kacper-daftcode/vLLM-Moet

vLLM-Moet

✅ Recently maintained

A vLLM patch + hand‑written SM120 SASS kernels: 2‑bit MoE experts + an FP4 "delta" cache that recovers precision — matching the official (NV)FP4 checkpoint's quality on consumer Blackwell cards

561Stars
50Forks
+117-day stars

Sass · Apache-2.0 · Last push 2026-10-30

Official repository & installation ↗

Star history

2026-06-152026-11-02
View historical data
DateStars
2026-10-04540
2026-10-05540
2026-10-06540
2026-10-07540
2026-10-08540
2026-10-09540
2026-10-10540
2026-10-11540
2026-10-12540
2026-10-13541
2026-10-14541
2026-10-15541
2026-10-16542
2026-10-17543
2026-10-18545
2026-10-19545
2026-10-20545
2026-10-21545
2026-10-22545
2026-10-23548
2026-10-24549
2026-10-25549
2026-10-26553
2026-10-27557
2026-10-28557
2026-10-29558
2026-10-30561
2026-10-31561
2026-11-01561
2026-11-02561

Get the project

git clone https://github.com/kacper-daftcode/vLLM-Moet.git

Related projects & alternative candidates

Related domains do not imply feature parity. Review official documentation and licenses.