You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Those have to either be requested from Meta via their project sign up, or from leaked sources. They do not ship with the project.
Once you have the modules downloaded, and there is a github that has a script to help with that, you will need to put them in a folder called 'models', and then run the two commands in the main readme.md of this repository that will convert and quantinize the files to work with this project.
------------------ 原始邮件 ------------------
发件人: "Jianqing ***@***.***>;
发送时间: 2023年3月16日(星期四) 晚上6:24
收件人: ***@***.***>;
抄送: ***@***.***>; ***@***.***>;
主题: Re: [ggerganov/llama.cpp] How to? (install models) (Issue #188)
Actually, we can download the models with aria2
meta-llama/llama#73
For download 7B:
aria2c --select-file 21-23,25,26 'magnet:?xt=urn:btih:b8287ebfa04f879b048d4d4404108cf3e8014352&dn=LLaMA'
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.Message ID: ***@***.***>
You can access the LLaMA model files by filling out the following form https://docs.google.com/forms/d/e/1FAIpQLSfqNECQnMkycAp2jP4Z9TFX0cGR4uf7b_fBxjY_OjhJILlKGA/viewform to request Meta's permission to use it. That is the primary source of the primary materials. Please do not post links to alternative sources on this issue tracker. This issue tracker is for bugs and feature requests. Thank you!
Activity
thezolon commentedon Mar 16, 2023
Those have to either be requested from Meta via their project sign up, or from leaked sources. They do not ship with the project.
Once you have the modules downloaded, and there is a github that has a script to help with that, you will need to put them in a folder called 'models', and then run the two commands in the main readme.md of this repository that will convert and quantinize the files to work with this project.
dnelyubenko commentedon Mar 16, 2023
I can't find, cuz I'm in the region under sanctions
And my VPN doesn't work now :(
tmzncty commentedon Mar 16, 2023
如果你来自中国，请使用链接：https://pan.baidu.com/s/1Y7YWdFWX1Yzy2Yuujp8Tqg?pwd=1p5n
提取码：1p5n
--来自百度网盘超级会员V4的分享
如果你来自其他国家，我还没上传到onedrive等，谷歌云盘暂时作为备份，不分享。
If you come from CHINA, please use this 链接：https://pan.baidu.com/s/1Y7YWdFWX1Yzy2Yuujp8Tqg?pwd=1p5n
提取码：1p5n
--来自百度网盘超级会员V4的分享
If you come from another contury ,please wait me upland.
tmzncty commentedon Mar 16, 2023
注意，这里都是原模型，非转换后的。
Notion: The models is real, not change.
xiliuya commentedon Mar 16, 2023
https://huggingface.co/datasets/nyanko7/LLaMA-65B/tree/main
tmzncty commentedon Mar 16, 2023
tmzncty commentedon Mar 16, 2023
mikewii commentedon Mar 16, 2023
what you do with .pth files? llama executable requires .bin model
tmzncty commentedon Mar 16, 2023
Use convert-pth-to-ggml.py
May be you can read this https://www.bilibili.com/read/cv22434771
Or this #103
.pth is real LLAMA, need to change to .bin
I am uploading them.
https://zl86n-my.sharepoint.com/:f:/g/personal/ticrawler_tmzn_top/EosXk2cjaGNKu_LdygSz7uMBQT8GHMWO14Y-0EqWhxphxw?e=7LdG9i
tmzncty commentedon Mar 16, 2023
There are changed models.
dnelyubenko commentedon Mar 16, 2023
I from Crimea and I'll be very grateful if u can upload through another service
UPD: tnx, files downloaded
tmzncty commentedon Mar 16, 2023
https://zl86n-my.sharepoint.com/:f:/g/personal/ticrawler_tmzn_top/EosXk2cjaGNKu_LdygSz7uMBQT8GHMWO14Y-0EqWhxphxw?e
If you can't download, please tell me.
pjq commentedon Mar 16, 2023
Actually, we can download the models with aria2
For download 7B:
tmzncty commentedon Mar 16, 2023
jart commentedon Mar 16, 2023
You can access the LLaMA model files by filling out the following form https://docs.google.com/forms/d/e/1FAIpQLSfqNECQnMkycAp2jP4Z9TFX0cGR4uf7b_fBxjY_OjhJILlKGA/viewform to request Meta's permission to use it. That is the primary source of the primary materials. Please do not post links to alternative sources on this issue tracker. This issue tracker is for bugs and feature requests. Thank you!
fix(meta): add GGML_OP_TURBO_WHT to tensor-split path (ggml-org#196)