Update multimodal.md #14122
Update multimodal.md #14122
Conversation
|
maybe @ngxson has some insights? |
docs/multimodal.md
Outdated
| Below are several popular models maintained by the community that should work with llama.cpp: | ||
| - Joy Caption Beta 1: https://huggingface.co/concedo/llama-joycaption-beta-one-hf-llava-mmproj-gguf/tree/main | ||
| - Bakllava: https://huggingface.co/AI-Engine/BakLLaVA1-MistralLLaVA-7B-GGUF/tree/main | ||
| - MiMo: https://huggingface.co/unsloth/MiMo-VL-7B-RL-GGUF | ||
| - Cosmos-Reason1: https://huggingface.co/unsloth/Cosmos-Reason1-7B-GGUF |
There was a problem hiding this comment.
I don't think we should include this part, it's just too hard to maintain in the long term. The link above should be enough
There was a problem hiding this comment.
removed them.
it's a shame though, joy caption in particular is a very powerful model.
There was a problem hiding this comment.
powerful or not is subjective - we cannot bias toward particular model(s) in the docs
There was a problem hiding this comment.
powerful or not is subjective - we cannot bias toward particular model(s) in the docs
true, but still sad nonetheless. it's a great model :)
There was a problem hiding this comment.
I started a helper doc here: https://github.com/ddpasa/cool_links/blob/main/llama.cpp.md
According to the internet (reddit, hackernews) llama.cpp has a huge model discovery problem. I hope it helps a little.
I am not too sure what to do with this. I saw people in Reddit complain about difficulties with model discovery, and made this doc change. Maybe it helps?
At the same time, most of this should be rather obvious.