Skip to content

397B v7x trainer: MoE wo FSDP sharding and valid-rows-only MoE kernels - #5760

Draft
khatwanimohit wants to merge 2 commits into
mohit/trainer-perf-v7xfrom
mohit/trainer-perf-v7x-pending
Draft

khatwanimohit wants to merge 2 commits into
mohit/trainer-perf-v7xfrom
mohit/trainer-perf-v7x-pending

MoE ring: unsort backward rows and the SiLU GLU visit only the packed…

6b58e2a
Select commit
Loading
Failed to load commit list.