Yuna AI General

[yukiarimo]

Hi guys. If you have any questions, feel free to ask!

[gitmanbloggus]

wonder if this could work on a raspberry pi 4B w/ 8gb ram, seems powerful enough possibly?

[gitmanbloggus]

I'm truly back: I don't think webbrowser is needed (isn't available on arm for some reason)

[yukiarimo]

Not sure about the webbrowser. There are some indirect requirements that are necessary, but you can test it out in the clear environment :)

[gitmanbloggus]

I did... its fixed but can yuna make images yet, just wondering

[yukiarimo]

Yes, she does! You just need to enable this in the config.json and download any model. In the upcoming release, I will upload the full SD because the single-file HF pipeline is working very badly! But you can give it a try :)

[gitmanbloggus]

there wasn't i folder and i created the folder but still, no images + its enabled

[Aaack-Aardvark]

I'm not sure if this' a bug (not likely) or a human operator issue (more likely), but it seems that in the new V7 Yuna can't or won't load the model at all, so it won't give any answer, I haven't been able to run the program at all.

It starts but every answer is an empty bubble, I tried to change the model name in the config (adding the ".gguf" extension that was missing) and I tried to change the model to the v4_fp16 without success, I also tried to add an absolute path but that won't do anything either.

Also I've tried the previous model that worked in the previous version: yuna-ai-v3-atomic-q_6_k.gguf

Am I missing something?

[yukiarimo]

Yes, the llama-cpp-python is broken. Use koboldcpp as a backend for now. It cannot parse special tokens correctly, I guess

[Aaack-Aardvark]

Sorry for this noob question but... how in the world do I change that? I checked the /static/config.json and there's no reference to that, I also tried to git pull but it didn't work either.