{"id":2047,"date":"2023-12-18T09:38:33","date_gmt":"2023-12-18T01:38:33","guid":{"rendered":"https:\/\/www.1ai.net\/?p=2047"},"modified":"2023-12-18T09:38:33","modified_gmt":"2023-12-18T01:38:33","slug":"%e9%98%bf%e9%87%8c%e9%80%9a%e4%b9%89%e5%8d%83%e9%97%ae1-8b%e8%b6%85%e8%bf%b7%e4%bd%a0%e5%a4%a7%e6%a8%a1%e5%9e%8b%ef%bc%8c5%e5%b9%b4%e5%89%8d%e7%9a%84%e8%80%81%e7%94%b5%e8%84%91%e4%b9%9f%e5%8f%af","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/2047.html","title":{"rendered":"Alitong Yi Qianwen 1.8B super mini model, even a 5-year-old computer can play it"},"content":{"rendered":"

I have recently, forSmall size AI big model<\/strong>More interested. Previously, domestic and foreign technology companies in the field of large models, generally increasing the model parameters to enhance the model capacity, from 700 million, 1.3 billion, 7 billion or even more than 20 billion, it is true that more and more powerful, but this will also make everyone's own home computer basically can not run.<\/p>\n

\"640\"<\/p>\n

Recently, Microsoft released only 270 million parameters of thePhi-2<\/strong>Large Model<\/strong>, Ali also released only 180 million parameters of theThousand Questions on Tongyi<\/a>Super Miniature Big Model<\/a><\/strong>. Not only do these small-sized large models outperform some of the larger models with 700 million or even 1.3 billion parameters in performance reviews, but because of their small size, they can basically run fast in sub-3G video memory or even CPU mode, so that theYou can play on an old computer from 5 years ago<\/strong>. Today, we're sharing aLazy pack for Tongyi Qianqian 1.8B super miniature big model<\/strong>Let's go and see.<\/p>\n

Tongyi Qianwen 1.8B Big Model<\/strong>Introduction<\/strong><\/h2>\n
The Qwen family, which now includes the Qwen base language models, Qwen-1.8B, Qwen-7B, Qwen-14B and Qwen-72B, and the Qwen-Chat chat models, Qwen-1.8B -Chat, Qwen-7B-Chat, Qwen-14B-Chat and Qwen-72B-Chat.<\/section>\n
\"640-3\"<\/section>\n
Qwen<\/strong>The Big Model has a powerful base language model, stably pre-trained for multilingual data of up to 3 trillion tokens, covering domains, languages (with a focus on Chinese and English), etc., and is capable of achieving competitive performance on benchmark datasets. There are also human-preference compliant chat models based on SFT and RLHF (not yet released) that can chat, create content, extract information, summarize, translate, encode, solve math problems, etc., and are able to use tools, play as an agent, or even play as a code interpreter, etc.<\/section>\n
And today's introduction ofQwen-1.8B large model<\/strong>, only 2.9GB of video memory is needed to run, in other words 5 years ago the video card can play, hurry up to download the lazy package, have a local AI assistant~!<\/section>\n

Offline lazy bag is here<\/strong><\/h2>\n

The old rules, we deployed on their own locally is more trouble, has made a good offline integration package for the ball everyone, download to the local unzip that is used.<\/p>\n

Please download and unzip it locally.<\/p>\n

\u2460Double-click "One-Click-Boot-Qwen_1.8B.exe<\/strong>\u201d.<\/p>\n

\u2461After double-clicking the one-click startup program, a command prompt window will open and the project will run automatically. After loading successfully, please copy the following URL by yourself to open \" http:\/\/127.0.0.1:8000\/ \" in your browser.<\/p>\n

\"640-4\"<\/p>\n

\u2462Open the link and you can seeQwen-1.8B Big Model Visualization Chat<\/strong>Main Interface:<\/p>\n

\"640-5\"<\/p>\n

The interface has already been Chineseized for you ~ it's very easy to use and you can get started in 10 seconds.<\/p>\n

    \n
  1. Input prompts: In the \"input\" field, enter prompts directly.<\/li>\n
  2. Send: Click on the \"Submit\" button.<\/li>\n<\/ol>\n

    Wait a few moments and you can get a reply from the Tongyi Thousand Questions 1.8B Big Model to you now:<\/p>\n

    \"640-6\"<\/p>\n

    From my own experience, the response is very quick and the answers are reliable, so you can try it out for yourself~<\/p>\n

    Caveats:<\/p>\n

    \u2460This project is recommended to use NVIDIA graphics card to run, only need 3G video memory, CPU can also run, but slow<\/p>\n

    \u2461Please make sure the installation path does not contain Chinese, otherwise it may cause strange adaptation problems<\/p>\n

    That's all for today, I'm Glen, thank you for reading my articles and I welcome your continued support, please<\/strong>Liking, Watching, Sharing<\/strong>Let's go for a three-legged walk.<\/strong><\/p>\n

    Qwen-1.8B large modelOffline Lazybag<\/a><\/strong><\/p>\n

    https:\/\/drive.uc.cn\/s\/e2872d9ba88c4<\/p>","protected":false},"excerpt":{"rendered":"

    I have recently, been more interested in small size AI large models. Previously, domestic and foreign technology companies in the field of large models, generally increasing the model parameters to enhance the model capacity, from 700 million, 1.3 billion, 7 billion or even more than 20 billion, it is true that it is getting stronger and stronger, but it will also make everyone's own home computer basically can not run. Recently, Microsoft released a Phi-2 large model with only 270 million parameters, and Ali also released a super-mini large model with only 180 million parameters of Tongyi Qianqian. These small-sized big model, not only in the performance evaluation, more than some 700 million or even 1.3 billion parameters of the big model, and because of the small size, basically can be in the 3G video memory or even under the CPU mode to run fast, so that 5 years ago, the old computer can also play. Today I share a Tongyi Qianqian 1.8B super mini large<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[149,144],"tags":[703,702,331],"collection":[],"class_list":["post-2047","post","type-post","status-publish","format-standard","hentry","category-jiaocheng","category-baike","tag-703","tag-702","tag-331"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/2047","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=2047"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/2047\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=2047"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=2047"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=2047"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=2047"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}