biondizzle 48b5b2b702 fix: TMA dimensions for packed FP4 must be in individual FP4 values (not bytes) ...

CUDA docs: 'Dimension for the packed data types must reflect the number
of individual U# values.' For 16U4_ALIGN8B, gmem/smem inner dims must be
FP4 value counts, not byte counts. Double the byte-oriented dimensions
passed by callers. gmem_outer_stride stays in bytes.

2026-05-12 17:39:07 +00:00

..

apis

fix: add kInt8 dtype support to TMA descriptor + change activation tensors to kInt8

2026-05-11 22:54:47 +00:00

indexing

[Public release 26/04] Introducing Mega MoE, FP4 Indexer and other features/fixes (#304)

2026-04-17 09:45:14 +08:00

jit

fix: gran_k=16 in transform_sf + sm_100a arch for NVFP4 mega_moe

2026-05-11 16:11:11 +00:00

jit_kernels

fix: TMA dimensions for packed FP4 must be in individual FP4 values (not bytes)

2026-05-12 17:39:07 +00:00

utils

[Public release 26/04] Introducing Mega MoE, FP4 Indexer and other features/fixes (#304)

2026-04-17 09:45:14 +08:00

python_api.cpp

feat: NVFP4 mega MoE kernel (scale_vec::4X, UE4M3 block scales)

2026-05-11 05:41:08 +00:00