Skip to content

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
modelscope:mainfrom
zb2313:fix/moe-nccl-lazy-init-deadlock
Open

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

Update swift/megatron/utils/megatron_lm_utils.py

89b7948
Select commit
Loading
Failed to load commit list.
Sign in for the full log view

Annotations

1 warning
unittest
succeeded Jun 4, 2026 in 50m 9s