Jev-like DiffusionGemma patch reportedly hits 22 structured generations per second at concurrency 32
A user reports testing the patch on a DGX Spark with almost no tuning. With one request at a time, the reported median (p50) latency was about 0.21 seconds.
TLDR
A user reports modifying DiffusionGemma to act like Jev and testing the patch on a DGX Spark with minimal—“basically zero”—tuning. They report about 0.21 seconds of median (p50) latency at concurrency 1, and throughput of 22 structured generations per second at concurrency 32.
Combined views
433.3K
4 Sources, first seen 3d ago
Jev-like DiffusionGemma patch reportedly hits 22 structured generations per second at concurrency 32
A user reports testing the patch on a DGX Spark with almost no tuning. With one request at a time, the reported median (p50) latency was about 0.21 seconds.