Moe inference optimizations: 15% lower expert load by request reordering (blog.doubleword.ai) 3 points by mezark 5mo ago ↗ HN
0 comments
[ 1.9 ms ] story [ 11.0 ms ] threadNo comments yet.