marcom / Llama.jl

Julia interface to llama.cpp, a C/C++ library for running language models
MIT License
25 stars 2 forks source link

Improve code in README #14

Closed marcom closed 4 months ago

marcom commented 4 months ago

Fixes #13

marcom commented 4 months ago

cc: @svilupp Any idea on the MacOS failures here? Seems to come from llama_cpp_jll

The errors are about an illegal instruction, perhaps there are some compile options active that use instructions on macos x86_64 that are not available on these cpus? I guess not that many people use that platform anymore, but might be worth fixing on the next llama_cpp_jll version bump.

I am merging this anyway as this is only a README fix.

The errors are of this kind:

[4298] signal (4.1): Illegal instruction: 4
in expression starting at /Users/runner/work/Llama.jl/Llama.jl/src/Llama.jl:12
_GLOBAL__sub_I_llama.cpp at /Users/runner/.julia/artifacts/242026219352e7e14af6046ae578398f1bce0d19/lib/libllama.dylib (unknown line)