So we can have meta packages such as "vllm" that automatically installs the corresponding package depending on the GPU that is being currently detected in the system. Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
