Server: Multimodal Model Input Parameter No longer Exists

[lyonsniall]

I have noticed that when using the server that the --mmproj parameter for multimodal models has been disabled. Although it still remains in the README. Is there an alternative to --mmproj , I cannot seem to find one in the code.

Any help on this would be great.

Code to reproduce:
./server -m ./ggml-model-q4_k.gguf --mmproj ./mmproj-model-f16.gguf -ngl 1

Error:
error: unknown argument: --mmproj usage: ./server [options]

[Jeximo]

Although it still remains in the README.

Yes, readme needs an update. It's temporarily disabled to be reimplemented later: #5882 (comment)

You might use older versions of llama.cpp to keep the feature, or try https://github.com/ggerganov/llama.cpp/tree/947d3ad27d94f1addef76b5d64c314618f063933/examples/llava#llava

[Single430]

Although it still remains in the README.

Yes, readme needs an update. It's temporarily disabled to be reimplemented later: #5882 (comment)

You might use older versions of llama.cpp to keep the feature, or try https://github.com/ggerganov/llama.cpp/tree/947d3ad27d94f1addef76b5d64c314618f063933/examples/llava#llava

Hello, may I ask when this parameter can be restored? --mmproj

[randoentity]

See Mozilla-Ocho/llamafile@8ff1283 for an example on how to selectively revert this change, or use llamafile if you weren't aware of that project.

[github-actions]

This issue was closed because it has been inactive for 14 days since being marked as stale.