Closed stygmate closed 4 days ago
I can support his use case, I want to be able to run multiple fairly rarely used models on the same hardware, but not being able to unload them quickly uses all available VRAM. If it would've been possible to specify the model to use in the API, that would've been a huge help.
This issue was closed because it has been inactive for 14 days since being marked as stale.
Prerequisites
Feature Description
Drive the loading of models from server api (like ollama) can be useful. More useful can be changing loras on the fly.
Motivation
Loras are small and switching between multiple finetunes can be very useful in complex application running on small computers.
Possible Implementation
No response