3 ms·
Or hopefully we will see an end of the LLM hype. Or at least models that don’t hog so much RAM.
by Kwpolska 3y ago
Or hopefully we will see an end of the LLM hype.
Or at least models that don’t hog so much RAM.
- ramesh31 3y ago>Or at least models that don’t hog so much RAM The RAM usage is kind of the point though; we're trading space for time. It's not a problem that the model is using it, it's just that with the default choice for UI being web based now, the unnecessary memory usage of browsers is actually starting to be a real pain point.
- 110 3y ago1. I hear you on going back to lightweight native apps. Unfortunately the Python ecosystem is not great for this. We use pyinstaller to create the native desktop app but it's a pain to manage. 2. The web UI isn't required if you use Obsidian or Emacs. That's just a convenient, generic interface that everyone can use.