Yunqi Yan

icon company Tsinghua University icon location Beijing, China icon location Hi. I'm a Phd student in Electrical Engineering at Tsinghua University. My research work mainly focuses on resilient distribution networks.

Results 1 issues of Yunqi Yan

Training Failure due to NCCL P2P Access Not Supported Between Devices

2 comment

# Description Training is failing with a NCCL error indicating that peer-to-peer (P2P) access is not supported between the GPU devices being used. Despite setting environment variables `NCCL_P2P_DISABLE=1` and `NCCL_IGNORE_DISABLED_P2P=1`,...