Files
myWiki/concepts/native-sparse-attention.md

15 lines
530 B
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# Native Sparse Attention (NSA)
**硬件对齐的原生可训练稀疏注意力**DeepSeek 2025 年提出。
## 核心创新
稀疏模式在训练阶段即被学习("原生"而非仅在推理时施加同时稀疏模式设计与硬件GPU Tensor Core天然对齐。
## 相关概念
- [[sparse-attention-patterns]] — 稀疏注意力全景
- [[seer-attention]] — 可学习稀疏对比
- [[deepseek-v4-million-token-context]] — 在实际模型中的应用
- [[llm-attention-survey-2026]] — 综述参考