I would like to request for tracks cuda , rocm, vulkan for the Workshop SDK llama-cpp
This is to mirror a similar track arrangement as is done for GitHub - canonical/ollama-sdk · GitHub . The ollama SDK’s division into cuda and rocm tracks is indeed because of the llama.cpp that is included in it, with different dependency chains per track, so there is a similar problem to solve here.
(For added context, llama-cpp is going to be a consumer for the similarly named llama-cpp snap, for which the same track request has been made here: Create new track for "llama-cpp" – once both approved, I will update the SDK packaging to bundle the snap, instead of fetching the upstream binary into the SDK).