llama3.2 vision models

[mudler]

Is your feature request related to a problem? Please describe. Llama3.2 was released, and as it has multimodal support would be great to have it in LocalAI

Describe the solution you'd like

Describe alternatives you've considered

Additional context llama.cpp have several issues wrt multimodal capabilities:

• https://github.com/ggerganov/llama.cpp/issues/9643
• https://github.com/ggerganov/llama.cpp/issues/8010

vLLM has already added support for it in https://github.com/vllm-project/vllm/pull/8811

See also:

• https://github.com/ggerganov/llama.cpp/pull/9687
• https://github.com/mudler/LocalAI/issues/3535
• https://github.com/ollama/ollama/issues/6972
• https://github.com/ollama/ollama/pull/6965
• https://github.com/ollama/ollama/pull/6971
• https://github.com/ollama/ollama/pull/6963

[briansan]

+1