-
Notifications
You must be signed in to change notification settings - Fork 718
model loading gets into swap #429
Copy link
Copy link
Closed
Description
if memory guard is on, when the server starts or when i load a model, it uses way too much vram, i.e. more than it needs. swap is being used. swap is only partially released after the model has loaded into the vram.
this problem does not occur when memory guard is off.
the problem is, memory guard turns itself back on everytime i restart the app or simply refresh the webpage.
i am on v.0.2.23
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels