biondizzle 0fefadedd4 KV-1: Fix FP8 round-trip mismatch in fused quantize ...

CRITICAL: quantize must use the FP8-round-tripped block scale, not the raw
pre-FP8 value. The dequant reads the FP8 bytes back, so the quantize must
match exactly. Same pattern as quantize_nvfp4.cu. This was the root cause
of cos=0.925 (should be ~0.995).

2026-06-02 09:46:32 +00:00

..

cache

E1: Wire LayerCacheHandle gather methods + CUDA gather kernels

2026-05-30 21:09:21 +00:00

kernels

KV-1: Fix FP8 round-trip mismatch in fused quantize

2026-06-02 09:46:32 +00:00

layers

wip: SE fused SwiGLU deinterleave fix

2026-06-02 08:41:00 +00:00

loader

Restructure: cutedsl/ -> dsv4/ with proper layering

2026-05-21 17:30:44 +00:00

model

P0 COMPLETE: Eliminate ALL .item() CPU-GPU syncs from NVFP4 activation path

2026-06-01 21:05:03 +00:00

ops

fix: rope_cuda path — kernels/cuda not ops/cuda

2026-06-02 09:06:36 +00:00

reference

Restructure: cutedsl/ -> dsv4/ with proper layering

2026-05-21 17:30:44 +00:00

__init__.py

Restructure: cutedsl/ -> dsv4/ with proper layering

2026-05-21 17:30:44 +00:00