盘古大模型通信优化实践：亲和昇腾平台的性能探索 Pangu's Training and Inference Communication Optimization on the Ascend Platform

精读摘要

AICon 深圳演讲介绍了盘古大模型在昇腾平台上的训练与推理通信优化实践。内容聚焦如何针对国产算力平台做亲和性优化，以提升大规模模型训练的通信效率与整体性能，为同类国产硬件上的大模型优化提供了工程参考。 An AICon Shenzhen talk presents Pangu's communication optimization practices for training and inference on Huawei's Ascend platform. It focuses on how platform-specific optimizations improve communication efficiency and overall performance for large-scale model training.