最新开源语言模型 Vicuna | 媲美GPT-4 | 支持本地化私有部署
作者：FancyPig | 发布时间： | 更新时间：
相关阅读
data-postsbox="{"id":34361,"title":"ChatGPT本地化 | 如何在你的电脑上部署开源GPT-3语言模型 | ChatGLM","author":"FancyPig","author_id":1,"cover_image":"https://static.pigsec.cn/wp-content/uploads/2023/03/20230326142701279.png","cover_video":"","views":5541,"comment_count":21,"category":"knowledge","is_forum_post":false}">{"id":34361,"title":"ChatGPT本地化 | 如何在你的电脑上部署开源GPT-3语言模型 | ChatGLM","author":"FancyPig","author_id":1,"cover_image":"https://static.pigsec.cn/wp-content/uploads/2023/03/20230326142701279.png","cover_video":"","views":5541,"comment_count":21,"category":"knowledge","is_forum_post":false}
data-postsbox="{"id":34010,"title":"ChatGPT本地化 | 如何在你的电脑上部署开源GPT-3语言模型 | LLaMA \u0026 Alpaca","author":"FancyPig","author_id":1,"cover_image":"https://static.pigsec.cn/wp-content/uploads/2023/03/20230321010140581.jpg","cover_video":"https://v.pigsec.cn/Install%20-ChatGPT-%20ON%20Your%20LOCAL%20Computer-%20LLaMA%20%26%20Alpaca_ts.m3u8","views":4833,"comment_count":11,"category":"knowledge","is_forum_post":false}">{"id":34010,"title":"ChatGPT本地化 | 如何在你的电脑上部署开源GPT-3语言模型 | LLaMA \u0026 Alpaca","author":"FancyPig","author_id":1,"cover_image":"https://static.pigsec.cn/wp-content/uploads/2023/03/20230321010140581.jpg","cover_video":"https://v.pigsec.cn/Install%20-ChatGPT-%20ON%20Your%20LOCAL%20Computer-%20LLaMA%20%26%20Alpaca_ts.m3u8","views":4833,"comment_count":11,"category":"knowledge","is_forum_post":false}
data-postsbox="{"id":32445,"title":"如何制作一个ChatGPT聊天页面 | ChatGPT国产化思路","author":"FancyPig","author_id":1,"cover_image":"https://static.pigsec.cn/wp-content/uploads/2023/02/20230216014624643.png","cover_video":"","views":92244,"comment_count":16,"category":"knowledge","is_forum_post":false}">{"id":32445,"title":"如何制作一个ChatGPT聊天页面 | ChatGPT国产化思路","author":"FancyPig","author_id":1,"cover_image":"https://static.pigsec.cn/wp-content/uploads/2023/02/20230216014624643.png","cover_video":"","views":92244,"comment_count":16,"category":"knowledge","is_forum_post":false}
视频讲解
本期视频中，我们将对比著名的语言模型ChatGPT和新开源的聊天机器人Vicuna。Vicuna声称其质量与ChatGPT相当，并使用了一种有趣的训练数据集和评估策略。该模型由来自ShareGPT网站上共享的聊天记录微调而成，这是一个允许用户共享ChatGPT对话的网站。研发人员从该网站收集了70,000个对话进行微调，在8台A100 GPU上训练了一天。他们还选择了80个多样化的问题，使用GPT-4来评估不同模型的输出效果，包括LLAMA，Alpaca，ChatGPT和Vicuna，最终发现其效果惊人，基本上已经达到了GPT-4的90%……