nvfp4-megamoe-kernel

Files

biondizzle dfd9c10ae9 Fix MHC import: don't import .torch from layers/mhc.py

The layers/mhc.py was trying to import kernels.mhc.torch which
failed because our __init__.py was breaking the package. Instead,
just import our mhc_torch_ops which has everything we need.

Also fix __init__.py to explicitly import mhc_pre_torch and
mhc_post_torch from .torch instead of using import *.

2026-05-19 05:36:35 +00:00

fused_moe

Keep MoE scale tensors: framework warmup needs them

2026-05-19 04:50:31 +00:00

kernels

Fix MHC custom op registration

2026-05-19 05:19:48 +00:00

layers

Fix MHC import: don't import .torch from layers/mhc.py

2026-05-19 05:36:35 +00:00

deepseek_v4_attention.py

Fix BF16 wo_a: per-group BMM instead of flat linear