Octopus V2

[iplayfast]

https://arxiv.org/abs/2404.01744
https://huggingface.co/NexaAIDev/Octopus-v2

Personally I would like better than gtp-4 on my laptop :)

[lin72h]

I think the it's finetuned from gemma-2b, so there already support for it #5631

[AndreasKunar]

FYI impactframes/Octopus-v2-Q4_K_M-GGUF seems to work. No need to download/convert yourself, like I did initially, because the 1st huggingface.co quantized GGUF I picked did not work.

Just as clarification, Octopus V2 is not a "normal" model like gemma-1.1-2b-it, its one specialized in function calling for agents.

[zhiyuan8]

Many thanks for your interest on Octopus model, we have also open-sourced V4 and published a quantized version

https://huggingface.co/NexaAIDev/Octopus-v4
https://huggingface.co/NexaAIDev/octopus-v4-gguf