[Feat] [FSDP2] [SFT] [ExpertParallel] [DeepSeekV4] fsdp2 dpsk v4 flash 0731 support on only 1Node with 8*H200 - #10197

[Feat] [FSDP2] [SFT] [ExpertParallel] [DeepSeekV4] fsdp2 dpsk v4 flash 0731 support on only 1Node with 8*H200#10197

mugglewei97 wants to merge 2 commits into

modelscope:mainmodelscope/ms-swift:mainfrom

mugglewei97:feat/fsdp2-dpsk-v4-flashmugglewei97/ms-swift:feat/fsdp2-dpsk-v4-flashCopy head branch name to clipboard