Alternate quantizations.

#1
by ZeroWw - opened

These are my own quantizations (updated almost daily).

The difference with normal quantizations is that I quantize the output and embed tensors to f16.
and the other tensors to 15_k,q6_k or q8_0.
This creates models that are little or not degraded at all and have a smaller size.
They run at about 3-6 t/sec on CPU only using llama.cpp
And obviously faster on computers with potent GPUs

Yo, thanks for quanting my model man. I've seen you floating around hf and used your quants before. Much appreciated, I've updated the model card to add you into the quants section

kromeurus changed discussion status to closed

Sign up or log in to comment