[Speculative Decoding] DFlash2 draft variant (sublayer convolution + candidate selector) - #2216

[Speculative Decoding] DFlash2 draft variant (sublayer convolution + candidate selector)#2216

h-guo18 wants to merge 8 commits into

mainNVIDIA/Model-Optimizer:mainfrom

haoguo/dflash2-supportNVIDIA/Model-Optimizer:haoguo/dflash2-supportCopy head branch name to clipboard