GGUF conversion fails
#3
by Lokator - opened
Hello!
I'm trying to use VILA / NVILA / LongVILA models in Ollama + OpenWebUI.
My problem is that when trying to use e.g ggml-org/gguf-my-repo or other GGUF converters, they always fail;
the error traces back to Llama.cpp backend and it says:
"Error converting to fp16: INFO:hf-to-gguf:Loading model: NVILA-15B
ERROR:hf-to-gguf:Model LlavaLlamaModel is not supported"
Llama.cpp doesn't seem to like 'LlavaLlamaModel' as a base-model.
(basically llama_cpp/convert_hf_to_gguf.py throws a NotImplementedError)
Can you offer any help with this?
I'd really appreciate any tips, or, best case scenario, .gguf files offered under your projects.