biondizzle b7c7e9fb50 refactor: clean up slot_token handling in cutlass_grouped_nvfp4_gemm ...

- Split provided_slot_token vs slot_token_out (returned to caller)
- No gather when slot_token=None (L2 path), no unnecessary alloc
- .contiguous() on gathered tensors for CUTLASS alignment
- Return slot_token_out consistently

2026-05-15 10:11:40 +00:00

..

cutlass_nvfp4_gemm

refactor: clean up slot_token handling in cutlass_grouped_nvfp4_gemm

2026-05-15 10:11:40 +00:00

__init__.py

cleanup: remove abandoned TileLang and Mojo files

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

nvfp4_mega_moe.py

fix: gather on slot_token presence, add shape asserts L1→L2

2026-05-15 10:06:07 +00:00

symm_buffer.py

Plumb global scale as GEMM alpha instead of folding into UE4M3

2026-05-15 03:32:19 +00:00

weight_transform.py

were getting close

2026-05-15 08:28:40 +00:00