Skip to main content

triton-ascend-attention

适用于注意力(attention)机制类算子的优化指南。当算子的核心计算是 Transformer 风格的注意力运算时应选择此指南,典型算子包括:self_attention, cross_attention, multi_head_attention, flash_attention, scaled_dot_product_attention, causal_attention, masked_attention 等。涵盖 QKV 矩阵乘分块、在线 softmax、因果 mask 处理、Flash Attention 分块策略等关键技巧。不适用于不含注意力结构的普通矩阵乘法或归约运算。

Jump to install

Source facts

Repository
mindspore-ai/akg
Last source activity
March 28, 2026 at 06:52
Detected SKILL.md language
Chinese
Stars
259
Forks
48

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.