Here's one less-than-helpful result from HuggingFace's takeover of ggml. When I launched the latest build of llama-server, it automatically did this: ================================================================================ WARNING: Migrating cache to HuggingFace cache directory Old cache: /h
