chadhurley25075-png/pd-bridge

Heterogeneous prefill/decode for DeepSeek-V4-Flash: CUDA prefill (DGX Spark, vLLM) -> Metal decode (Mac Studio, oMLX) over plain 10GbE

3 best points · 1 comments · 1 Hacker News post · 151 GitHub stars · View on GitHub

Hacker News discussions

• Prefill a 284B model on Nvidia. Decode it on Apple Silicon. Over plain 10GbE (3 points, 1 comments, by jnaina, 2026-09-16)