llama : add llama_model_ftype_name()

[angt]

Overview

Add llama_model_ftype_name()

Additional information

Expose the model file type (quantization) name, e.g. "Q8_0" or "Q4_K - Medium", through a new public C API. The returned pointer is valid for the lifetime of the model and nullptr when the model is invalid or the file type is unknown.

$ ./build/bin/llama cli -hf unsloth/Qwen3.5-0.8B-GGUF
Downloading Qwen3.5-0.8B-Q4_K_M.gguf ─────────────────────────────── 100%
Downloading mmproj-BF16.gguf ─────────────────────────────────────── 100%

Loading model...


▄▄ ▄▄
██ ██
██ ██  ▀▀█▄ ███▄███▄  ▀▀█▄    ▄████ ████▄ ████▄
██ ██ ▄█▀██ ██ ██ ██ ▄█▀██    ██    ██ ██ ██ ██
██ ██ ▀█▄██ ██ ██ ██ ▀█▄██ ██ ▀████ ████▀ ████▀
                                    ██    ██
                                    ▀▀    ▀▀

build      : b9841-e1bbcdc25
model      : unsloth/Qwen3.5-0.8B-GGUF
ftype      : Q4_K - Medium
modalities : text, vision

available commands:
  /exit or Ctrl+C     stop or exit
  /regen              regenerate the last response
  /clear              clear the chat history
  /read <file>        add a text file
  /glob <pattern>     add text files using globbing pattern
  /image <file>       add an image file
  /video <file>       add a video file
  

See #25020

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: YES

[ngxson]

no strong opinion about this feature, but IMO the more important thing is to get the :tag of the running model. IIRC I had that logic before the HF migration, but seems like it's no longer there

for example: if I run without the tag: -hf user/model, then common_params_model::get_name() currently have no info about the tag being used. would be useful if you can fix that @angt

[angt]

for example: if I run without the tag: -hf user/model, then common_params_model::get_name() currently have no info about the tag being used. would be useful if you can fix that @angt

I'll make a PR for that, but I still think this one is useful for the mentioned issue

[ngxson]

sounds ok to me, not sure if there are any edge cases, but let's see