FP8 models
#1
by ecopoiesis - opened
Thanks for the amazing work. If you could distill into the official FP8 versions, it would be phenomenal!
I would think the ideal path would be distill into full weights and perform a new well-calibrated FP8 quant from this model.
nubie asked. can i using rtx3090 to run it?
FP8 thanks