How to convert .safesensors to mmproj-*.gguf

[F5Psyche]

Deploying the Qwen2.5-VL-7B model using llama box requires mmproj-qwen2.5-vl-7b-f16.gguf and qwen2.5-vl-7b-f16.gguf. How can I use llama.cpp to obtain mmproj-qwen2.5-vl-7b-f16.gguf and qwen2.5-vl-7b-f16.gguf

[CISC]

You have to run convert_hf_to_gguf.py twice, the second time with the --mmproj parameter.

[F5Psyche]

python convert_hf_to_gguf.py /data/deepseek/hub/Qwen/Qwen2.5-VL-7B-Instruct --mmproj --outfile /data/deepseek/hub/Qwen/Qwen2.5-VL-7B-GGUF/mmproj-Qwen2.5-VL-7B.gguf --outtype f16 and python convert_hf_to_gguf.py /data/deepseek/hub/Qwen/Qwen2.5-VL-7B-Instruct --outfile /data/deepseek/hub/Qwen/Qwen2.5-VL-7B-GGUF/Qwen2.5-VL-7B.gguf --outtype f16, are these two commands correct?

[CISC]

Yes.