fix(megatron): pre-initialize NCCL communicator for MoE expert DP group to prevent lazy-init deadlock#9486
Open
zb2313 wants to merge 2 commits into
Open
fix(megatron): pre-initialize NCCL communicator for MoE expert DP group to prevent lazy-init deadlock#9486zb2313 wants to merge 2 commits into
zb2313 wants to merge 2 commits into