mirror of
https://github.com/likelovewant/ollama-for-amd.git
synced 2025-12-21 22:33:56 +00:00
If the model doesn't fit any layers on metal, and we load zero layers we would panic trying to look up the GPU size during scheduling ops