Talkie 1930 vintage LLM support #22687
Replies: 5 comments 4 replies
-
|
Interest is here. I'll try building it and see how it works at some point tonight and I'll report back over the week after seeing how it behaves. |
Beta Was this translation helpful? Give feedback.
-
|
Built from source and confirming it works 👌 |
Beta Was this translation helpful? Give feedback.
-
|
Just to clarify, if you're using @thomasgauthier 's GGUF you need to use his branch as well, the branches are incompatible. I'd recommend using his for now since mine is somewhat under construction, and they produce bit for bit identical outputs anyways. |
Beta Was this translation helpful? Give feedback.
-
|
That was the one I planned since he also had GGUFs |
Beta Was this translation helpful? Give feedback.
-
|
I forgot to specify, I used the @niklassheth version (works) |
Beta Was this translation helpful? Give feedback.
-
|
Thomasgauthier fork and GGUFs work and now I've realized that it has very low attention span but it is very interesting 😂 |
Beta Was this translation helpful? Give feedback.
-
|
Until it gets merged to main what build do we do? Do we stick with the thomasgauthier build? I think that the niklassheth one has the closer architecture to what will be merged right? But I don't know if it's going to be stable enough to use right now. |
Beta Was this translation helpful? Give feedback.
-
|
You can use either fork, they produce identical results when it comes to text generation. My fork will not be merged nor updated but it works fine for local inference right now. If you need some newer feature that have been added to llama.cpp since I created my fork you're better of using niklassheth's as they're keeping it in sync with the main branch, just make sure you are using their GGUF. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you. Hopefully it gets merged soon. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
A few members of the community have been working on supporting the new Talkie 1930 vintage LLM. There is already a PR opened and my GGUF upload on Hugging Face is nearing 13 000 downloads. I think this is a clear signal that the community wants this model to be supported.
We just need to get the PR merged, we haven't heard from the maintainers yet so I think we just need to get their attention (attention is all you need 🙂). Anyone who wants Talkie supported in llama.cpp please voice your interest here to help us get momentum.
Beta Was this translation helpful? Give feedback.
All reactions