mirror of
https://github.com/dogkeeper886/ollama37.git
synced 2025-12-10 15:57:04 +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