Skip to content

[AutoTP] Support tied embeddings for row-parallel output-head training - #8680

Open
jinyouzhi wants to merge 4 commits into
deepspeedai:masterfrom
jinyouzhi:fix/autotp-row-head-tie-snapshot
Open

jinyouzhi wants to merge 4 commits into
deepspeedai:masterfrom
jinyouzhi:fix/autotp-row-head-tie-snapshot