人工智能 [2]

Native Sparse Attention: Hardware-Aligned and Natively  Trainable Sparse Attention 置顶

Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention

https://arxiv.org/abs/2502.11089 DeepSeek在AI领域的发展一直备受关注,其最新论文《Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention》更是引发了行