biondizzle 7a05d3d3af NVFP4 router gate: use Nvfp4Linear for both checkpoint and quantized paths ...

- Checkpoint path: load NVFP4 gate weight directly into Nvfp4Linear
- BF16 path: quantize and load into Nvfp4Linear
- Both paths use proven production GEMM (no custom kernel)
- load_nvfp4_fused_gate now creates Nvfp4Linear from BF16 weight

2026-06-01 11:25:50 +00:00

..

cache

E1: Wire LayerCacheHandle gather methods + CUDA gather kernels

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

kernels

Switch router to Nvfp4Linear production GEMM (custom CuTeDSL kernel crashes MLIR)

2026-06-01 11:17:54 +00:00

layers

NVFP4 router gate: use Nvfp4Linear for both checkpoint and quantized paths

2026-06-01 11:25:50 +00:00

loader

Restructure: cutedsl/ -> dsv4/ with proper layering

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

model

Fix remaining mHC API references: layer_compare.py, layer.py comment

2026-05-31 18:38:34 +00:00

ops

fix: clamp block_amax to E4M3 max (448) in quantize_activation_nvfp4 — prevents NaN from overflow

2026-06-01 04:59:06 +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