optimization

2 posts tagged “optimization”.

1. We Wrote Our Own Inference Engine. It Started at 0.9 tok/s. (2026-09-06)
2. Unfusing a Kernel Made It 1.63x Faster (2026-09-01)