2 GGUF

#4
by Gomer99 - opened

is it possible to use such GGUF format which is stacked with two parts? and if yes, what should be done. I understood that before using the model it should be connected, but I am not familiar with either bibiotechniques or understanding of how it is done and that it is necessary to remember in your system.

KoboldAI org

Yes this is possible but not with the bundled Dockerfile.
You'd have to modify it slightly to where you hardcode the downloading of both files with their original filenames (or manually combine them all into one file if its not the 00001-of style GGUF's).
Then make sure it launches KoboldCpp with the filename of the first part.

Sign up or log in to comment