biondizzle f259d63930 CRITICAL FIX: SE swizzled buffers were allocated then overwritten with None — graph capture would fall through to broken Python path

2026-06-06 07:01:52 +00:00

..

__init__.py

Restructure: cutedsl/ -> dsv4/ with proper layering

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

grouped_linear.py

Restore A/B split + gsa scalar fix (error is pre-existing, not regression)

2026-06-04 01:03:36 +00:00

linear.py

Add safety check for swizzled buffers: fall through to Python path if None

2026-06-04 04:32:00 +00:00

mhc.py

CUDA graph: Fix per-step allocations in decode loop

2026-06-03 16:38:35 +00:00

moe.py

Fix NameError: add rows/cols variables to MoE swizzle

2026-06-04 03:14:27 +00:00

router.py

Fix dense router BF16 dispatch: use torch.matmul instead of F.linear

2026-06-04 05:58:24 +00:00

shared_expert.py

CRITICAL FIX: SE swizzled buffers were allocated then overwritten with None — graph capture would fall through to broken Python path

2026-06-06 07:01:52 +00:00