Version: 0.7.1
Llamacpp-extension seems to always set --no-mmap flag, despite of the setting in UI
/extensions/llamacpp-extension/src/index.ts
line 1658:
args.push('--no-mmap')
Does it need a
if (cfg.no_mmap)
check?
This leads to suboptimal memory management when model is idle and longer load times.
Operating System
- [x ] MacOS
- [ x] Windows
- [ x] Linux