biondizzle f023b3b2c6 Fix: wrap dummy MoE weights in nn.Parameter ...

PyTorch requires module attributes to be nn.Parameter or None.
torch.empty can't be assigned to a registered parameter slot.

2026-05-19 04:21:35 +00:00

..

kernels/linear/nvfp4

Replace autograd.Function with torch.library.custom_op for Dynamo compat

2026-05-19 01:54:48 +00:00

patches

Fix: wrap dummy MoE weights in nn.Parameter

2026-05-19 04:21:35 +00:00

cutedsl_quant_method.py

Replace autograd.Function with torch.library.custom_op for Dynamo compat

2026-05-19 01:54:48 +00:00

nvfp4_cutedsl.py

Replace autograd.Function with torch.library.custom_op for Dynamo compat

2026-05-19 01:54:48 +00:00