mirror of
https://github.com/ollama/ollama.git
synced 2026-04-28 23:18:14 -05:00
Provide a mechanism for users to set aside an amount of VRAM on each GPU to make room for other applications they want to start after Ollama, or workaround memory prediction bugs