llama.cpp
#1
by
FrescoHF
- opened
Do you always use +- last llama.cpp?
depends on the +., yes. I do not update very often, but usually once per month, usually driven by some change I want/need.
it would be nice to have the llama version as metadata for various reasons, but for even more reasons, this is quite hard.
@nicoboss since I now have a nice upstream maintainer, how difficult, would you think, would it be to add code to quantize to add the llama or git build id to the resulting gguf? actually that is only half the solution, convert_hf_to_gguf.py should also do that, and quantize should then copy that. so not super trivial.