DeepSeek 发布昇腾工具包，对齐其 Nvidia 组件
DeepSeek released an Ascend toolkit matching its Nvidia components: TileLang, DeepGEMM, DeepEP, TileKernels, FlashMLA and DeepSelect.
DeepSeek 发布了一套昇腾工具包，与其 Nvidia 组件一一对应，包括 TileLang、DeepGEMM、DeepEP、TileKernels、FlashMLA 和 DeepSelect。DeepSeek 称其训练中使用的每个 TileLang kernel 现在都有高性能昇腾实现，TileLang 支撑了 V4 系列训练的大部分 kernel。DeepSeek 还与华为联合优化了 128 卡昇腾 950 超节点，关键计算与通信测试据称接近硬件极限。
正文
当前显示原文，中文全文译文尚未就绪。
DeepSeek released an Ascend toolkit matching its Nvidia components: TileLang, DeepGEMM, DeepEP, TileKernels, FlashMLA and DeepSelect.
It says every TileLang kernel used in its training now has a high-performance Ascend implementation. TileLang powers most kernels in V4-series training.
DeepSeek and Huawei also jointly optimized a 128-card Ascend 950 supernode, with key compute and communication tests reportedly approaching hardware limits.