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.
post
there doesn't seem to be anything here
1 comment in c/technology@lemmy.ml view that post
no comments (yet)