Native Sparse Attention: Hardware-Aligned and Natively Trainable (arxiv.org)

2 points by teepo 1y ago ↗ HN

0 comments

[ 2.5 ms ] story [ 10.8 ms ] thread

No comments yet.

A showcase of shard-db. Data from the Hacker News API. Refreshed every 15 minutes.

Source · Live DB stats