Re-uploads needed?
#2
by bharatcoder - opened
Based on the updates, patches and fixes to llama.cpp after the initial launch, do we need a re-upload of these quants?
No most if not all the fixes will get retroactively applied to existing quants as they mainly affect how llama.cpp runs inference on Gemma 4 models. As far I'm aware no serious issues were found with the model itself. Models done very early after release lacked vision but for those, we are requeing for mmproj extraction. Was there any major bugfix that requires quants to be redone that we missed?