main
update llama.cpp submodule to `c29af7e` (#2868)
Omit build date from gzip headers
Switch to local dlopen symbols
update llama.cpp submodule to `66c1968f7` (#2618)
Always dynamically load the llm server library
add gguf file types (#2532)
use `llm.ImageData`
Ensure the libraries are present
Detect AMD GPU info via sysfs and block old cards
Add multiple CPU variants for Intel Mac
Fix up the CPU fallback selection
partial decode ggml bin for more info