RIS-Kernel: Running 64k context LLMs on CPU via sparse attention(github.com)
2 pointsby santosardr7 hours ago1 comment