CPPLLAMA_VERSION?=XXX in Makefile is not used for llama.cpp build #1436
Labels
bug
Something isn't working
good first issue
Good for newcomers
high prio
up for grabs
Tickets that no-one is currently working on
LocalAI version: Master (7641f92)
Describe the bug
current commit for llama.cpp in Makefile is:
but backend/cpp/llama/Makefile uses a default of
which is from around november 2023
CPPLLAMA_VERSION is not forwarded to LLAMA_VERSION and because of this: commit d9b33* of llama.cpp is build instead of 8a7b* (or whatever is defined in CPPLLAMA_VERSION)
To Reproduce
make CPPLLAMA_VERSION=948ff137ec37f1ec74c02905917fa0afc9b97514 build
Expected behavior
localai builds llama.cpp out of commit 948ff*
Actual result
localai builds llamap.cpp out of commit d9b33*
Additional Resources
i made a archlinux package for localai,
https://aur.archlinux.org/packages?K=localai-git
with CPU, CUDA and ROCM version.
See https://github.com/wuxxin/aur-packages/blob/main/localai-git/PKGBUILD
for the patches including this issue
The text was updated successfully, but these errors were encountered: