bert.cpp icon indicating copy to clipboard operation
bert.cpp copied to clipboard

Is this ever going to be updated?

Open BBC-Esq opened this issue 2 years ago • 6 comments

Is this repository ever going to be updated and/or worked on or has it been abandoned?

BBC-Esq avatar Dec 01 '23 12:12 BBC-Esq

Possibly. I've been in talks with people who want to help make this production quality.

skeskinen avatar Dec 01 '23 19:12 skeskinen

Interesting. So does that mean that it will no longer be open source or what not?

BBC-Esq avatar Dec 01 '23 19:12 BBC-Esq

It would stay open source.

skeskinen avatar Dec 01 '23 19:12 skeskinen

Oh, I'm not inherently opposed to open source for-profit stuff, it just affects to the extent to which I contribute and/or use something. I use lots of stuff that was created by for-profit companies. My only recommendation though it to do thorough testing and don't challenge other technologies without providing accurate apples-to-apples comparisons. Just went through an ordeal with a guy at Huggingface at this link However, I just checked the repository again and it looks like he might have opened up some re-testing so...we'll see. I haven't read it yet.

Just my 2 cents. :-) I look forward to seeing what you produce...if I knew anything about C++ I might even consider contributing, but I'll definitely be following!

BBC-Esq avatar Dec 01 '23 19:12 BBC-Esq

@skeskinen There are talks in llama.cpp to support BERT, although I don't know the current status.

https://github.com/users/ggerganov/projects/7?pane=issue&itemId=37112645

dranger003 avatar Dec 04 '23 22:12 dranger003

Hey @skeskinen and guys, I've made another fork to support multilingual tokenizer, real batch inference and the bge series model. Check this out if you're interested. embeddings.cpp

xyzhang626 avatar Dec 20 '23 11:12 xyzhang626