Skip to main content

in reply to ☆ Yσɠƚԋσʂ ☆

I'm surprised that the model's benchmark performance isn't that much worse than the base given that it's running with 1-bit quantization. Impressive.