MiMo-v2.5-Pro

#4
by RedDragonGecko - opened

Any plans to make quants of the Pro model?

I do have quants of the Pro model, but haven't uploaded them since they're quite large and I wanted the PR to get merged first. I did get feedback from ngxson recommending keeping the QKV fused instead of split like I did initially, so that change will mean I have to re-quant everything anyways.

Basically, I will get the Pro quants uploaded, when the PR is merged so I don't have to redo them :)

Sign up or log in to comment