[2/2] refactor(speculative): train the DFlash family through one forward with shared loss terms - #2990

[2/2] refactor(speculative): train the DFlash family through one forward with shared loss terms#2990

h-guo18 wants to merge 6 commits into

haoguo/dflash-loss-contextNVIDIA/Model-Optimizer:haoguo/dflash-loss-contextfrom

haoguo/dflash-shared-forwardNVIDIA/Model-Optimizer:haoguo/dflash-shared-forwardCopy head branch name to clipboard