3 ms·
> When running larger models that don't fit into VRAM on macOS, Ollama will now split the model between GPU and CPU to maximize performance. Is there a way to
by _ink_ 2y ago
> When running larger models that don't fit into VRAM on macOS, Ollama will now split the model between GPU and CPU to maximize performance.
Is there a way to achieve this on a PC (preferably Windows)?
- fermuch 2y agoOn linux at least it already works like that
- planetafro 2y agoI use WSL2 on Win11 and it uses my GPU, no issues. I'm not a power user so YMMV.