[LLM] fix moe using on tensor parallelism - #11099

[LLM] fix moe using on tensor parallelism #11099

tianshuo78520a merged 3 commits into

PaddlePaddle:incubate/paddlenlp-fleety_20250421PaddlePaddle/PaddleNLP:incubate/paddlenlp-fleety_20250421from

KB-Ding:moe_fleetKB-Ding/PaddleNLP:moe_fleetCopy head branch name to clipboard