Helpers for interacting with llama.cpp.
Llama.cpp has CLI options that automatically download models from huggingface, but does not have clear, documented facilities for doing so without loading them. This can be problematic for cases where model files must be downloaded via computers without the available RAM to run them, for example when the llama.cpp server lives on a network with slow downlink speeds.
Paca provides a download subcommand for this purpose.
paca download unsloth/GLM-4.7-GGUF:BF16