Implementation of the sparse attention pattern proposed by the Deepseek team in their "Native Sparse Attention" paper
Catalog snapshot
Repository health snapshot: 52 forks and 12 open issues at review time. Platform support is inferred conservatively from repository topics.
Implementation of the sparse attention pattern proposed by the Deepseek team in their "Native Sparse Attention" paper Fossfits includes native-sparse-attention-pytorch as an actively maintained llm infra option: its public repository was updated on 2025-08-15, uses the MIT license, and had 813 stars when reviewed. Check its documentation and release notes before production adoption.
artificial intelligence · attention · deep learning · sparse attention