[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

llama.cpp split into several subpackages



Hi,

having gathered some experience with this package, I felt it was ready
to be split from the catch-all bin:llama.cpp to something more granular.

The new upload of src:llama.cpp now contains the following binary
packages:

llama.cpp: Metapackage to help transition to the new packages
libllama0: Libraries (private, for lack of stable interface)
libllama-dev: Header files
llama-tools: The main utilities: llama-cli, -bench, -quantize, -server
llama-tools-extra: All other utilities
llama-examples: The binaries shipped built from upstream's examples/
llama-tests: Test binaries
python3-gguf: The Python library for gguf files

@Kohei, the offer still stands of course: if you have improvements on
the Python package, I'd be happy to merge them.

Also, shout-out to Mathieu for sharing his experience based on his own
builds of llama.cpp.

Best,
Christian

PS: Had to rebase and force-push debian/experimental on the recently
changed debian/unstable, but the nature of debian/experimental is more
ephemeral anyway.


Reply to: