[3/4] perf(speculative): fuse the DFlash-family loss with the lm_head projection - #3010

[3/4] perf(speculative): fuse the DFlash-family loss with the lm_head projection#3010

h-guo18 wants to merge 1 commit into

haoguo/spec-train-inorder-compileNVIDIA/Model-Optimizer:haoguo/spec-train-inorder-compilefrom

haoguo/dflash-fused-lossNVIDIA/Model-Optimizer:haoguo/dflash-fused-lossCopy head branch name to clipboard