Spaces:

---

Naphula

/

model_tools

like 7

Running

Ever since Grpyhe realeased his style tune, I've come to realize how much of a VRAM hog tied embeddings are. It uses ~1GB of extra VRAM (For Gemma4 31B) with no performance gains to show for it what-so-ever, so it would be really handy to have a script to untie the embeddings within GGUF files when there is no official transformers release of said model.