Skip to content

LSE v0.4.22 — faster Q4 prefill

Choose a tag to compare

View all tags
@Geramy Geramy released this 30 Sep 16:20
· 21 commits to master since this release
v0.4.22
cc53abd
  1. Faster Q4 prefill with optimized nibble expansion and paired activation staging: 597.9 pp/s peak on a warm full prompt with BF16 KV and no prompt cache reuse.
  2. Fused eligible single-token Q4 SwiGLU gate/up projections and activation; updated dispatch regression coverage.
  3. Fixed RDNA4 prefetch address/span lowering in bundled Loom. Previously observed live DFlash2 peaks: 67.6 tok/s decode and 96% acceptance on Qwen3.8-27B Q4, R9700/macOS. These are separate workload peaks.
Assets 6
Loading