biondizzle 1c15dadaa5 cleanup: remove dead _pack_ue4m3_to_uint32, fix data format docs ...

weight_transform.py returns float8_e4m3fn scales, NOT packed uint32.
The _pack_ue4m3_to_uint32 function was never called. Removed it.
Updated README data formats to accurately reflect the pipeline:
- Weight scales: float8_e4m3fn (direct to CUTLASS, no unpack)
- Activation scales: uint32 packed (from staging kernel, unpacked to float8)

2026-05-14 17:28:12 +00:00

..

cutlass_nvfp4_gemm

cleanup: remove all debug printfs from CUDA kernel and weight_transform

2026-05-14 16:57:32 +00:00

__init__.py

cleanup: remove abandoned TileLang and Mojo files

2026-05-14 12:44:47 +00:00

nvfp4_mega_moe.py

fix: unpack_ue4m3_u32 — uint32 lacks CUDA bitwise ops, use int32

2026-05-14 13:44:42 +00:00

symm_buffer.py

Initial: TileLang NVFP4 mega_moe kernel package

2026-05-13 15:44:51 +00:00

weight_transform.py

cleanup: remove dead _pack_ue4m3_to_uint32, fix data format docs

2026-05-14 17:28:12 +00:00