Chore: Model.yml new parameters

[hahuyhoang411]

Problems

Current model.yml only supports text

New features

• Add support parameters for VLM

  • Parameters
  • Cortex support: planning: Supporting vision model (Llava and Llama3.2) cortex.cpp#1493

• Add support parameters for tool use

  • Parameters
  • Cortex support: feat: Cortex supports Function Calling cortex.cpp#295

[hahuyhoang411]

1. For VLM/ SLM:

• new parameters should be added in
• VLM: img_end: ... path to mmproj
  • convert mmproj to gguf: https://github.com/ggerganov/llama.cpp/tree/master/examples/llava#llava-16-gguf-conversion
  • latest VLM doesn't have mmproj, they include the projector inside the model: https://github.com/ggerganov/llama.cpp/blob/master/examples/llava/README-minicpmv2.6.md
• SLM: speech_end: ... path to Whisper
• (Future) any2any model has both encoder and decoder for modality:
  • we might need to introduce speech_end, img_end and img_dec, speech_dec
  • Example Emu3: https://huggingface.co/BAAI/Emu3-Gen

2. For tool use:

• new parameters should be added in tool: None as default value.tool accepts list (e.g tool: [])
  Ref: https://platform.openai.com/docs/guides/function-calling