llama_model_load: error loading model: invalid split file

#1
by Stefanvarunix - opened

Hi,
I tried to merge the files with
cat dbrx-16x12b-instruct-q6_k-* > dbrx-16x12b-instruct-q6_k.gguf

When running
./server -m ./dbrx-16x12b-instruct-q6_k.gguf
I get the error message
"llama_model_load: error loading model: invalid split file:"

My mistake?

Llama.cpp sharded model support is built-in now, do not concat files in a binary way. You can use gguf-split to merge or simply pass the first split to llama_model_loader.

phymbert changed discussion status to closed

Sign up or log in to comment