user experience.In addition to providing insights into user engagement, loyalty, and safety, this data also offers valuable information for businesses looking to improve their customer service and support. By analyzing the data, businesses can identify areas where they need to improve and take steps to address those issues.Overall, the Moss-002-SFT-Data is a valuable resource for anyone interested in understanding user engagement, loyalty, and safety in multi-round dialogue scenarios. It provides insights that can be used to improve the overall user experience and help businesses provide better customer service and support.
moss-003-sft-data: moss-moon-003-sft 所使用的多轮对话数据,基于 MOSS-002 内测阶段采集的约 10 万用户输入数据和 gpt-3.5-turbo 构造而成,相比 moss-002-sft-data,moss-003-sft-data 更加符合真实用户意图分布,包含更细粒度的有用性类别标记、更广泛的无害性数据和更长对话轮数,约含 110 万条对话数据。目前仅开源少量示例数据,完整数据将在近期开源。
moss-003-sft-plugin-data: moss-moon-003-sft-plugin 所使用的插件增强的多轮对话数据,包含支持搜索引擎、文生图、计算器、解方程等四个插件在内的约 30 万条多轮对话数据。目前仅开源少量示例数据,完整数据将在近期开源。
moss-003-pm-data: moss-moon-003-pm 所使用的偏好数据,包含在约 18 万额外对话上下文数据及使用 moss-moon-003-sft 所产生的回复数据上构造得到的偏好对比数据,将在近期开源。
MOSS 的 GitHub 页面:点此查看
AI时代,掌握AI大模型第一手资讯!AI时代不落人后!
免费ChatGPT问答,办公、写作、生活好得力助手!
扫码右边公众号,驾驭AI生产力!