Optimizing DeepSeek's NSA for TPUs – A Kernel Worklog

Show HN: Optimizing DeepSeek's NSA for TPUs – A Kernel Worklog

Operating in show hn launch, Optimizing DeepSeek's NSA for TPUs – A Kernel Worklog operates.

Profile

I enjoyed reading DeepSeek's Natively Sparse Attention(NSA) and implemented a JAX version and Pallas Kernel(JAX's kernel language).There's a Google Colab link to follow along at the top of the blogpost.Enjoy!

Within the show hn launch space, the company focuses on its core offering rather than a broad product line.

It is recorded at the Show HN launch stage.

Website: henryhmko.github.io/posts/nsa_tpu/nsa_tpu.html