Hacker News new | ask | show | jobs
by gunalx 58 days ago
I habe found llama-fit sometimes just selects a way to conservative load with VRAM to spare.