Invalid Split File
#1
by
simusid
- opened
Did a git pull of llama.cpp, and clean build. Running llama-cli or llama-server results in:
llama_model_load: error loading model: invalid split file: models/Meta-Llama-3.1-70B-Instruct-Q8.gguf
What OS/GPU? did you download both files? I was able to load it (without combining them)
OS is RHEL, GPU is 8 X V100
I did download both and yes, I did combine them (via cat part1 part2 > Model.GGUF which has worked in the past)
honestly did not know that I do not have to combine, will try loading them split right now.
cheers!
oh yeah you can't cat them together like that
you need to use llama-split to combine, but the easiest is to just not bother, if you load the first part it'll search for the second automagically