Junda ChenandClaude Opus 5.5 6397822a5b [perf] FastH3 on DGX Spark: exact fused kernels, SP exchange overlap, Omni prefix attention ...

Exact-lane work from the FastH3 Spark agent team (2026-10-23): same computation
as the shipped QAD-v2 Spark recipe, measured warm on GB10.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

2026-10-23 22:53:50 +00:00

..

minimax_h3_swiglu_gemm.cu

[perf] FastH3 on DGX Spark: exact fused kernels, SP exchange overlap, Omni prefix attention

2026-10-23 22:53:50 +00:00

minimax_h3_swiglu_quant.cuh

[perf] FastH3 on DGX Spark: exact fused kernels, SP exchange overlap, Omni prefix attention

2026-10-23 22:53:50 +00:00