v0.32.10-rc0: nn: speed up prefill on double-scale nvfp4 models

Summary

ModelOpt checkpoints apply a float32 global scale to every projection output on top of the per-group quantization scales. Running the multiply and the cast back to the activation dtype as separate eager ops costs an extra kernel launch and a materialized intermediate per projection. Compile the multiply and cast into one kernel. On an M5 Max (medians of order-swapped A/B runs against main; greedy outputs byte-identical): qwen3.6:27b prefill 703 -> 769 t/s +7.9% muse-glimmer:30b prefill 790 -> 843 t/s +6.7% Speculative decode is unchanged within noise on both models. Only checkpoints with a global scale are affected; single-scale nvfp4, mxfp8, and affine checkpoints take the unchanged path.

Lotu Radar provides attributed news summaries and links to the original publisher. Full reporting and copyright remain with the source.